Handbook of Applied Health Economics

in Vaccines
 HA ND BO O KS IN HEA LTH ECONOMICS
EVA LUAT IO N SER IES
Series editors: Alastair Gray and Andrew Briggs
Existing volumes in the series:
Decision Modelling for Health Economic Evaluation
Andrew Briggs, Mark Sculpher, and Karl Claxton
Applied Methods of Cost-​effectiveness Analysis in Healthcare
Alastair M. Gray, Philip M. Clarke, Jane L. Wolstenholme,
and Sarah Wordsworth
Applied Methods of Cost-​Benefit Analysis in Health Care
Emma McIntosh, Philip Clarke, Emma Frew, and Jordan Louviere
Economic Evaluation in Clinical Trials 2e
Henry A. Glick, Jalpa A. Doshi, Seema S. Sonnad, and Daniel Polsky
Applied Health Economics for Public Health Practice and Research
Rhiannon Tudor Edwards and Emma McIntosh
Distributional Cost-​Effectiveness Analysis
Richard Cookson, Susan Griffin, Ole F. Norheim, and Anthony J. Culyer
Handbook of Applied
Health Economics
in Vaccines
Edited by

David Bishai, MD, MPH, PhD

Logan Brenzel, PhD
and
William V. Padula, PhD
 Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Oxford University Press 2023
The moral rights of the authors have been asserted
First Edition published in 2023
Impression: 1
It is permitted to copy, distribute, transmit, and adapt this work, including for commercial purposes,
provided that appropriate credit is given to the creator and copyright holder, a link is provided to
the licence, and any changes made to the work are properly indicated. You may do so in any reasonable
manner, but not in any way that suggests the licensor endorses you or your use.

This is an open access publication, available online and distributed under the terms of a Creative
Commons Attribution 4.0 International licence (CC BY 4.0), a copy of which is available
at http://creativecommons.org/licenses/by/4.0/.
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2022936459
ISBN 978–0–19–289608–7
DOI: 10.1093/​oso/​9780192896087.001.0001
Printed in the UK by
Ashford Colour Press Ltd, Gosport, Hampshire
Oxford University Press makes no representation, express or implied, that the
drug dosages in this book are correct. Readers must therefore always check
the product information and clinical procedures with the most up-​to-​date
published product information and data sheets provided by the manufacturers
and the most recent codes of conduct and safety regulations. The authors and
the publishers do not accept responsibility or legal liability for any errors in the
text or for the misuse or misapplication of material in this work. Except where
otherwise stated, drug dosages and recommendations are for the non-​pregnant
adult who is not breast-​feeding
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Series preface

Economic evaluation in healthcare is a thriving international acti­vity that is

increasingly used to allocate scarce health resources, and within which ap­plied
and methodological research, teaching, and publication are flourishing.
Several widely respected texts are already well established in the market, so
what is the rationale for not just one more book, but for a series? We believe
that the books in the series Handbooks in Health Economic Evaluation share
a strong distinguishing feature, which is to cover as much as possible of this
broad field with a much stronger practical flavor than existing texts, using
plenty of illustrative material and worked examples. We hope that readers
will use this series not only for authoritative views on the current practice of
economic evaluation and likely future developments, but for practical and
detailed guidance on how to undertake an analysis. The books in the series are
textbooks, but first and foremost they are handbooks.
Our conviction that there is a place for the series has been nurtured by
the continuing success of two short courses we helped develop—Advanced
Methods of Cost-Effectiveness Analysis, and Advanced Modelling Methods
for Economic Evaluation. Advanced Methods was developed in Oxford in
1999 and has run several times a year ever since, in Oxford, Canberra, and
Hong Kong. Advanced Modelling was developed in York and Oxford in 2002
and has also run several times a year ever since, in Oxford, York, Glasgow, and
Toronto. Both courses were explicitly designed to provide computer-based
teaching that would take participants through the theory but also the methods
and practical steps required to undertake a robust economic evaluation or
construct a decision-analytic model to current standards. The proof-of-
concept was the strong international demand for the courses—from academic
researchers, government agencies, and the pharmaceutical industry—and the
very positive feedback on their practical orientation.
So the original concept of the Handbooks series, as well as many of the
specific ideas and illustrative material, can be traced to these courses. The
Advanced Modelling course is in the phenotype of the first book in the series,
Decision Modelling for Health Economic Evaluation, which focuses on the role
and methods of decision analysis in economic evaluation. The Advanced
Methods course has been an equally important influence on Applied Methods of
vi Series preface

Cost-Effectiveness, the third book in the series which sets out the key elements
of analyzing costs and outcomes, calculating cost-effectiveness, and reporting
results. The concept was then extended to cover several other important topic
areas. First, the design, conduct, and analysis of economic evaluations along-
side clinical trials have become a specialized area of activity with distinctive
methodological and practical issues, and its own debates and controversies. It
seemed worthy of a dedicated volume, hence the second book in the series,
Economic Evaluation in Clinical Trials. Next, while the use of cost–benefit
analysis in healthcare has spawned a substantial literature, this is mostly
theoretical, polemical, or focused on specific issues such as willingness to
pay. We believe the fourth book in the series, Applied Methods of Cost-Benefit
Analysis in Health Care, fills an important gap in the literature by providing a com-
prehensive guide to the theory but also the practical conduct of cost–benefit
analysis, again with copious illustrative material and worked out examples.
Each book in the series is an integrated text prepared by several contributing
authors, widely drawn from academic centers in the United Kingdom, the
United States, Australia, and elsewhere. Part of our role as editors has been to
foster a consistent style, but not to try to impose any particular line: that would
have been unwelcome and also unwise amidst the diversity of an evolving field.
News and information about the series, as well as supplementary material
for each book, can be found at the series website: <http://www.herc.ox.ac.
uk/books>.
Alastair Gray
Oxford
Andrew Briggs
Glasgow
Foreword

Vaccination to prevent contagious diseases resulted from a happy accident—​

the observation by British physician and scientist Edward Jenner that milk-
maids were essentially immune from smallpox, and (subsequently) that being
infected by the (relatively mild) cowpox disease conferred immunity against
smallpox. Our word “vaccination” derives from the Latin word for “cow”
(vacca). Smallpox was an incredibly deadly disease, killing 300 million people
in the 20th century, and 400 million in the last 100 years before it was eradi-
cated near the end of the 20th century. By inventing vaccines, Jenner is said to
have saved more lives than any other person in human history.
Since then, the science of vaccines has moved through numerous important
steps. For over a century, vaccines were made by creating weakened or inacti-
vated pathogens. The famous Salk and Sabin polio vaccines used this same
basic approach in the 1950s. Addition of adjuvants strengthened the immune
response created by vaccines. New coronavirus disease 2019 (COVID-​19)
vaccines—​developed at an unprecedented speed and with outstanding pro-
tection from the disease—​instead use “messenger RNA” (mRNA) to teach a
body’s cells how to make a protein (or even a piece of one) to create an im-
mune response without actually infecting the subject. This technique offers
considerable promise for future vaccines.
A key feature of vaccines is their role in creating “herd immunity.” In simple
terms, herd immunity arises when the percent of the population who become
immune (either from vaccines or by surviving an infection) exceeds (R0 –​1) /​R0,
where R0 (called the “reproduction rate”) is the number of persons natu-
rally infected by a newly infected person. Some pathogens have a very low
R0 so vaccination is often not necessary to limit outbreaks. Others have very
high contagion levels. Measles, for example, has R0 =​12 to 18, so to prevent
spreading of measles requires (using the midpoint) that 14/​15/​=​93.3% of the
population must either become immunized by vaccination or survive a nat-
ural infection to prevent the disease from spreading. COVID-​19 has an R0 of
about 2.0–​3.0 depending on the variant, thus requiring effective vaccination
coverage (combined with disease-​based immunity) of about two-​thirds of
any self-​contained population. The number needed to be vaccinated interacts
with the vaccine efficacy (the rate of protection provided) in obvious ways.
viii Foreword

This raises another key feature of vaccines—​they are “public goods” in the
sense that all people receiving a vaccination not only protect themselves but
also confer a small benefit on the entire remaining population. This benefit is
obviously larger as the R0 of the pathogen increases. Standard economic anal-
ysis (Phelps, 2017) shows that private incentives to become vaccinated lead
to vaccination rates that are too low, so public policy interventions can be-
come necessary to reach optimal levels of vaccination coverage in any given
population.1
Multiple issues can reduce vaccine uptake. Things that deter vaccination
coverage include painful or health-​risking side effects, the necessity of mul-
tiple shots to achieve full immunity, and the mode of administration (in de-
scending order of preference, oral, intramuscular injection, and intravenous
injection). Apparently simple issues can also confound distribution through
the supply chain, including the “cold chain” requirements for storage from
manufacturing up to the point of final administration (both temperature and
volume of space), and even requisite shelf space for storage of supplies.
In the production process itself, supply chain availability of key compo-
nents can rate-​limit production, as can the simple issue of availability of
glass vials of appropriate size and characteristics, and even the availability
of needles to give injections. Complete consideration of these issues requires
a comprehensive systems analysis review of all facets of vaccine produc-
tion, distribution, financial, and logistics issues that can deter patients’ ac-
cess to vaccines, and information campaigns (Madhavan, Phelps, Rouse, &
Rappuoli, 2018).
In addition to their primary health effects, vaccines can have profound ec-
onomic implications that extend far beyond avoided healthcare costs. Worker
productivity rises when contagious diseases are suppressed. Particularly in
areas where endemic diseases such as malaria exist, school participation and
final educational attainment suffer, so vaccines that either prevent the disease
or reduce disease severity can lead to long-​term economic gains from im-
proved education and higher final attainment levels. These will increase future
worker productivity, make for a more informed electorate, and even reduce
the rate at which people undertake harmful consumption choices (tobacco,
alcohol abuse, lack of exercise, and obesity) (Phelps, 2010).

1 The “cost” of vaccination can be monetary, physical, or psychological, and may be based on misinfor-

mation. In rural and lower-​income areas, travel costs to receive second and third shots may reduce vacci-
nation rates in a way similar to the effect of monetary fees. Fear of physical pain or other adverse reaction
also inhibits vaccination acceptance. Sometimes, misinformation deters vaccination acceptance, such as in
individuals who believe the now-​refuted concept that vaccine adjuvants lead to autism in children.
 Foreword ix

The worldwide COVID-​19 pandemic highlighted another important eco-

nomic consequence of vaccine development. The economies of most nations
of the world came to a near standstill when the only available public interven-
tions were “shelter in place” and quarantine of those either actually or poten-
tially affected. Only the emerging hope of vaccination success brought these
economies back toward full employment, which can only be achieved when
worldwide vaccination rates reach necessary levels.
The first chapter of this handbook explores these and other related is-
sues relating to vaccine production and use. Chapter 2 goes into detail about
methods to evaluate vaccines using appropriate methods of estimating costs,
while Chapters 3 and 4 provide important details about proper methods of
evaluation using up-​to-​date methods of cost-​effectiveness analysis. An ap-
plication of these methods to evaluate the cost-​effectiveness of the vaccines
against COVID-​19 is presented in Chapter 4. Chapter 5 covers the global
landscape for immunization financing, providing insights about the structure
of key global initiatives to expand and sustain immunization programs and on
how the different actors within them interact.
Vaccines are some of the greatest miracles of medical science. Those who
invent them, finance them, distribute them, promote them, administer them,
and, yes, those who receive them are advancing human well-​being. This hand-
book aims at assisting those undertaking economic evaluation of various vac-
cine programs to sharpen their skills, increase their credibility, and through
their work, hopefully, help to focus vaccine development and administration on
those diseases where vaccines matter most.

Charles E. Phelps

References
Madhavan, G., Phelps, C. E., Rouse, W. B., & Rappuoli, R. (2018). Vision for a systems ar-
chitecture to integrate and transform population health. Proceedings of the National
Academy of Sciences of the United States of America, 115(50), 12595–​12602. doi:10.1073/​
pnas.1809919115
Phelps, C. E. (2010). Eight questions you should ask about our health care system (even if the an-
swers make you sick). Stanford, CA: Hoover Institution Press.
Phelps, C. E. (2017). Externalities in health and medical care. In Health economics (6th ed., pp.
387–​411). New York, NY: Routledge Press.
Contents

Introduction to the handbook xv

David Bishai
Acknowledgments xvii
List of abbreviations xix
Contributors xxi

1 Principles of vaccine economics 1

Edited by David Bishai and Chrispus Mayora
1.0 Section introduction: principles of vaccine economics 3
David Bishai and Chrispus Mayora
1.1 Introduction to global vaccine systems 5
Gatien de Broucker
1.2 Relevance of health economics to vaccines 16
David Bishai and Chrispus Mayora
1.3 Cost of finding and making vaccines: implications for
immunization programs 27
Clarke B. Cole, Beth Evans, and Soleine Scotney
1.4 Vaccination as investment in human capital 47
J. P. Sevilla, David Bloom, Dan Salmon, and David Bishai
1.5 Economics of vaccine delivery 67
George Pariyo and Onaopemipo Abiodun
2 Estimating the cost of immunization services 81
Edited by Logan Brenzel
2.0 Section introduction: estimating the cost of
immunization services 83
Logan Brenzel
2.1 Why costing studies are needed 85
Ann Levin, Stephen Resch, and Logan Brenzel
2.2 Defining immunization costs 95
Logan Brenzel
2.3 Designing a primary costing study or analysis 103
Ijeoma Edoka, Stephen Resch, and Logan Brenzel
xii Contents

2.4 Data analysis 115

Stephen Resch and Logan Brenzel
2.5 Costing new vaccine introduction 124
Susmita Chatterjee, Siriporn Pooripussarakul,
and Logan Brenzel
3 Economic evaluation of vaccines and vaccine programs 135
Edited by William V. Padula, Emmanuel F. Drabo,
and Ijeoma Edoka
3.0 Section introduction: economic evaluation of vaccines
and vaccine programs 137
William V. Padula, Emmanuel F. Drabo, and Ijeoma Edoka
3.1 Overview of decision analysis and cost-​effectiveness 140
Emmanuel F. Drabo, Ijeoma Edoka, and William V. Padula
3.2 Defining the scope and study design of cost-​effectiveness
analysis 156
Joseph F. Levy and Charles E. Phelps
3.3 Parameter estimation 167
Emmanuel F. Drabo and David W. Dowdy
3.4 Measuring and valuing health outcomes 198
Y. Natalia Alfonso, Stéphane Verguet, and Ankur Pandya
3.5 Reporting and interpreting results of economic
evaluation 213
Ijeoma Edoka, Carleigh Krubiner, Andrew Mirelman,
R. Brett McQueen, Mark Sculpher, Julia F. Slejko,
and Tommy Wilkinson
3.6 Budget impact analysis and return on investment 239
Elizabeth Watts
3.7 Introduction to decision tree modeling 246
William V. Padula
4 Advanced methods in economic evaluation 259
Edited by William V. Padula, Emmanuel F. Drabo,
and Ijeoma Edoka
4.0 Section introduction: advanced methods in economic
evaluation 261
Emmanuel F. Drabo and William V. Padula
4.1 Introduction to Markov modeling 264
Emmanuel F. Drabo and William V. Padula
 Contents xiii

4.2 Static and dynamic modeling 279

Ann Levin and Colleen Burgess
4.3 Probabilistic sensitivity analysis and value of
information analysis 290
Ciaran N. Kohli-​Lynch
4.4 Economic evaluation reference case with Markov model 310
William V. Padula, Shreena Malaviya, Natalie M. Reid,
Jonothan Tierce, and G. Caleb Alexander
5 Financing and resource tracking of vaccination programs 329
Edited by Logan Brenzel and Shreena Malaviya
5.0 Section introduction: financing and resource tracking of
vaccination programs 331
Logan Brenzel and Shreena Malaviya
5.1 Introduction to immunization financing and expenditure 332
Logan Brenzel and Shreena Malaviya
5.2 Financing of immunization programs 340
Logan Brenzel
5.3 Donor architecture for immunization financing 351
Grace Chee, George Pariyo, and Shreena Malaviya

Appendix 1. Exercise: a case study on estimating the total and

unit routine immunization costs from the facility to the
national level 365
Ijeoma Edoka
Appendix 2. Exercise: estimating new vaccine introduction costs 371
Susmita Chatterjee
Appendix 3. Immunization activities and line item costs 375
Logan Brenzel
Appendix 4. Markov decision processes 379
Emmanuel F. Drabo and William V. Padula
Appendix 5. Derivation of the annual expected costs associated
with the Infected state 381
Emmanuel F. Drabo and William V. Padula
Appendix 6. Making models probabilistic and estimating the
value of information 383
Ciaran N. Kohli-​Lynch
Appendix 7. Decision model 395
Index 397
Introduction to the handbook
David Bishai

Complexity surrounding the development, production, distribution, storage,

injection, and surveillance of vaccines has fascinated economists for good
reason. Much can go wrong. Much can go right. Triumph can save millions of
lives and billions of dollars. But failures can multiply out of control and cost
jobs, reputations, and set back progress for decades.
Economists are also drawn to the study of vaccines because they are ex-
quisite economic products. They are supremely valuable to societies yet often
scarce and neglected except in a crisis. Because they prevent something from
happening, their invisibility creates a constant need to create and spread in-
formation about their value. The invisibility demands transparent and metic-
ulous models of the value of vaccines so that the right choices can be made by
myriad stakeholders.
Vaccine stakeholders include literally every human being. From the day of
birth until the end of life, there is always a choice to be made about whether
to receive a vaccine and how valuable it will be. From newborns getting BCG
shots in the nursery to hospice patients deciding on a COVID-​19 vaccine,
these products are inescapable. Most choices about vaccines occur in an
information-​scarce environment. Price signals that could guide an efficient
choice are seldom functional because public subsidies abound. Information
about the benefits and risks of a shot change over time as epidemics wax and
wane and safety data emerge. The stakeholders responsible for financing
the subsidies or setting prices are forced to make vital decisions that proxy
what fully informed people would pay, but there are never any fully informed
people anywhere. The authors and editors of this handbook were drawn to
vaccine economics because of both its importance and the appeal of con-
necting models of economic value to life-​saving decisions.
Many of the chapters in this handbook had their genesis in classroom ses-
sions with policymakers and practitioners who needed to apply economic
tools to their work in immunization programs and health systems. In 2017,

David Bishai, Introduction to the handbook In: Handbook of Applied Health Economics in Vaccines. Edited by:
David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.001.0001
xvi Introduction to the handbook

with a grant from the Bill & Melinda Gates Foundation, a consortium known
as Teaching Vaccine Economics Everywhere (TVEE) started with faculty
from Johns Hopkins University, Aga Khan University, Indian Institute of
Hospital Management Research University, Makerere University, and
Witwatersrand University. (The University of Ouagadougou and Mahidol
University joined in 2019.) The goal of TVEE was to prepare and deliver a
curriculum in vaccine economics that stretched from introductory material
to advanced methods with an audience ranging from policymakers and prac-
titioners to economics graduate students. After several workshops to de-
velop outlines of the necessary fundamentals in the field, the curriculum
was organized around modules on economic principles, costing, economic
evaluation, financing, and resource tracking. Courses were co-​taught live
in university settings and online with slides and videos available in French
and English. Many of the participants in these courses were practitioners so
there was a focus on immediately applying principles to problems. The ex-
ercises accompanying this handbook have undergone extensive classroom-​
based refinement.
This handbook goes beyond the original classroom material by including
material from leaders in the field to fill in essential areas and connect readers
to emerging consensus in the areas of vaccine costing, evaluation, and guid-
ance. The economics lessons learned during the COVID-​19 pandemic are still
emerging, but the authors have incorporated them whenever possible.
If nothing else, the ongoing struggle to solve the economic problems sur-
rounding the deployment of COVID-​19 vaccines will stimulate many more
readers and practitioners to consider the economics of vaccines. This is a
beautiful field and promises life-​changing rewards.
Acknowledgments

Bishai, Brenzel, and Padula wish to thank the countless individuals who
have studied vaccine economics through the Teaching Vaccine Economics
Everywhere (TVEE) program. Faculty and workshop participants in Burkina
Faso, India, Pakistan, South Africa, Thailand, and Uganda spent weeks dis-
cussing the elements of vaccine economics that were central to both research
and policymaking. Their support of TVEE, and feedback, instilled a sense of
confidence that the content in this handbook could deliver change in vaccine
capacity building throughout countries and communities worldwide.
We would like to acknowledge the supporting roles of Shreena Malaviya,
Gatien de Broucker, and Mandy Chen, whose efforts to manage elements of
the manuscript development from start to finish were critical in its success.
The editors and authors also wish to thank their families, whose daily sup-
port of efforts to develop this manuscript during the midst of the COVID-​19
pandemic was instrumental to completing this work.
Financial support was provided through a grant to the Johns Hopkins
University (INV-​009627). This funding source has made this handbook
openly accessible to individuals seeking to learn more about excellence in vac-
cine economics.
List of abbreviations

AMC Advance Market Commitment

BCR benefit–​cost ratio
BIA budget impact analysis
BMGF Bill & Melinda Gates Foundation
CBA cost–​benefit analysis
CCA cost–​consequence analysis
CEA cost-​effectiveness analysis
CET cost-​effectiveness threshold
CFA cost-​finding analysis
CHEERS Consolidated Health Economic Evaluation Reporting Standards
CI confidence interval
CIA cost-​identification analysis
CMA cost-​minimization analysis
cMYP Comprehensive Multi-​Year Plan
COI cost of illness
COVAX COVID-​19 Vaccines Global Access
COVID-​19 coronavirus disease 2019
CUA cost–​utility analysis
DALY disability-​adjusted life year
DCVM Developing Country Vaccine Manufacturer
DOF Department of Finance
DOH Department of Health
ED-​5Q EuroQol five-​dimensions
EPI Expanded Program on Immunization
EVPI expected value of perfect information
EVPPI expected value of perfect parameter information
EVSI expected value of sample information
FDI Federation Dentaire Internationale
GDP gross domestic product
GNI gross national income
HALY health-​adjusted life year
hib Haemophilus influenzae type b
HPV human papillomavirus
HRQoL health-​related quality of life
HSIS health system and immunization strengthening
ICER incremental cost-​effectiveness ratio
IFFIm International Finance Facility for Immunization
JCVI Joint Committee on Vaccination and Immunisation
xx List of abbreviations

JRF Joint Reporting Form

LMICs low-​and middle-​income countries
MDP Markov decision process
MI4A Market Information for Access to Vaccines
MNC multinational corporation
mRNA messenger RNA
MVP Meningitis Vaccine Project
NHA National Health Accounts
NHB net health benefit
NHI national health insurance
NHIS National Health Insurance Scheme
NIP national immunization program
NITAG National Immunisation Technical Advisory Group
NMB net monetary benefit
NUVI new and underutilized vaccine introduction
OOP out of pocket
PAHO Pan American Health Organization
PCV pneumococcal conjugate vaccine
PFM public financial management
PFP private for-​profit
PNFP private not-​for-​profit
PPS post-​polio syndrome
PSA probabilistic sensitivity analysis
QALY quality-​adjusted life year
QHES Quality of Health Economic Studies
R&D research and development
RCT randomized controlled trial
ROI return on investment
SE standard error
SEIR Susceptible–​Exposed–​Infected–​Recovered
SG standard gamble
SIA supplementary immunization activity
SIR Susceptible–​Infected–​Recovered
SIRV Susceptible–​Infected–​Recovered–​Vaccinated
TTO time trade-​off
UK United Kingdom
UNICEF SD UNICEF Supply Division
US United States
VAS visual analog scale
VFC Vaccines for Children
VPD vaccine-​preventable disease
WHO World Health Organization
Contributors

Editors

David Bishai, MD, MPH, PhD, is Adjunct Professor of Population, Family

and Reproductive Health in the Bloomberg School of Public Health at Johns
Hopkins University in Baltimore, Maryland, USA.

Logan Brenzel, PhD, is Senior Program Officer at the Bill & Melinda Gates
Foundation in the District of Columbia, USA.

William V. Padula, PhD, is Assistant Professor of Pharmaceutical &

Health Economics in the School of Pharmacy at the University of Southern
California, Los Angeles, and Fellow in the Leonard D. Schaeffer Center for
Health Policy & Economics in Los Angeles, California, USA.

Contributors
Onaopemipo Abiodun
PhD Candidate
International Health
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
G. Caleb Alexander
Professor of Epidemiology and Medicine
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
Y. Natalia Alfonso
Health Economist, PhD Candidate
International Health
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
xxii Contributors

David Bishai
Adjunct Professor
Population Family and Reproductive Health
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
David Bloom
Professor
Global Health & Population
Harvard T.H. Chan School of Public Health
Boston, MA, USA
Logan Brenzel
Senior Program Officer
Bill & Melinda Gates Foundation
Seattle, WA, USA
Gatien de Broucker
Senior Health Economist
International Vaccine Access Center, Department of International Health
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
Colleen Burgess
Principal Consultant
Ramboll Health Sciences
Phoenix, AZ, USA
Susmita Chatterjee
Senior Health Economist
George Institute for Global Health
New Delhi, India
Grace Chee
Project Director
MOMENTUM Routine Immunization Transformation and Equity
JSI Research & Training Institute, Inc.
Arlington, VA, USA
Clarke B. Cole
Manager
Non-​Communicable Diseases
Clinton Health Access Initiative
Accra, GH
David W. Dowdy
Associate Professor
Department of Epidemiology
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
 Contributors xxiii

Emmanuel F. Drabo
Assistant Professor
Health Policy and Management
Johns Hopkins University
Baltimore, MD, USA
Ijeoma Edoka
Health Economics and Epidemiology Research Office
Department of Internal Medicine, School of Clinical Medicine,
Faculty of Health Sciences
University of the Witwatersrand
Johannesburg, South Africa
School of Public Health, Faculty of Health Sciences
University of the Witwatersrand
Johannesburg, South Africa
Beth Evans
Program Manager
Global Vaccines Team
Clinton Health Access Initiative
Boston, MA, USA
Ciaran N. Kohli-​Lynch
Research Fellow
Center for Health Services & Outcomes Research
Northwestern University
Chicago, IL, USA
Carleigh Krubiner
Bioethics Lead
Research Environment
Wellcome Trust
London, GB
Ann Levin
President
Levin & Morgan LLC
Bethesda, MD, USA
Joseph F. Levy
Assistant Professor
Department of Health Policy and Management
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
Shreena Malaviya
Senior Health Economist
Purple Squirrel Economics
Toronto, ON, CA
xxiv Contributors

Chrispus Mayora
Lecturer
Health Policy Planning and Management
Makerere University School of Public Health
Kampala, UG
R. Brett McQueen
Assistant Professor
Skaggs School of Pharmacy and Pharmaceutical Sciences
University of Colorado Anschutz Medical Campus
Aurora, CO, USA
Andrew Mirelman
Technical Officer
Health Systems Governance and Financing
World Health Organization
Geneva, CH
William V. Padula
Assistant Professor
Department of Pharmaceutical & Health Economics
University of Southern California
Los Angeles, CA, USA
Ankur Pandya
Associate Professor of Health Decision Science
Health Policy and Management
Harvard T.H. Chan School of Public Health
Boston, MA, USA
George Pariyo
Chief of Operations
Senior Management Team
Serum Africa Medical Research Institute (SAMRI)
Kampala, UG
Charles E. Phelps
Professor and Provost Emeritus
University Professor and Provost Emeritus
Economics and Public Health Sciences
University of Rochester
Rochester, NY, USA
Siriporn Pooripussarakul
Independent Researcher
Bangkok, Thailand
 Contributors xxv

Natalie M. Reid
Director
Monument Analytics
Baltimore, MD, USA
Stephen Resch
Lecturer
Department of Health Policy and Management
Harvard T.H. Chan School of Public Health
Boston, MA, USA
Dan Salmon
Professor
Department of International Health
Johns Hopkins Bloomberg School of Public Health
Baltimore, MD, USA
Soleine Scotney
Country Director
Clinton Health Access Initiative (CHAI) Cambodia
Phnom Penh, KH
Mark Sculpher
Professor and Director
Centre for Health Economics
University of York
York, GB
J. P. Sevilla
Research Associate
Global Health and Population
Harvard T.H. Chan School of Public Health
Boston, MA, USA
Julia F. Slejko
Associate Professor
Department of Pharmaceutical Health Services Research
University of Maryland School of Pharmacy
Baltimore, MD, USA
Jonothan Tierce
Principal
Monument Analytics, Inc.
Baltimore, MD, USA
Stéphane Verguet
Associate Professor of Global Health
Global Health and Population
Harvard T.H. Chan School of Public Health
Boston, MA, USA
xxvi Contributors

Elizabeth Watts
Doctoral Researcher
Health Policy & Management
University of Minnesota
Minneapolis, MN, USA
Tommy Wilkinson
Senior Researcher
Health Economics Unit, School of Public Health
University of Cape Town
Cape Town, ZA
 1
PRINCIPLES OF VACCINE ECONOMICS
Edited by David Bishai and Chrispus Mayora
1.0
Section introduction: principles
of vaccine economics
David Bishai and Chrispus Mayora

Fundamentals matter. Vaccines are some of the most complex molecules ever
invented and the social systems that deploy, monitor, and finance them are
equally complex. Much of the peculiarities of vaccine economics arises from
the peculiarities of vaccines. The most eccentric thing about vaccines is that
many of them can benefit the person who gets the shot as well as the people
around them. Furthermore, not everyone will obtain the same amount of ben-
efit because of varying risk levels in the population. Coronavirus disease 2019
(COVID-​19) vaccines illustrate these points well. The massive public invest-
ments to subsidize supply of and stimulate demand for COVID-​19 vaccines
were predicated on a sure forecast that without public investment the free
market was doomed to fail. In many countries, even with the public sector in-
vestments, a COVID-​19 vaccine at a price of $0 was not attractive to many who
were at lower risk and did not appreciate the benefits their shot would offer
their community.
There are also many technical peculiarities of vaccines such as their produc-
tion and delivery that have economic policy implications, which will be the
focus of Chapter 1.1. The need to closely monitor storage, delivery practices,
and wastage as well as the methods of reaching children either through rou-
tine pediatric visits or supplemental immunization activities and campaigns
are key special features of vaccines with implications for costs and financing.
The fundamental elements of economics such as price, demand, and supply
take on a new meaning in the world of vaccines, which are often “free” to
most consumers. Chapter 1.2 adapts these fundamental tools of economics
to highly subsidized goods like vaccines. By increasing supply and demand
while keeping prices low, these tools will be fundamental in understanding
policies to achieve better coverage. With the extensive role of public sector

David Bishai and Chrispus Mayora, Section introduction: principles of vaccine economics In: Handbook of Applied Health
Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0001
4 Section introduction: principles of vaccine economics

subsidies and regulations, the prices charged for vaccines also deviate sub-
stantially from the marginal cost per dose. Chapter 1.3 describes how costs
of vaccines are determined, highlighting the cost differences between newer
vaccines where companies must recover their substantial outlay required for
discovery and traditional older vaccines whose production costs resemble
undifferentiated commodities produced in competitive markets. Specifically,
approaches to procurement of vaccines that keep prices affordable for low-​
income countries, but preserve a robust supply environment will be discussed.
On the demand side of vaccine economics lies the perennial problem of
substantial social benefits in terms of herd immunity that might remain elu-
sive if consumers choose to free ride. This topic is raised in Chapter 1.4 where
the foundations of vaccine hesitancy are rooted in the behavioral economics
violations of rational perception of risks. Fear and dread of the unknowns sur-
rounding either a new vaccine or a new disease can sway rational choices away
from the cold calculation of expected health risks and health benefits. Finally,
we close the chapter with an economic look at how vaccines are delivered to
patients and the potential economies and diseconomies of reaching people in
outreach programs versus in routine clinics (Chapter 1.5).
1.1
Introduction to global vaccine systems
Gatien de Broucker

Developing vaccines requires balancing different characteristics to ensure ef-

fectiveness and safety, from the choice of biotechnology (vaccine type) to its
delivery system(s) to users. This handbook will discuss how these character-
istics all factor into costs, from research and development to manufacturing
and distribution.
The biotechnology must trigger an immune response associated with a spe-
cific pathogen that will persist beyond taking the vaccine. This immune re-
sponse must be able to recognize subsequent exposure to the infectious agent
quickly and curb its replication and pathology. Additionally, the immune re-
sponse upon vaccination must be harmless to the vaccinated individual and
impose negligible risk of adverse effects in the long term.
Beyond provoking an immune response in clinical settings, vaccines must
work effectively in the real world—​in combination with other vaccines and
treatments, and in a variety of climates and settings. Standards are set for
transporting and storing vaccines from production to use using a cold chain,
and for the use of each vaccine in specific populations through vaccination
schedules and recommendations based on medical status. Vaccines are sen-
sitive to environmental and biological factors: variations in temperature,
delay between production and usage, as well as deviations from the vacci-
nation schedule are known shortcomings to their protective effect (Babirye
et al., 2012).
In addition to upholding these standards, immunization programs are de-
signed to maximize vaccine coverage, with the goal of achieving herd immu-
nity where everyone in the population benefits either directly or indirectly
from the protective effect of vaccines. People who do not benefit directly from
vaccine protection for a variety of reasons (e.g., being unvaccinated, having
a weakened immune system, vaccine protection has lost its potency) can still
benefit from others being directly protected and not transmitting the disease.

Gatien de Broucker, Introduction to global vaccine systems In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0002
6 Introduction to global vaccine systems

In this chapter, we briefly introduce the biotechnology behind vaccines

along with the different vaccine types and discuss the strategy and logistics of
immunization programs. Regulatory and monitoring systems are discussed
in Chapter 1.3 of this handbook.

1.1.1 Vaccine types

The vaccine technology encompasses the molecular make-​up of the vaccine,

most often referred to as the vaccine type, and aims to ensure an appropriate
immune response to different pathogens, and their viability beyond the clin-
ical settings. We distinguish three types of vaccines based on the biological
unit used: vaccines containing the whole pathogen, or a subunit of it, or based
on nucleic acid (either DNA or RNA). All types aim to stimulate an immune
response to a specific pathogen. Their development and production costs
differ greatly.
Whole pathogen vaccines include live-​attenuated and inactivated vaccines,
which present, respectively, a weakened or dead pathogen to the immune
system. Recent advances in genetic engineering introduced chimeric vac-
cines, where the genes or proteins of the pathogen of interest are mounted
on a different pathogen. Vaccines against the following pathogens or disease
use this technology: cholera, coronavirus disease 2019 (COVID-​19), Ebola,
hepatitis A, influenza, Japanese encephalitis, measles, mumps, pertussis
(whooping cough), poliomyelitis, rabies, rotavirus, rubella, smallpox, tuber-
culosis, typhoid, varicella (chickenpox), varicella (zoster), and yellow fever.
Subunit vaccines present only the antigen, that is, the part of the pathogen
used to identify it by the immune system. They include polysaccharide, con-
jugate, and recombinant vaccines, using either polysaccharides or proteins
often combined with adjuvants to stimulate an immune response. Other sub-
unit vaccines present products related to the pathogen, such as toxoid vac-
cines, which present the (inactivated) toxin generated by the bacterium rather
than the bacterium itself, and virus-​like particle vaccines. Vaccines against the
following pathogens or disease are subunit vaccines: dengue, diphtheria, hep-
atitis B, Haemophilus influenzae type b, human papillomavirus (HPV), me-
ningococcal meningitis, pertussis, pneumococcal disease, rotavirus, shingles,
tetanus, and typhoid.
Developed more recently, nucleic acid vaccines supply the genetic code
for the body’s cells to produce the antigen and provoke an immune response
using DNA plasmids, messenger RNA (mRNA), or a viral vector. Nucleic acid
 1.1.2 The strategy and logistics of immunization programs 7

vaccines, particularly those using a viral vector, promise to be faster and easier
to develop and produce on a large scale (Pardi, Hogan, Porter, & Weissman,
2018; Siegrist, 2018). The vaccines against COVID-​19 manufactured by Pfizer
and Moderna are the first vaccines using the mRNA technology to be approved
by regulatory agencies. Merck’s vaccine against Ebola uses the viral vector tech-
nology and is approved for use in several countries, including the US (Centers
for Disease Control and Prevention, 2021a; Dolzhikova et al., 2017).

1.1.2 The strategy and logistics of immunization

programs

After manufacturing, vaccines have a finite lifespan, subject to environmental

factors. To minimize wastage and optimize vaccination coverage, govern-
ments manage a large procurement, storage, and delivery system which is part
of the Expanded Program on Immunization (EPI) or National Program on
Immunization, built to sustain routine immunization and any supplemental
immunization activities.
The inclusion/​exclusion of each vaccine in a country’s EPI is determined
partly by the country’s disease burden, but also by a vaccine’s cost to the gov-
ernment, safety, shelf life, and the duration of protective effect. Several vac-
cines prevent diseases that are common to most regions of the world, such as
measles, tetanus, diphtheria, or pertussis. Others are more relevant to specific
areas, such as yellow fever and dengue in equatorial and tropical countries,
and influenza in colder climates. Finally, some vaccines are most useful to pre-
vent outbreaks caused by a combination of environmental and human factors
often brought forward in humanitarian crises, such as the cholera vaccine.
Alternatively, vaccines excluded from the EPI can still be used in a country,
for instance, as a traveler’s vaccine: rabies, yellow fever (in countries where it
is not endemic), and typhoid fever can be provided to travelers visiting coun-
tries where the risk of contracting these diseases is increased.

1.1.2.1 Routine vaccination and supplemental

immunization activities

The choice of strategy for each vaccine, either to vaccinate systematically or

punctually, depends on several factors both endogenous and exogenous to a
country’s EPI.
8 Introduction to global vaccine systems

Ongoing, planned, and contingent on population growth, the systematic

immunization of newborns and infants through routine immunization con-
stitutes the main share of immunization activities in a country. Routine im-
munization tends to be the most efficient at increasing vaccine coverage,
although inequities persist (Chopra et al., 2020). Routine immunization can
leverage (or strengthen) postnatal medical visits for infants, in synergy with
the public health effort in maternal and child health. In high-​income countries,
where older adults are also a target population, routine immunization further
strengthens regular medical check-​ups. However, sustaining adequate vaccine
coverage through routine immunization is subject to supply-​and demand-​side
constraints and is challenging to do. Known constraints include vaccine dose
stockouts and wastage, which often constrict the supply of vaccines to specific
areas in a country. Deficient vaccine confidence (or increased vaccine hesi-
tancy) and missing facility visits for neonatal care are common factors that in-
fluence the demand for vaccines (discussed further in Chapter 1.4).
In situations when the vaccine coverage falls short either nationally or sub-​
nationally, increasing the risk of, or causing, outbreaks, governments may
organize vaccination campaigns or child health days and child health weeks
known as supplementary immunization activities (SIAs), often with the support
of international health agencies. A large amount of healthcare resources can be
redirected towards an SIA for specific vaccines (e.g., measles), aiming to “catch-​
up” and reach those children who are unimmunized. In recent years, the term
“zero-​dose” child was coined to define inequalities in immunization and assess
the impact of SIAs more accurately (Portnoy, Jit, Helleringer, & Verguet, 2018).
While they can be a cost-​effective intervention to prevent major outbreaks, as
is the case for measles (Bishai et al., 2011), SIAs are resource intensive and, in
some circumstances, can undermine the delivery of routine immunization in
resource-​constrained settings (Chakrabarti, Grépin, & Helleringer, 2019).
The international effort to vaccinate against COVID-​19 using vaccines
from different manufacturers qualifies as an SIA. If vaccination continues
beyond the current crisis caused by the pandemic and is added to the vac-
cine schedules, it would become part of routine immunization in a similar
pathway as for the influenza vaccine in the aftermath of the 2009 H1N1 pan-
demic (Centers for Disease Control and Prevention, 2010).

1.1.2.2 Target population

For optimal and lasting immunogenicity, and to account for the age-​specific
morbidity and mortality of diseases, vaccines are typically associated with a
target population.
 1.1.2 The strategy and logistics of immunization programs 9

Newborns and infants are the primary target population for all routine im-
munization schedules as vaccines effectively prevent deadly or debilitating
diseases, which not only would affect the child’s life, but also that of their care-
givers. The routine immunization of children globally provides a very high
value of return-​on-​investment: $21 saved in healthcare cost and produc-
tivity loss averted for every dollar invested in immunization in countries el-
igible for Gavi support over two decades (Sim, Watts, Constenla, Brenzel, &
Patenaude, 2020). Furthermore, infants are also the target population of many
other public health interventions under the scope of neonatal care and hence
more accustomed to regular medical visits. Conversely, conducting SIAs and
strengthening routine immunization for zero-​dose children may also provide
them with these other healthcare services they would not otherwise be ex-
posed to.
Comparatively, adults tend to interact less often with the healthcare
system than children, and to be less inclined to follow their vaccination
schedule for a variety of reasons. The general adult population is a target
population for specific vaccines such as the influenza vaccine (recom-
mended annually), for boosters (e.g., Td or Tdap booster), and for traveler’s
vaccines (when relevant). There are three groups within the adult popula-
tion more specifically targeted for routine immunization: immunocom-
promised adults, older adults (65 years and above), and pregnant women.
Immunocompromised adults, older adults, and pregnant women may re-
ceive additional doses of the vaccines provided during childhood (e.g.,
pneumococcal, meningococcal, and hepatitis vaccines), subject to clinical
advice (Centers for Disease Control and Prevention, 2021b).

1.1.2.3 Vaccine selection

National Immunization Technical Advisory Groups (NITAGs) provide re-

commendations on new vaccine introductions, the choice of product, the
vaccine schedule, and priorities within the country (Gavi, 2021b; Howard,
Walls, Bell, & Mounier-​Jack, 2018; Steffen et al., 2021). Examples of national
level NITAGs in high-​income countries include the Advisory Committee
on Immunization Practices (ACIP) in the US and Joint Committee on
Vaccination and Immunisation (JCVI) in the UK. There are over 170 NITAGs
in countries around the world listed at https://​ www.nitag-​
resou​ rce.org/​
.
At the global level, a group called the Strategic Advisory Group of Experts
on Immunization (SAGE) meets to advise the World Health Organization
(WHO). The WHO then compiles a list of prequalified vaccines to guide
UN agencies’ procurement of vaccines and national governments. Countries
10 Introduction to global vaccine systems

use NITAG recommendations (Howard et al., 2018), or those from the their
Ministry of Health and other agencies when these is no NITAG in place, to
prioritize vaccines that can alleviate the most disease cases—​and, thus, related
healthcare costs—​and deaths, based on their budget and on the availability
of external funding (Steffen et al., 2021). The WHO and Gavi (officially, Gavi,
the Vaccine Alliance, formerly known as the Global Alliance for Vaccines
and Immunization) guide the vaccine selection process in low-​and middle-​
income countries with severe budget constraints, providing them with tech-
nical assistance and a financing plan to afford their introduction and sustain
their use. With shared funding and negotiated vaccine dose prices, Gavi fa-
cilitates the procurement of 12 vaccines for routine immunization and SIAs
(“catch-​up campaigns”): pentavalent, rotavirus, pneumococcal, HPV, inacti-
vated polio, Japanese encephalitis, measles, measles–​rubella, meningitis A,
typhoid, cholera, and yellow fever vaccines (Gavi, 2019, 2021a). These policy
and financing mechanisms are discussed in Chapter 5 of this handbook.
In the last decades, countries (in particular, high-​income countries) also
considered population demand for vaccines, or lack thereof, when making
decisions about vaccine selection. In 2013, the Japanese government intro-
duced the HPV vaccine in its routine immunization program, targeting girls
aged 12–​16 years, aiming to prevent HPV infection and cervical cancer.
This introduction was met with considerable negative news coverage in the
Japanese media, and a growing, but unfounded, reticence from parents for
their daughters to get the vaccine. While the HPV vaccine still technically
appears in the Japanese vaccine schedule, the Ministry of Health suspended
its use indefinitely (Ikeda et al., 2019). In Denmark, the HPV vaccine intro-
duction also met with adverse public opinion. However, while the uptake of
the HPV vaccine significantly decreased, it was still offered through the rou-
tine immunization program (Suppli et al., 2018).

1.1.2.4 Vaccine schedules

Each country’s EPI plans its routine immunization program using a vaccine
schedule. The schedule aligns every vaccine procured by the government for
the EPI with their target population, defined by age and health status (e.g.,
healthy, immunocompromised). As an essential part of primary healthcare,
pediatricians and general physicians use this schedule at every medical visit
to assess the need to vaccinate and plan with patients and caregivers for
follow-​up visits.
 1.1.2 The strategy and logistics of immunization programs 11

Timely vaccination is essential for optimal immunogenicity and pro-

tection, particularly in infants, and to manage supply-​side constraints effi-
ciently. Late vaccination for specific vaccines, such as the measles vaccine,
is considered as a risk factor for outbreaks (Babirye et al., 2012). To manage
such risk, vaccine schedules can also include recommendations on catch-​up
vaccination for missed doses.
Detailed vaccine schedules are available through national and regional
health agencies. Selected examples of vaccine schedules are listed below:

• WHO recommendations for routine immunization (global): http://​www.

who.int/​immun​izat​ion/​pol​icy/​immu​niza​tion​_​tab​les/​en/​.
• US Centers for Disease Control and Prevention: https://​www.cdc.gov/​
vacci​nes/​schedu​les/​index.html.
• European Center for Disease Prevention and Control: https://​vacc​ine-​
sched​ule.ecdc.eur​opa.eu/​.
• National Health Mission of India: http://​www.nrh​mhp.gov.in/​cont​ent/​
immun​izat​ion.
• Immunization schedules for Africa: http://​www.vacfa.uct.ac.za/​immun​
izat​ion-​schedu​les-​afr​ica (hosted by the University of Cape Town, South
Africa).

1.1.2.5 Delivery vehicle and logistics

Among environmental factors, vaccines are most sensitive to changes in

temperature and exposure to sunlight and air. Vaccine packaging, delivery,
and handling are, therefore, carefully planned and monitored. Vaccine
doses are packaged in vials, tubes, or ampoules. Vials typically contain mul-
tiple doses of vaccines, which are meant for several people. Once a vial is
opened, vaccine doses must be used rapidly: within 6 hours or up to 28 days
for select vaccines. Healthcare workers refer to the WHO Multi-​Dose Vial
Policy and description of each vaccine to manage the use of multi-​dose vials
(WHO, 2014).
Before and between uses, vials are stored and transported in temperature-​
controlled containers or refrigerators, shaping the vaccine “cold chain.” Most
vaccines are sensitive to warmer temperatures, hence the need for constant
refrigeration. Malfunctioning or mishandled refrigerators can bring the tem-
perature below the freezing point, which also affects vaccines in liquid form.
Several vaccines can be freeze dried to prevent freeze damage. Vaccine vials
12 Introduction to global vaccine systems

usually feature a vaccine vial monitor, a heat-​sensitive label, which indicates

whether they were exposed to excessive heat (WHO, 2020).
Such constraints lead to wastage, where doses issued to a healthcare fa-
cility or outreach team were not administered to patients (Equation 1.1.1 and
Table 1.1.1) (WHO, 2005). Wastage costs are distinct from waste manage-
ment costs. The latter refers to the costs associated with the management of
used disposable items, such as syringes and gloves. Wastage is inextricable to
vaccine programming: efficient vaccine program management reduces, but
never eliminates, wastage. Wastage rates vary significantly between vaccines,
regions, and immunization programs, and must be appraised to cost and plan
immunization programs properly. For instance, vaccine wastage rates in India
varied between 13.2% (pentavalent vaccine; Pune district) and 50.8% (oral
polio vaccine; Kangra district) in one study (Das et al., 2020).
Equation 1.1.1. Vaccine wastage rate:

Number of doses administered

Vaccine wastage rate = 1 −
Number of doses issued

Assumptions and methods to integrate the wastage rate into economic

models are discussed in Chapter 2.5. Common reasons for wastage are listed
in Table 1.1.1.
Maintaining and operating cold chains and assuming some level of wastage
ensure that the available vaccine doses are in optimal condition to stimulate
an immune response. Using doses subjected to inadequate storage risks af-
fecting the effectiveness of the vaccine. For instance, recognized shortcom-
ings in the measles vaccine effectiveness in Uganda include settings where the
cold chain does not perform adequately (Mupere et al., 2006).

Table 1.1.1 Types of vaccine wastage

Vaccine wastage in unopened vials Vaccine wastage in opened vials

• Expiry In addition to the types listed in the previous

• Vaccine vial monitor indication and column:
heat exposure • Discarding remaining doses at the end
• Freezing of session
• Breakage • Not being able to draw the number of
• Missing inventory doses indicated on the label of a vial
• Theft • Poor reconstitution practices
• Discarding unused vials returned • Submergence of opened vials in water
from an outreach session • Suspected contamination
• Patient reaction requiring more than
one dose
 1.1.3 Conclusion 13

1.1.3 Conclusion

In this chapter, we introduced several considerations associated with devel-

oping and using vaccines. This web of actors, programs, regulations, and con-
ventions connecting global and local policies and constraints form the global
vaccine systems.
Vaccine research and implementation appear as distinct steps enacted in a
strict sequence, separated only, yet firmly, by regulatory approval processes.
However, the considerations to implementation we highlighted hint to the
contrary. Whereas vaccine rollout is guided by the clinicians who created
and tested it, vaccine research is sensitive to rollout constraints and can priori-
tize the development of vaccines that can accommodate different “real-​world”
applications beyond the clinical settings. Companies and researchers may
favor different biotechnologies and vaccine platforms to make the vaccine
highly effective after only one dose was inoculated, or transportable and stor-
able in resource-​constrained settings with a deficient cold chain. For instance,
the Johnson & Johnson COVID-​19 vaccine marketed in 2021 distinguishes
itself from the others as a one-​dose (rather than two-​dose) vaccine—​a strong
advantage for implementation in low-​income settings (in both high-​and low-​
income countries).
Economic evaluations play an essential role in providing evidence of af-
fordability, sustainability, and relevance to NITAGs and governments, but
also in suggesting research priorities for pharmaceutical companies seeking
to expand their product portfolio. Furthermore, more fundamental research
led by economists, and behavioral and social scientists also provides insights
on factors influencing demand, such as vaccine hesitancy.
The study of vaccine economics, in both its applied and fundamental forms,
must accompany progress in vaccine development and rollout so we do not
spoil this effective protection against humanity’s microscopic foe.

References
Babirye, J. N., Engebretsen, I. M. S., Makumbi, F., Fadnes, L. T., Wamani, H., Tylleskar,
T., & Nuwaha, F. (2012). Timeliness of childhood vaccinations in Kampala Uganda: A
community-​based cross-​sectional study. PloS One, 7(4), e35432–​e35432. doi:10.1371/​
journal.pone.0035432
Bishai, D., Johns, B., Nair, D., Nabyonga-​Orem, J., Fiona-​Makmot, B., Simons, E., & Dabbagh,
A. (2011). The cost-​ effectiveness of supplementary immunization activities for mea-
sles: A stochastic model for Uganda. Journal of Infectious Diseases, 204(Suppl 1), S107–​S115.
doi:10.1093/​infdis/​jir131
14 Introduction to global vaccine systems

Centers for Disease Control and Prevention. (2010). Prevention and control of influenza
with vaccines: Recommendations of the Advisory Committee on Immunization Practices
(ACIP), 2010. Morbidity and Mortality Weekly Report, 59(RR08), 1–​62.
Centers for Disease Control and Prevention. (2021a). Ebola vaccine: Information about Ervebo®.
https://​www.cdc.gov/​vhf/​ebola/​cli​nici​ans/​vacc​ine/​index.html
Centers for Disease Control and Prevention. (2021b). Recommended adult immunization
schedule for ages 19 years or older, United States, 2021. https://​www.cdc.gov/​vacci​nes/​schedu​
les/​hcp/​imz/​adult.html
Chakrabarti, A., Grépin, K. A., & Helleringer, S. (2019). The impact of supplementary immu-
nization activities on routine vaccination coverage: An instrumental variable analysis in five
low-​income countries. PloS One, 14(2), e0212049. doi:10.1371/​journal.pone.0212049
Chopra, M., Bhutta, Z., Blanc, D. C., Checchi, F., Gupta, A., Lemango, E. T., . . . Victora, C. G.
(2020). Addressing the persistent inequities in immunization coverage. Bulletin of the World
Health Organization, 98(2), 146–​148. doi:10.2471/​BLT.19.241620
Das, M. K., Sood, M., Tambe, M. P., Sharma, T. D., Parande, M. A. G., Surwade, J. B., . . . Sindhu,
M. (2020). Documentation of vaccine wastage in two different geographic contexts under
the universal immunization program in India. BMC Public Health, 20(1), 556. doi:10.1186/​
s12889-​020-​08637-​1
Dolzhikova, I. V., Tokarskaya, E. A., Dzharullaeva, A. S., Tukhvatulin, A. I., Shcheblyakov, D.
V., Voronina, O. L., . . . Gintsburg, A. L. (2017). Virus-​vectored Ebola vaccines. Acta Naturae,
9(3), 4–​11.
Gavi. (2019). Annual program report. https://​www.gavi.org/​sites/​defa​ult/​files/​pro​gram​mes-​
imp​act/​our-​imp​act/​apr/​Gavi-​Progr​ess-​Rep​ort-​201​9_​1.pdf
Gavi. (2021a). Detailed product profiles. https://​www.gavi.org/​news/​docum​ent-​libr​ary/​detai​
led-​prod​uct-​profi​les
Gavi. (2021b). Gavi vaccine funding guidelines. https://​www.gavi.org/​sites/​defa​ult/​files/​supp​
ort/​Vaccin​e_​Fu​ndin​gGui​deli​nes.pdf
Howard, N., Walls, H., Bell, S., & Mounier-​Jack, S. (2018). The role of National Immunisation
Technical Advisory Groups (NITAGs) in strengthening national vaccine decision-​
making: A comparative case study of Armenia, Ghana, Indonesia, Nigeria, Senegal and
Uganda. Vaccine, 36(37), 5536–​5543. doi:10.1016/​j.vaccine.2018.07.063
Ikeda, S., Ueda, Y., Yagi, A., Matsuzaki, S., Kobayashi, E., Kimura, T., . . . Kudoh, K. (2019). HPV
vaccination in Japan: What is happening in Japan? Expert Review of Vaccines, 18(4), 323–​325.
doi:10.1080/​14760584.2019.1584040
Mupere, E., Karamagi, C., Zirembuzi, G., Grabowsky, M., de Swart, R. L., Nanyunja, M., &
Mayanja, H. (2006). Measles vaccination effectiveness among children under 5 years of age
in Kampala, Uganda. Vaccine, 24(19), 4111–​4115. doi:10.1016/​j.vaccine.2006.02.038
Pardi, N., Hogan, M. J., Porter, F. W., & Weissman, D. (2018). mRNA vaccines—​a new era in
vaccinology. Nature Reviews. Drug Discovery, 17(4), 261–​279. doi:10.1038/​nrd.2017.243
Portnoy, A., Jit, M., Helleringer, S., & Verguet, S. (2018). Impact of measles supplementary
immunization activities on reaching children missed by routine programs. Vaccine, 36(1),
170–​178. doi:10.1016/​j.vaccine.2017.10.080
Siegrist, C.-​A. (2018). Vaccine immunology. In S. A. Plotkin, W. A. Orenstein, & P. A. Offit
(Eds.), Plotkin’s vaccines (7th ed., pp. 16–​34). Philadelphia, PA: Elsevier.
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353.
Steffen, C. A., Henaff, L., Durupt, A., Omeiri, N. E., Ndiaye, S., Batmunkh, N., . . . Hombach, J.
(2021). Evidence-​informed vaccination decision-​making in countries: Progress, challenges
and opportunities. Vaccine, 39(15), 2146–​2152. doi:10.1016/​j.vaccine.2021.02.055
 1.1.3 Conclusion 15

Suppli, C. H., Hansen, N. D., Rasmussen, M., Valentiner-​Branth, P., Krause, T. G., & Mølbak,
K. (2018). Decline in HPV-​vaccination uptake in Denmark—​the association between HPV-​
related media coverage and HPV-​vaccination. BMC Public Health, 18(1), 1360. doi:10.1186/​
s12889-​018-​6268-​x
World Health Organization. (2005). Monitoring vaccine wastage at country level: Guidelines
for programme managers (WHO/​V&B/​03.18/​Rev.1). https://​apps.who.int/​iris/​bitstr​eam/​han​
dle/​10665/​68463/​WHO_​VB​_​03.18.Rev.1_​eng.pdf
World Health Organization. (2014). WHO policy statement: Multi-​dose vial policy (MDVP).
https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​135​972/​WHO​_​IVB​_​14.07_​eng.pdf
World Health Organization. (2020). What is VVM and how does it work? https://​www.who.
int/​imm​uniz​atio​n_​st​anda​rds/​vacc​ine_​qual​ity/​What%20is%20VVM%20and%20how%20d​
oes%20it%20w​ork.pdf
1.2
Relevance of health economics
to vaccines
David Bishai and Chrispus Mayora

Economics informs problems of choice and allocation when there is uncer-

tainty and scarcity. The world of vaccines and vaccinations abounds with
these dilemmas. There are many choices, few resources, and never enough in-
formation. Enter vaccine economics.
This chapter offers an introduction to the multiple ways that economics
can be applied to systems tasked with discovering, financing, distributing,
tracking, and administering the world’s vaccines. It sets the stage for deeper
coverage of these topics in subsequent chapters.
The case of vaccines for coronavirus disease 2019 (COVID-​19) starkly illus-
trates the many applications of vaccine economics. As the epidemic emerged,
securing safe and effective vaccines became a priority for politicians, busi-
ness leaders, and public health officials. Speed was paramount. The past pace
of vaccine discovery based on private pharmaceutical companies financed
mostly by private investors would be too slow. Relying solely on private rev-
enue streams to incentivize research and pay for production would be too in-
equitable. Scientific uncertainty clouded whether a safe and effective vaccine
against SARS-​CoV-​2 virus could even be developed.
Even after safe and effective vaccines emerged, the world’s systems for allo-
cation of vaccines between countries and people in countries faced challenges
in overcoming inequalities, misinformation, and misallocation. Guidelines
for prioritization of vaccine allocation to countries and subpopulations in
countries appropriately stressed a combination of moral considerations be-
sides just economic efficiency. However, the human systems tasked with fol-
lowing these guidelines when they clashed with economic self-​interest broke
down. Data systems to track gaps in vaccine distribution and available human
and physical resources relevant to vaccines were wanting.

David Bishai and Chrispus Mayora, Relevance of health economics to vaccines In: Handbook of Applied Health Economics in
Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0003
 Relevance of health economics to vaccines 17

One answer to the predictable barriers to efficient and fair vaccine al-
location is to create institutions designed to break through these barriers.
Governments instituted COVID-​19 vaccine priority groupings internally and
the World Health Organization (WHO) developed global guidelines to sup-
port countries in their allocation of COVID-​19 vaccines (WHO, 2020, 2021).
One example is the COVID-​19 Vaccines Global Access (COVAX) Facility,
led by WHO, the Coalition for Epidemic Preparedness Innovations (CEPI),
Gavi, and UNICEF. Within COVAX, the global collaboration of the Access
to COVID Tools Accelerator (ACT) includes government and global health
and civil society organizations such as CEPI, Gavi, the Global Fund, WHO,
UNICEF, the Pan American Health Organization (PAHO), and the World
Bank. COVAX maintains a pooled fund for an advance market commitment
wing where donor countries can support access to COVID-​19 vaccines on be-
half of low-​income countries.
Vaccine economics was able to shed light and offer guidance in many of
the areas highlighted in the case of COVID-​19 vaccines. In the area of finance
for vaccine discovery, years of work in developing alternative financial incen-
tives for pharmaceutical discovery were at hand (Kickbusch, Krech, Franz,
& Wells, 2018; Kremer, Levin, & Snyder, 2020; Yamey et al., 2020). Advance
market commitments are public sector promises to guarantee future demand
and help assure private companies that a successful product will generate
the revenue to pay off prior investments and maintain production. In addi-
tion, governments and groups of governments became the largest investors in
private firms seeking to discover COVID-​19 vaccines. As investors, govern-
ments had to manage various types of uncertainty in assembling a portfolio
of candidate vaccines to invest in (McDonnell et al., 2020; Shnaydman, 2020).
Multigovernmental purchasing cooperatives helped to pool buying power
of middle-​and lower-​income countries through COVAX (Kuehn, 2020).
Frameworks for how to distribute the vaccine fairly while achieving goals
of controlling morbidity, mortality, and negative societal impact were ap-
proached by applying both ethical and economic principles (Grauer, Löwen,
& Liebchen, 2020; National Academies of Sciences, Engineering, & Medicine,
2020). The logistics for getting COVID-​19 vaccines from factories to ware-
houses to vaccination sites disclosed long-​standing weaknesses in vaccine
distribution systems. In the public–​private partnerships tasked to distribute
and deliver vaccine, bottlenecks emerged and economic principles that high-
lighted information flows helped to solve them (Lee, Mueller, & Tilchin, 2017).
The COVID-​19 vaccines case highlights the importance of uncertainty
and information as key factors that can create inefficiency. Vaccine makers
18 Relevance of health economics to vaccines

faced uncertainty about the odds of successful discovery and changes in de-
mand, revenue, and liability. Vaccination systems faced uncertainty about the
delivery schedules and the role of vaccine hesitancy. The economic princi-
ples used to manage uncertainty required measures to pool risks into larger
and larger groups and ultimately required both national governments and
multigovernmental entities to share and manage risk.
The role of scarcity loomed even larger in the roll-​out of COVID-​19 vac-
cines. The traditional market-​driven approach to handle scarcity through
prices set by supply and demand could not be used because of equity con-
cerns and the large footprint of government investments and advance pur-
chase commitments. The large role of government and multigovernmental
agencies in vaccine procurement has always been a major feature of vac-
cine economics. Government involvement in vaccine markets does
not mean that market economics is of no use in understanding vaccine dis-
tribution. Instead, it implies the need to appreciate the function of heavily
regulated markets instead of free markets. For many in public health, vac-
cines are “free,” so the price of a free thing seems strange. This chapter will
outline why the price of a vaccine—​even a “free” one—​is extremely useful to
manage and minimize the burden of scarcity.
The first part of this chapter will develop the relevance of health economics
to vaccines by describing the system responsible for vaccines and vaccina-
tion. This approach makes sure we cover all the elements and highlight the
many areas where economic principles apply. The second part of this chapter
develops the foundational microeconomics of vaccine supply and demand,
highlighting the role of scarcity and public goods.

1.2.1 Vaccines and the immunization system

An economic system is composed of organizational units of human agents

who abide by norms, rules, and cultural expectations called “institu-
tions” in order to adjust and adapt to a changing environment (Grossman,
1967). The WHO Building Blocks Framework (WHO, 2010) has identi-
fied the typical units that one will find in a health system. The building
blocks of medical products, financing, information, service delivery, gov-
ernance, and workforce are all at work in the system responsible for vac-
cines and vaccination (Fig. 1.2.1). As shown in the figure, changes in the
institutions governing financing for discovery, procurement, distribution,
and delivery will create actions and reactions by the units responsible for
 1.2.1 Vaccines and the immunization system 19

Financing for
Financing for Financing for
distribution and
discovery procurement
delivery

Procurement and distribution

to countries

Vaccine discovery Vaccine production Distribution within countries

Distribution to clinics and

people

Information to regulate Information to regulate

safety distribution and delivery

Fig. 1.2.1 Vaccines and vaccination system.

vaccine discovery, production, and distribution. For example, if institu-

tional changes provide investments for vaccine discovery out of public
funds instead of private capital markets, then vaccine discovery can be pri-
oritized by what governments say is important (e.g., COVID-​19 vaccines)
as opposed to what financial investors think will generate the most rev-
enue. The informational state shown by the diamonds at the bottom of the
figure can also affect the system. If new information arises about social dis-
parities in vaccine distribution, then institutional reforms can be crafted to
make new incentives for the vaccine to be distributed to unreached priority
populations.

1.2.1.1 The role of financing

There are financial reasons limiting private investment in vaccines for the
world’s most needed vaccines: HIV, tuberculosis, and malaria (Bishai, Lin,
& Kiyonga, 2001; Bishai & Mercer, 2001; Kremer, 2002). A financing system
for vaccine discovery that relies heavily on private capital markets will under-
value diseases that affect people who cannot pay high prices. With privately
financed vaccine discovery, firms are motivated to discover new products so
they can obtain a patent that awards them monopoly rights to produce and sell
it. The investments they make in discovery will be paid back after the product
is approved because the patent allows them to mark up the price without fear
20 Relevance of health economics to vaccines

Table 1.2.1 The range of vaccine prices negotiated by the Pan American
Health Organization’s Revolving Fund

Vaccine Weighted average

price per dose

DTaP/​IPV/​Hep B/​Hib (hexavalent acellular) $21.12

Meningococcal ACYW135 $20.30
Varicella $15.85
Inactivated polio (IPV) $3.10
DPT Hib lyophilized (pediatric) $3.00
Measles/​mumps (Zagreb)/​rubella $2.75
BCG $0.23
DPT (pediatric) $0.18
Oral polio bivalent (bOPV) $0.17

DTaP/​IPV/​Hep B/​Hib, diphtheria, tetanus, pertussis, polio, hepatitis B, and Haemophilus

influenzae type b.

of direct competition. Higher prices will lead to less demand as poorer con-
sumers are priced out of the market. These high prices build in financial in-
equity in who will receive new vaccines which skews decisions about what
products to produce towards vaccines for diseases of affluent countries and
not low-​and middle-​income countries.
Because newer vaccines are covered by patents or have not yet garnered
competition, their prices are higher. Table 1.2.1 shows the price variation be-
tween newer vaccines like hexavalent, meningococcal, and varicella as op-
posed to the older generic vaccines like bacillus Calmette–​Guérin (BCG),
diphtheria, pertussis, and tetanus (DPT), and oral polio. Table 1.2.1 is pre-
senting prices that have been negotiated by PAHO (2020). These prices are
much lower than any one individual could obtain because PAHO represents
the purchasing power of 41 countries. Firms can offer reduced prices because
PAHO purchases in bulk and can offer reliable forecasts of future demand.
Pooled purchasing to achieve lower prices is a basic principle of vaccine
economics.
Once vaccines are procured, the financing and economic choices for their
distribution and delivery inside a country play a large role in determining
success. Most countries have both public and private systems for distribution
of vaccines to venues and to pay the staff to store, schedule, and administer
vaccines. The advantages of the public sector include the ability to main-
tain a focus on the most vulnerable populations and to occasionally deploy
campaigns and door-​to-​door delivery strategies. However, the public sector
 1.2.1 Vaccines and the immunization system 21

requires taxation or foreign assistance to secure public finances. The private

sector can deliver vaccines with finances raised by fees, but private providers
will seldom focus on the most vulnerable groups. Costs per fully immun-
ized child vary widely depending on the delivery strategy selected (Vaughan
et al., 2019).

1.2.1.2 The role of information

Removing barriers to the flow of information emerges as an important area

to improve system function. Scientific data on technologies to enhance safety,
and efficacy, as well as credible data on the burden of vaccine-​preventable
disease can help spur rapid vaccine development. For example, when mem-
bers of the Meningitis Vaccine Project (MVP) began to develop a vaccine
for meningitis A in sub-​Saharan Africa, they searched for a way to bind the
bacterial antigen to a carrier molecule that would incite immunity. When
they approached pharmaceutical companies who kept their technology li-
censes behind a paywall, they found their way blocked by licensing fees that
would make the vaccine price out of range for most African countries. But
MVP kept searching and found an appropriate chemical pathway that had
been developed by a scientist at the Food and Drug Administration with a
license owned by the US government. This offered a low-​cost and effective
method to conjugate polysaccharides to a carrier protein. This small snippet
of information kept the final price of the Menafrivac vaccine well below $1.00
and shaved off years of product development (Bishai, Champion, Steele, &
Thompson, 2011).
Information about the cost-​effectiveness of various antigens has had a dra-
matic effect on scaling up vaccine programs and new vaccine introduction
(Hutubessy, Henao, Namgyal, Moorthy, & Hombach, 2011). Estimates of the
cost per life saved or disability-​adjusted life year averted is a critical element
among many criteria used to assess whether or not to introduce a new vaccine
(Pooripussarakul, Riewpaiboon, Bishai, Muangchana, & Tantivess, 2016).
Studies comparing the cost of vaccination to the cost of vaccine-​preventable
disease in low-​and middle-​income countries routinely find that money spent
on vaccines generates financial savings. If one includes the dollar value of a
death averted, one dollar spent on vaccines in low-​to middle-​income coun-
tries would return $19.80 in averted cost of illness or $52.20 based on a value-​
of-​statistical-​life approach from 2021 to 2030 (Sim, Watts, Constenla, Brenzel,
& Patenaude, 2020).
22 Relevance of health economics to vaccines

1.2.2 The price of free things

From the perspective of consumers, vaccines often appear to have a zero price.
The absence of prices that would ordinarily signal value, scarcity, and abun-
dance to consumers increases the importance of economic analysis to deter-
mine the efficient level of supply. Furthermore, even when vaccines’ monetary
price to a consumer is $0, there will still be costs that the individual must bear
in the form of transportation, lost work time, and minor or major side effects.
These private disincentives to vaccination face off against the public interests
to achieve herd immunity.
Vaccines are dual public and private goods. They offer private benefits by
helping protect the vaccine recipient, but they also provide public benefits
to the community by lowering rates of contagious disease. They are neither
a wholly pure public good nor a wholly pure private good. The duality is the
root of an eternal policy dilemma in the economics of vaccines. The perspec-
tive of an individual trying to know how much cost to bear to get the pri-
vate benefits of vaccination clashes with the perspective of a health planner
trying to invest in procurement and subsidies to drive up the vaccine cov-
erage rates to offer the public benefit. The public investments can crowd out
reliance on the private willingness to pay for vaccination. Furthermore, as
the coverage rates get higher and higher, the disease prevalence goes lower
and lower. The personal benefit from vaccination becomes smaller (Geoffard
& Philipson, 1997).
Ordinarily, consumers’ demand curve and industry’s supply curve would
meet at a point of equilibrium and efficiency. However, for vaccines, the money
price to the consumer is zero and the demand curve (Fig. 1.2.2) is driven by
out-​of-​pocket costs required for transportation and waiting time. Meanwhile,
the supply curve is driven by public procurement of vaccines. Total market
supply and total market demand are not running off of the same price point.
Furthermore, the private demand expressed by consumers is based on their
own private calculation of benefit. In contrast, the public supply secured by
the public procurement is based on a calculation of public benefit that in-
cludes benefits of herd immunity, future labor productivity, and avoidance of
publicly financed losses due to vaccine-​preventable disease.
The demand curve (Fig. 1.2.2) is a graph relating prices on the vertical axis
to the number of units sold on the horizontal axis. The downward slope is
common sense: fewer people want and can afford a thing with a high price and
more people want and can afford a thing with a low price. Trying to minimize
both monetary and non-​monetary costs of obtaining vaccinations is a basic
policy to improve uptake.
 1.2.2 The price of free things 23

Total personal out-of-pocket spending per dose

Demand curve B
Subpopulation perceives higher
benefit or lower risk

Demand curve A
Subpopulation perceives lower
benefit or higher risk

Proportion of population getting vaccine

Fig. 1.2.2 Downward-​sloping demand curves.

Implied in the downward-​sloping demand curve is an inherent disparity.

Those who access something with a high price have social and economic ad-
vantages compared to those who have to wait for lower prices. Even when
monetary prices are zero, disparities in vaccine uptake can be seen because the
out-​of-​pocket costs for transportation and waiting are more easily borne by
those with more means. Vaccine hesitancy (discussed further in Chapter 1.4)
can also create disparities if misinformation about vaccine benefits or risks
is more concentrated in a particular subpopulation. Fig. 1.2.2 shows two de-
mand curves. The solid demand curve on the left might be a subgroup with
an information set making them think that vaccine benefits are lower/​vaccine
risks are higher compared to the subpopulation on the right with the dotted
line. Interventions that improve information, beliefs, and trust can shift de-
mand curves to the right and improve vaccine uptake.
The upward-​sloping supply curves (Fig. 1.2.3) also reflect common sense.
Holding all else equal, there are fewer firms who are able to make a profit when
the price per dose is low. Thus, the number of doses supplied is lower at low
prices and higher at high prices. The supply curves in Fig. 1.2.3 can be shifted
from curve A on the left to curve B on the right if there are innovations that
lower production costs. Pooled purchasing helps lower companies’ unit cost
of marketing and contracting and lowers their future risk by offering reliable
demand forecasts backed up by larger numbers of buyers. Other ways to lower
production costs are to scale up the production process. Larger production
facilities based on larger orders are able to cut costs and shift the supply curve
24 Relevance of health economics to vaccines

Supply curve A
When production costs
are higher
Dollars received per dose

Supply curve B
When production costs
are lower

Dose supplied

Fig. 1.2.3 Upward-​sloping supply curves.

to the right. Public subsidies to scientific and engineering progress can also
lower firms’ production costs and generate more vaccines at lower costs as
long as these innovations remain in the public domain. To the extent that
companies find a way to patent technology that was made possible by publicly
funded scientists, they will use their monopolistic advantage to raise prices
without expanding supply.
The disjunction between vaccine supply and demand is what makes vac-
cine economics exquisite. In markets for pizza, the social goal of getting every
pizza-​eater the optimal amount of pizza is achieved by letting the market work
things out on its own. Pizza-​eating might have unrecognized spillover bene-
fits, but they are nothing like what can be achieved by vaccine uptake.
The public benefits of vaccines have led to massive subsidies of their mon-
etary prices and efforts to improve the information that helps the vaccine
system work. Vaccine economists recognize the rationale for this government
involvement in the vaccine system and the need to stay deeply involved in a
market that can never work things out on its own to achieve the optimal result
for society.

1.2.3 Conclusion

Choices with uncertainty and scarcity typify the work of vaccination sys-
tems. These are areas where economic analysis can be extremely helpful. The
 1.2.3 Conclusion 25

chapter put forth a systems framework showing a reliance on information and

finance to chip away uncertainty and scarcity. This is a complex system: ac-
tion breeds reaction. Setting up one source of finance (e.g., public) crowds out
another (e.g., private). Opening up one information flow offers advantages to
some and not others. A policy to fix one problem creates another.
The basic tenets of market economics that celebrate a happy equilibrium
where supply meets demand cannot be applied because the private demand
for vaccines will not automatically align with the demand necessary to pro-
duce the public good of herd immunity. However, on vaccine demand, there
is plenty of rich economic work to do in measuring factors that affect the
response of people to the non-​monetary obstacles to vaccination. There is
plenty to understand about what makes some subgroups experience right-
ward and leftward shifts in their demand for vaccines. On vaccine supply,
there is much left to uncover about the forces governing the pace of vaccine
discovery: the entry of firms and efficiency improvements in vaccine produc-
tion. Economic research has only begun to develop better insights into the
information flows that can improve the vaccine system. The perpetual chal-
lenges of financing discovery, production, and distribution of vaccines will
remain rich areas for study.
In the wake of the COVID-​19 pandemic, access to a safe and effective
COVID-​19 vaccine became the most important issue in health, politics, and
economics. Vaccine economics justly deserves more attention in the wake of
COVID-​19. However, beyond this one vaccine, there lie horizons where mil-
lions more lives can be saved in the 21st century with improvements in the
world’s vaccine systems.

References
Bishai, D., Champion, C., Steele, M. E., & Thompson, L. (2011). Product development partner-
ships hit their stride: Lessons from developing a meningitis vaccine for Africa. Health Affairs,
30(6), 1058–​1064. doi.org/​10.1377/​hlthaff.2011.0295
Bishai, D., Lin, M. K., & Kiyonga, C. (2001). Modeling the economic benefits of an AIDS vaccine.
Vaccine, 20(3–​4), 526–​531. doi.org/​10.1016/​s0264-​410x(01)00335-​8
Bishai, D., & Mercer, D. (2001). Modeling the economic benefits of better TB vaccines.
International Journal of Tuberculosis and Lung Disease, 5(11), 984–​993.
Geoffard, P.-​Y., & Philipson, T. (1997). Disease eradication: Private versus public vaccination.
American Economic Review, 87(1), 222–​230.
Grauer, J., Löwen, H., & Liebchen, B. (2020). Strategic spatiotemporal vaccine distribution in-
creases the survival rate in an infectious disease like Covid-​19. Scientific Reports, 10(1), 1–​10.
doi.org/​10.1038/​s41598-​020-​78447-​3
Grossman, G. (1967). Economic systems. Englewood Cliffs, NJ: Prentice-​Hall.
26 Relevance of health economics to vaccines

Hutubessy, R., Henao, A. M., Namgyal, P., Moorthy, V., & Hombach, J. (2011). Results from
evaluations of models and cost-​effectiveness tools to support introduction decisions for new
vaccines need critical appraisal. BMC Medicine, 9(1), 1–​4. doi.org/​10.1186/​1741-​7015-​9-​55
Kickbusch, I., Krech, R., Franz, C., & Wells, N. (2018). Banking for health: Opportunities in
cooperation between banking and health applying innovation from other sectors. BMJ Global
Health, 3(Suppl 1), e000598. doi.org/​10.1136/​bmjgh-​2017-​000598
Kremer, M. (2002). Pharmaceuticals and the developing world. Journal of Economic Perspectives,
16(4), 67–​90. doi.org/​10.1257/​089533002320950984
Kremer, M., Levin, J., & Snyder, C. M. (2020). Advance market commitments: Insights from
theory and experience. AEA Papers and Proceedings, 110, 269–​273. https://​doi.org/​10.1257/​
pandp.20201​017
Kuehn, B. M. (2020). Interest in global COVID-​19 vaccine alliance grows. JAMA, 324(11),
1025–​1025. doi.org/​10.1001/​jama.2020.16966
Lee, B. Y., Mueller, L. E., & Tilchin, C. G. (2017). A systems approach to vaccine decision
making. Vaccine, 35, A36–​A42. doi.org/​10.1016/​j.vaccine.2016.11.033
McDonnell, A., Van Exan, R., Lloyd, S., Subramanian, L., Chalkidou, K., La Porta,
A., . . . Rosenberg, J. (2020). COVID-​19 vaccine predictions: Using mathematical modelling
and expert opinions to estimate timelines and probabilities of success of COVID-​19 vaccines.
Washington, DC: Center for Global Development.
National Academies of Sciences, Engineering, & Medicine. (2020). Framework for equitable
allocation of COVID-​19 vaccine. Washington, DC: National Academies Press.
Pan American Health Organization. (2020). PAHO revolving fund vaccine prices for 2020.
https://​www.paho.org/​en/​docume​nts/​paho-​revolv​ing-​fund-​vacc​ine-​pri​ces-​2020
Pooripussarakul, S., Riewpaiboon, A., Bishai, D., Muangchana, C., & Tantivess, S. (2016). What cri-
teria do decision makers in Thailand use to set priorities for vaccine introduction? BMC Public
Health, 16(1), 684. doi.org/​10.1186/​s12889-​016-​3382-​5
Shnaydman, V. (2020). Simulation model for productivity, risk and GDP impact forecasting of
the COVID-​19 portfolio vaccines. medRxiv. doi.org/​10.1101/​2020.11.01.20214122
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353. doi.org/​10.1377/​hlthaff.2020.00103
Vaughan, K., Ozaltin, A., Mallow, M., Moi, F., Wilkason, C., Stone, J., & Brenzel, L. (2019). The costs
of delivering vaccines in low-​and middle-​income countries: Findings from a systematic review.
Vaccine: X, 2, 100034. doi:10.1016/​j.jvacx.2019.100034
World Health Organization. (2010). Monitoring the building blocks of health systems: A hand-
book of indicators and their measurement strategies. https://​apps.who.int/​iris/​bitstr​eam/​han​
dle/​10665/​258​734/​978924​1564​052-​eng.pdf
World Health Organization. (2020). WHO SAGE values framework for the allocation and prior-
itization of COVID-​19 vaccination. https://​www.who.int/​publi​cati​ons/​i/​item/​who-​sage-​val​
ues-​framew​ork-​for-​the-​all​ocat​ion-​and-​pri​orit​izat​ion-​of-​covid-​19-​vacc​inat​ion
World Health Organization. (2021). Guidance on developing a national deployment and vacci-
nation plan for COVID-​19 vaccines. https://​www.who.int/​publi​cati​ons/​i/​item/​WHO-​2019-​
nCoV-​Vacc​ine-​dep​loym​ent-​2021.1-​eng
Yamey, G., Schäferhoff, M., Hatchett, R., Pate, M., Zhao, F., & McDade, K. K. (2020). Ensuring
global access to COVID-​19 vaccines. Lancet, 395(10234), 1405–​1406. doi.org/​10.1016/​
S0140-​6736(20)30763-​7
1.3
Cost of finding and making
vaccines
Implications for immunization programs
Clarke B. Cole, Beth Evans, and Soleine Scotney

1.3.1 Problem statement

Ensuring access to affordable vaccines is a critical function of public health

systems. To achieve this, two questions need to be answered: What vaccines
are available? Which vaccines are more affordable? This chapter provides an
overview of the vaccine market landscape and then discusses the strategies
used by decision makers and policymakers to help improve affordability of
and sustainable access to vaccines.
The first part of this chapter covers the costs and risks of vaccine develop-
ment and production and how this influences which vaccines are developed
and how much they cost to produce and buy. The second part provides an over-
view of the vaccine supply landscape to provide an understanding of the range
of vaccine companies and products and the ways in which market dynamics
incentivize company behaviors. The third part details how procurement mo-
dalities and mechanisms impact a country’s level of access to vaccines. By the
end of this chapter, readers will have an improved understanding of how dif-
ferent decisions and actions can influence access to sustainable supply of af-
fordable vaccines.

1.3.2 Introduction to vaccine costs and pricing

What drives the pricing of vaccines? In highly competitive markets, prices are
tied closely to production costs, and in perfect markets, price theoretically

Clarke B. Cole, Beth Evans, and Soleine Scotney, Cost of finding and making vaccines In: Handbook of Applied Health Economics
in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0004
28 Cost of finding and making vaccines

equals marginal cost. However, many vaccines markets are oligopolies with
a small number of firms holding the greatest market share. As a result, cost
plus pricing (i.e., pricing vaccines at cost plus a given margin) is rarely used
in vaccines markets. Understanding the costs of vaccine development and
production are crucial for recognizing the theoretical floor price of vaccines,
particularly given the high costs typically associated with vaccine research
and development (R&D). This section explores the components that con-
tribute to the cost of developing and producing vaccines, and how these costs
link to prices charged on the market. This is followed by a discussion on dif-
ferent approaches used to set vaccine prices.

1.3.2.1 Vaccine research and development costs

R&D is expensive and risky. Typically, taking a candidate from discovery and
preclinical testing through clinical trials and to licensure, including regula-
tory approvals, is estimated to take between 5 and 18 years (Médecins Sans
Frontières, 2010) and cost between US $200 million and US $500 million
(André, 2002). The key categories of R&D costs include costs incurred to dis-
cover, develop, and bring a vaccine to market (e.g., clinical trials, regulatory
approval including World Health Organization (WHO) prequalification of
medicines). In some cases, vaccine R&D costs can include costs to in-​license
product-​related intellectual property to further develop a product in-​house.
Further third-​party contributions (e.g., grants, loans, subsidies) are common
to support R&D; these should be included in product cost calculations to
offset costs.
R&D spending can be considered a fixed cost. R&D costs vary by supplier
and product based on scientific hurdles associated with different product
types, the size and complexity of clinical trials required, and the developer’s
operational costs, among other factors.
If development is successful, vaccine developers have significant costs to
recoup through sales (unless development funding has been secured exter-
nally, e.g., via a grant). Due to the continuous nature of innovation, suppliers
may expect to face a roughly normally distributed curve in terms of sales over
time—​with a period of ramp-​up when products become licensed and intro-
duced, a period of steady sales, and then a period of ramp-​down as technology
is replaced by new, innovative, or competitor products. Thus, not only must
developers recoup initial development investments, but they also face a lim-
ited period in which they may be able to recoup those costs.
 1.3.2 Introduction to vaccine costs and pricing 29

Sometimes, vaccine development is unsuccessful. Analysis of vaccine de-

velopment data from 1998 to 2009 found that there was just a 6% probability
of market entry (i.e., 94% chance of failure) for vaccine candidates from the
preclinical phase (Pronker, Weenen, Commandeur, Claassen, & Osterhaus,
2013). When failures occur, the developer no longer has prospects of direct
returns on the investments. It has sunk into a failed vaccine candidate. Most
suppliers will aim to recover such investments, at least partially, through sales
of successful products.

1.3.2.2 Vaccine production costs

Costs of production vary by supplier, product, and geography. In general,

costs incurred by vaccine companies throughout the production cycle of a
vaccine can be grouped into the following categories (Bill & Melinda Gates
Foundation, 2016):

• Facilities and equipment: costs associated with fixed assets, including

capitalized costs that depreciate over time (e.g., land, buildings, ma-
chinery) as well as ongoing costs of upkeep (e.g., repairs and mainte-
nance, utilities).
• Direct labor: employee costs (e.g., wages, benefits) directly attributable to
a specific vaccine.
• Consumables: raw materials used as inputs in production of a specific vac-
cine such as biological and chemical agents; fill/​finish consumables such
as vials, stoppers, and seals; packaging consumables such as labels and
vaccine vial monitors; and quality control consumables such as inputs for
testing kits.
• Overheads: operational costs not directly attributable to a specific vaccine
such as salaries, training, back-​office functions, and insurance.
• Commercialization: expenses incurred post regulatory approval asso-
ciated with selling and marketing the product in the relevant market
(e.g., advertising, marketing, distribution, etc.).
• Licensing: expenses paid for licensing the right to use product-​related
intellectual property in order to develop, produce, and/​or sell a
vaccine.
• Interest costs: interest payments as a result of company financing, such
as interest costs for working capital (inventories plus receivables minus
payables) and interest for capital investments.
30 Cost of finding and making vaccines

• Third-​party contributions: all contributions (e.g., grants, loans, subsidies)

from governments and other organizations should be included in product
cost calculations to offset costs.

1.3.2.3 Accounting for vaccine development and

production costs

Cost categories can be broadly classified into four groups: fixed, variable,
semi-​variable, and mixed. The relative contribution of each cost classifica-
tion determines how changes in volume impact production economics (Bill &
Melinda Gates Foundation, 2016).

• Fixed costs are costs that do not change as output increases or decreases.
Thus, increasing sales volume can spread fixed costs across more doses.
Examples include product development costs, facilities, equipment, and
third-​party financing costs.
• Variable costs increase directly with additional output, for example, con-
sumables such as vaccine vials.
• Semi-​variable costs are correlated with output in aggregate, but not as di-
rectly as variable costs. Direct labor costs do not materially increase with
each additional vaccine unit produced, but do increase in aggregate pro-
portional to total output, for example, in a stepwise manner as additional
shifts are required.
• Mixed costs refer to groupings of costs that include both fixed and var-
iable components, for example, commercialization and licensing. In
order to understand, model, and plan for the effects of volume on price,
it is best to separate mixed costs into their component fixed and variable
subcomponents.

1.3.2.4 Impact of vaccine cost structure on vaccine

economics

The types of costs faced when developing and producing vaccines are not
static; they vary based on several factors. The cost structure of vaccines influ-
ences the lower bound of pricing possible for a given vaccine product (pre-
suming suppliers will not sell below fully loaded cost of goods). It is important
 1.3.2 Introduction to vaccine costs and pricing 31

for procurers to understand the dynamics that influence costs and determine
what levers are available to access lower prices.
Costs vary depending on the following factors (Bill & Melinda Gates
Foundation, 2016):

• Production volume. As production volume increases (up until capacity

constraints are reached), the fixed production costs per dose decrease—​
allowing price to decrease.
• Predictability. Predictability of output volumes over time benefits sup-
pliers through:
• Enabling appropriate amortization of fixed costs.
• Appropriate sizing of the workforce and efficient operations—​for ex-
ample, purchasing the correct volume of (perishable) consumables.
• Enabling supply to meet demand: it can take up to 3 years to produce
a batch of vaccines due to the complexity of manufacturing (Vidor &
Soubeyrand, 2016), meaning that production needs to be well planned
in advance to meet target output in the future.
• Product complexity. Development of novel vaccines that target areas of
unmet health need where there are no existing vaccines may be more
expensive and complex than development of vaccine products where a
product already exists (Heaton, 2020). This is driven, for example, by the
need to generate completely new data and navigate new pathways to li-
censure. Partnerships with donor agencies or public procurers that help
to spread risks (either in terms of upfront funding for R&D costs (an ex-
ample of push funding) or increased demand certainty (an example of
pull funding)) can help to overcome such barriers.
• Product packaging requirements. Consumables, including the primary
and secondary packages and labels for vaccines, are variable costs, which
increase as production increases. Packaging costs can become further
compounded with rising production when multiple different vial pres-
entations, secondary packages, and labels are required to serve interna-
tional markets due to country-​specific product preferences and language
requirements. Smaller production runs impose additional direct labor
costs associated with process management. Furthermore, formulation of
both multidose and single-​dose vials adds new costs related to additional
clinical testing.
• Regulatory requirements. Country-​ specific regulatory approval pro-
cesses place an administrative burden on suppliers, which increases
32 Cost of finding and making vaccines

resource requirements and costs. When countries simplify or harmonize

regulatory requirements, this can speed up time to market entry (Wolf
et al., 2020).
• Fixed and indirect cost allocation possibilities. The allocation of fixed and
indirect costs is a decision that depends on feasible options influenced by
the number of products, markets, and time that costs can be split over.
While some of these factors are primarily within the suppliers’ control,
others can be influenced by governments and other purchasers. For ex-
ample, procurers can improve predictability of demand for suppliers by
structuring tenders over longer-​term periods or they can signal an in-
tention to introduce particular vaccines in the future, clearly indicating
preferred product characteristics. Countries purchasing vaccines can
also harmonize regulatory approval processes with international norms,
increasing ease of market entry for suppliers and potentially facilitating
access to lower prices.

1.3.2.5 Vaccine pricing

Vaccine prices vary from country to country depending on a range of factors,

particularly supplier-​specific targets for returns on investment, supply-​side
and demand-​side impacts on cost of goods sold, and the extent of uncertainty
or predictability in supply forecasts. This section focuses on pricing models
used by firms in vaccine markets; however, it is worth noting that the pricing
models used by firms are also influenced by purchaser levers (e.g., procure-
ment mechanisms, negotiating power, and ability-​or willingness-​to-​pay). In
perfectly competitive markets, firms theoretically use marginal cost pricing,
which is the practice of setting a price equal to the extra cost of producing an
extra unit of output (e.g., vial) including normal profit. This pricing model is
not typically used in vaccine markets because few have close to perfect market
conditions.
Given high market entry barriers, most vaccine markets are typically ol-
igopolistic with a small number of firms gaining the greatest market share.
As a result, firms can be price setters, rather than price takers and alternative
pricing models are often used, for example:

• Reference or competitive pricing: the practice of setting a price by com-

paring prices set by competitors in similar markets. This may include
external price-​setting—​in which firms set prices achieved in similar
 1.3.3 Vaccine supply landscape 33

countries matched by geography and/​or income level. This is often seen

in middle-​income countries to inform negotiations or decision-​making
on product selection.
• Value-​based pricing: pricing based on the value a product brings to the
market or health system can be measured in various ways (e.g., in-
cremental benefit compared to incumbent product, costs averted to
a health system by preventing a disease compared to treating it). This
method could be used by firms as the basis for setting vaccine prices.
Health technology assessment is increasingly being conducted by high-​
income countries to help make evidence-​based decisions on reim-
bursement and funding based on the value of new product options over
existing ones.
• Tiered pricing: tiered pricing or price differentiation is the practice of of-
fering different prices to different markets based on ability to pay. This
strategy can enable revenue growth through enabling nationwide vac-
cine introductions at a price affordable to a given government, thus
increasing volumes of sales—​and potentially in turn spreading fixed
production costs over more doses. For example, GlaxoSmithKline uses
World Bank gross national income per capita groupings to tier prices of
vaccines, in addition to other factors. (GlaxoSmithKline, 2019). While
GlaxoSmithKline does not publish prices by tier, review of anonym-
ized vaccine purchase data provided by WHO Market Information for
Access to Vaccines (MI4A) shows that prices (e.g., for pneumococcal
conjugate vaccine (PCV)) may vary by a factor of up to 25 times between
low-​income countries and high-​income countries (WHO: Immunization
Vaccines and Biologicals, 2019). If tiered pricing only consists of one high
price for high-​income countries and one low price set near marginal costs
for low-​and middle-​income countries (LMICs) then it will not lead to
higher profit than simply having one high price for high-​income coun-
tries. It is the use of multiple middle tiers with mid-​level prices above
marginal cost that offers firms the ability to capture revenue in excess of
costs that would otherwise be lost.

1.3.3 Vaccine supply landscape

This section of the chapter describes the range of vaccine companies devel-
oping and manufacturing vaccines. It describes how suppliers perceive and
respond to market signals, thus influencing product development.
34 Cost of finding and making vaccines

1.3.3.1 Vaccine suppliers

The landscape of vaccine suppliers has important implications for the availa-
bility of affordable vaccines that meet population health needs. The global vac-
cine market is highly consolidated on the supply side, with four multinational
corporations (MNCs) accounting for approximately 80% of global vaccine
revenues: GlaxoSmithKline plc (which acquired Novartis’ vaccines business
in 2015), Merck & Co., Inc. (known as Merck, Sharp & Dohme outside the US
and Canada), Pfizer Inc., and Sanofi (Access to Medicine Foundation, 2017).
Such multinational companies have historically been responsible for bringing
high-​impact, innovative vaccines (such as Pfizer Inc.’s PCV and Merck & Co.,
Inc.’s HPV vaccine) to market for the first time. Typically, products are devel-
oped in response to high-​income country demand and first launched there,
before rolling out to lower-​income countries (this time lag has decreased in
recent years).
There is also a growing number of vaccine companies based in emerging
markets, organized as members of the Developing Countries Vaccine
Manufacturers Network (DCVMN). This network of over 40 manufacturers
typically supplies lower-​priced, traditional vaccines to the global market
(such as pentavalent, measles and rubella, and typhoid vaccines). These
products have traditionally been either biosimilar vaccines developed in-​
house or in-​licensed products from innovating companies. While they do
not account for a large portion of global vaccine revenues, DCVMN mem-
bers’ contributions to the vaccine market are substantial: in combination,
they supply vaccines for approximately 84% of the world’s birth cohort
each year and represent 64 out of 147 WHO prequalified vaccines (Batson,
2016). In recent years, some DCVMN members have increased their focus
on vaccine R&D, including both adapting existing vaccines to meet the
specific needs of LMICs and developing novel first-​generation products.
Provided they adhere to globally accepted quality and safety standards
such as WHO prequalification, they should be considered comparable to
MNC products.
Adding to the complexity of the vaccine supply landscape are smaller
biotechnology firms, which are increasingly responsible for conducting
early-​stage vaccines R&D. When their research shows promise, these
vaccine products are usually out licensed to or acquired (sometimes in
the form of the entire biotech firm) by larger companies like MNCs
or members of the DCVMN for phase II clinical testing onwards and
commercialization.
 1.3.3 Vaccine supply landscape 35

1.3.3.2 Competition in vaccines markets

Markets for newer vaccines are not competitive; they usually include a lim-
ited number of suppliers, typically MNCs. The lack of price competition has
contributed to high prices in vaccine markets. Because products are usually
heterogeneous with regard to product presentation (e.g., different number of
doses, cold-​chain requirements, or efficacy), it is difficult for buyers to com-
pare products, which further impedes competition. For example, transpar-
ency on prices paid by different countries for the same product is relatively
low. However, the WHO MI4A vaccine purchase database has helped to in-
crease transparency in this area by publishing anonymized data on prices paid
by countries for different vaccine products, which can be viewed based on
characteristics such as purchase order volume, procurement mechanism, and
country income level.
There are high barriers to entry due to the complexity and cost of vaccine
development and production. There are also barriers to exit due to the com-
plexity and cost of vaccine development and production. These barriers result
in limited suppliers of innovative vaccines for a number of years, after which
additional competitors may come to market with next-​generation vaccines.
There may sometimes even be barriers, or at least delays, to exit due to pro-
duction timelines and potentially advance tenders. Market-​shaping teams in
global health organizations work to identify potential gaps in supply or afford-
ability and provide incentives to suppliers to enter or remain in the market. The
Healthy Markets Framework, developed jointly by Gavi, UNICEF, and the Bill
& Melinda Gates Foundation (BMGF), is a tool that helps to support such teams
in assessing the “health” of a given vaccine market and develop strategies to im-
prove it (Gavi, 2020).

1.3.3.3 Market signals and their impact on product

development

Given the significant scientific hurdles associated with vaccine R&D, vac-
cine developers face strong incentives to respond to the needs of high-​income
countries and governments with high abilities to pay or other lucrative mar-
kets (e.g., private travel vaccines). This means that private financing for
R&D for vaccines that disproportionately affect poor populations tends to
be insufficient—​resulting in slower development or deprioritization of vac-
cines to target unmet health needs focused in low-​and lower-​middle-​income
36 Cost of finding and making vaccines

countries. In addition, where vaccines do come to market, their characteristics

do not always meet the specific needs of populations in lower-​income coun-
tries, for example, in terms of serotype coverage (which influences efficacy).
Global donors and product development partnerships have played an im-
portant role in the vaccine supply landscape, providing incentives for vaccine
companies to develop and sell products that meet health needs, where com-
mercial incentives are insufficient to drive such R&D. These entities collabo-
rate with vaccine companies to share the risk that their investments will not
be recouped in the market (Cole & Iyer, 2016). One example is the Meningitis
Vaccine Project, in which the WHO, PATH (formerly known as the Program
for Appropriate Technology in Health), and Serum Institute of India, funded
by the BMGF, worked with African public health officials to develop a low-​cost
meningitis A vaccine (MenAfriVac) designed specifically to meet the needs of
populations along the African meningitis belt (PATH & WHO, 2017).
Furthermore, DCVMN members typically face different incentives than
MNCs: the domestic markets they operate in provide incentives to develop
products that meet the needs of local populations. In the case of Chinese and
Indian DCVMN members, the combination of large domestic markets (in ad-
dition to demand from Gavi-​eligible markets through pooled procurement)
provides incentives for R&D to focus on meeting the needs of LMIC popu-
lations. DCVMN members’ role in developing for and supplying to LMICs
is expected to increase. For example, in 2019, the Serum Institute of India
received Indian licensure and WHO prequalification for its PCV, which was
designed to protect against ten serotypes of Streptococcus pneumoniae most
likely to cause serious disease in Africa and Asia (Serum Institute of India &
PATH, 2020). In addition to increasing fit to local disease burden in LMICs,
this second-​generation vaccine lowered the price of PCV immunization by
one-​third, compared to the lowest price offered by a competitor (Gavi, 2021c).
Suppliers benefit from country signals, which may indicate market appetite
and stimulate vaccine development. These signals can include what prod-
ucts should be prioritized for development and under what conditions they
would be preferred by countries. However, clear country signals on future
demand for innovative vaccines are limited. Global think tanks, such as the
Centre for Global Development, have proposed interventions to improve
market signals and commitments to encourage product development. For
example, the market-​driven value-​based advance commitment (Chalkidou,
Garau, Silverman, & Towse, 2020), focuses on accelerating development of
health products to minimize the burden of tuberculosis. The proposal is based
on extending health technology assessments in-​country to not only evaluate
 1.3.4 Vaccine procurement 37

the value of potential products but also to commit to future procurement

from interested countries provided products meet the required target product
profile. Such demand signals and commitments may accelerate private sector
development. In the absence of such signals, suppliers may be hesitant to in-
vest the significant sums required to develop innovative vaccines.

1.3.4 Vaccine procurement

Vaccine procurement is the process whereby vaccines are purchased and de-
livered to governments or government agencies for public immunization
programs, and to private purchasers for the private market. A range of pro-
curement modalities exist—​each with different eligibility criteria and impli-
cations for national immunization programs. This section first outlines the
broad influence of procurement on vaccine economics, and then details the
range of procurement modalities, followed by a discussion of trends and
changes in the procurement landscape. It finishes with an exploration of inno-
vative vaccine procurement mechanisms.

1.3.4.1 Impact of procurement levers on vaccine

economics

The structure, eligibility, and processes used for different vaccine procurement
modalities can influence the price, quality, availability, and supply security
of different vaccine products. This is due to the influence that procurement
modalities have on the relationship between supply and demand. The main
levers are volume, visibility/​predictability, country context, and extent of op-
erational and regulatory requirements.
As explained above, suppliers typically wish to sell vaccines at scale and
in an efficient manner. Procurement modalities that pool demand from a
range of countries can be attractive to suppliers (ideally over multiple years),
since they provide clear demand signals allowing allocation and potentially
even expansion of production capacity to meet demand. This increased vis-
ibility and economic predictability of aggregated demand can enable sup-
pliers to lower prices while ensuring economies of scale and “right sizing”
of facilities. Typically, pooled and coordinated procurement modalities that
procure larger volumes than individual countries can make lower prices
accessible.
38 Cost of finding and making vaccines

Suppliers also benefit from streamlined processes that simplify product

presentation requirements (i.e., labelling, languages, doses per vial) since
these lower the cost of consumables and labor required to prepare appropriate
product presentations, shifting the cost curve. If aggregate volumes rise, but
require diverse country-​specific packaging, then suppliers will face increased
costs to prepare different product presentations, which may counteract the
benefits of scale. As a result, pooled/​coordinated procurement mechanisms
that agree on labelling harmonization may allow access to further lowered
prices, provided cost savings from the supplier are passed onto purchasers.
Once clinical trials have been successfully completed, vaccines undergo
quality and safety assurance steps through registration with a relevant regula-
tory authority for the production site. For in-​country use, vaccines typically re-
quire local registration with the national regulatory authority and, potentially,
further evidence of quality and safety—​often in a different format and with
different requirements to previous reviews. The existence of multiple different
registration and licensure requirements across countries can limit access to
vaccines in different geographies, due to the country-​specific administrative
requirements placed on suppliers. As a result, regulatory harmonization—​
such as acceptance of WHO prequalification in lieu of local regulatory as-
sessment, and use of the WHO Collaborative Procedure for Accelerated
Registration—​can ease the market entry pathway in countries. This improves
access to a wider range of vaccines as well as potentially increasing demand
and lowering prices which reflect these greater efficiencies.

1.3.4.2 Procurement modalities

At the highest level, there are two main procurement modalities for vac-
cines: self-​procurement and pooled procurement. In practice, many coun-
tries exploit a hybrid strategy with different products procured via different
routes. Most high-​income countries self-​procure vaccines, and—​due to the
existence of and eligibility required for Gavi and the Pan American Health
Organization (PAHO) (see section 1.3.4.2.2)—​many LMICs procure through
pooled procurement mechanisms.

1.3.4.2.1 Self-​procurement
Self-​procurement refers to the practice of procuring vaccines directly, either
from internal domestic suppliers or bilaterally with international companies,
and offers autonomy in terms of procurement schedule, tender frequency,
product choice, and other, broader goals. Domestic suppliers are most
 1.3.4 Vaccine procurement 39

prevalent in high-​income countries (e.g., the US, Europe, Japan, and South
Korea) as well as some large middle-​income countries (e.g., Brazil, China,
India, Thailand, Indonesia, and Vietnam). Where countries have domestic
vaccine companies, they often exhibit a preference for domestic supply—​
often due to the supply security offered and the ability to meet broader ec-
onomic and domestic security goals. These domestic suppliers may also be
subsidized by local governments, facilitating access to lower prices and ena-
bling more control on domestic production (e.g., which vaccines are devel-
oped and what characteristics they have).
Access to affordable vaccines through self-​procurement will typically de-
pend on countries’ requirements for vaccine registration and procurement,
the volumes being secured, and negotiation power. If suppliers are required
to conduct additional vaccine trials in the procuring country or navigate a
complex registration process, this may decrease the number of interested sup-
pliers. Bid prices may depend on population size (particularly if this is part-
nered with unique presentation requirements), meaning countries with small
population sizes and complex registration requirements may encounter ac-
cess or pricing challenges if they choose self-​procurement.
Taking supplier preferences into account in tender and contract structures
can lead to price reductions. Suppliers have communicated to countries and
partners that they prefer longer duration tenders (e.g., 3 years) to shorter 1-​
year tenders. However, some countries may have cash flow or other financial
restrictions, limiting the current tender and contract length and structure.
Countries could investigate changes to the country-​specific tender and con-
tract processes, coordinated with any additional government procurement
departments that may need to be involved in order to increase attractiveness
to suppliers and potentially access lower prices.
In addition, countries can leverage publicly shared insights from compa-
rable countries (i.e., in terms of income level or region) to understand market
prices achieved via self-​procurement and help negotiate a good price for the
given country. Countries can refer to the WHO MI4A vaccine purchase da-
tabase to review public procurement information (WHO: Immunization
Vaccines and Biologicals, 2019), for example, reviewing vaccine prices cat-
egorized by product type, presentation, and volume.

1.3.4.2.2 Pooled procurement
Pooled procurement refers to the practice of central procurement agencies
procuring on behalf of a group of countries, usually with defined conditions
on eligibility. This section covers the two most prominent vaccine pooled pro-
curement agencies: UNICEF Supply Division (SD) and the PAHO Revolving
40 Cost of finding and making vaccines

Fund (RF). Both procurement bodies aggregate demand for a limited number
of products for eligible countries. Other pooled procurement bodies have
been proposed, for example, for the ASEAN region (Siripitayakunkit, 2018),
and may emerge over time, as hurdles to achieve regulatory harmonization
are overcome.
Most vaccines procured by UNICEF SD are accessible to all countries,
provided they sign a memorandum of understanding with UNICEF SD—​
though prices depend on Gavi eligibility, and other income group criteria
for PCV, HPV, and rotavirus vaccine prices in particular. Many countries
procuring through UNICEF SD are Gavi eligible, based on country gross
national income (GNI) per capita.
PAHO RF procurement is open to countries in the Americas—​currently 34
economies in Latin America and the Caribbean have become PAHO mem-
bers (Pan American Health Organization, 2020). Countries are required to
pre-​pay for orders placed through UNICEF SD or PAHO RF.
Pooled procurement bodies are typically responsible for some logistical,
procedural, and regulatory steps in the procurement process such as ship-
ping, tendering, and negotiations. It can be beneficial to countries to have
these processes conducted by an independent body, since this decreases the
in-​country operational requirements and can leverage internationally ac-
cepted standards and processes that may not be available (or may be expen-
sive to provide) in-​country. This may come at a financial cost, for example,
UNICEF SD’s vaccines handling fee for Gavi countries is 1.4% (UNICEF,
2021). It may also be a requirement to accept specific tendering terms with
regard to the timing or duration of tenders and deliveries in order to procure
via the pooled procurement mechanism. This requires alignment in-​country
to forecast for the appropriate time period and organize the supply chain.
For countries participating in pooled procurement, the choice of vaccines
will generally be limited to a predetermined menu—​both in terms of products
and product presentation (e.g., only multidose vials may be available). The
consolidation of demand in a subset of products and product presentations
facilitates lower prices and attracts suppliers that benefit from streamlined co-
ordination of (stable) supply.

1.3.4.3 Trends in the procurement landscape

This section of the chapter describes some key trends in the procurement
landscape that may shape the way vaccines are procured in the coming decade.
 1.3.4 Vaccine procurement 41

Table 1.3.1 Gavi classification and eligibility for support

Gavi classification Eligibility conditions Co-​financing terms

Initial self-​financing GNI per capita below World Countries pay flat rate of US
Bank threshold for low-​income $0.20/​dose for each vaccine
countries
Preparatory transition GNI per capita above low-​income Country co-​financing
country threshold, and below contributions increase 15%
lower-​middle-​income country year-​on-​year
threshold
Accelerated transition GNI per capita “Gavi eligibility Country co-​financing
threshold” (US $1,580 per capita requirements increase
in 2019) linearly year-​on-​year for
5 years to reach 100% at end
Fully self-​financing After 5 years of accelerated Countries pay the full
transition, provided GNI per procurement cost
capita remains above threshold

1.3.4.3.1 Gavi transition
As a country’s GNI per capita rises, it will transition out of Gavi support from
full co-​financing, to preparatory transition, to accelerated transition, to fully
self-​financing (see Chapter 5). Countries are classified based on where their
average GNI per capita over the past 3 years falls compared to cut-​off thresh-
olds governing transition between phases (Gavi, 2018) (Table 1.3.1).
Growing independence from Gavi support has implications for procure-
ment, particularly for the newly fully self-​financing countries. Importantly,
even when countries transition from Gavi support they are still eligible to
continue to procure from UNICEF SD—​offering continued logistical advan-
tages. However, the prices they are eligible to access may evolve, depending
on the status of post-​transition pricing agreements. As it stands, post-​
transition pricing agreements exist for rotavirus and HPV vaccines as well
as PCV (WHO, 2018). In the future, it is expected that middle-​income coun-
tries that do not qualify for Gavi support may participate in pooled procure-
ment for vaccines, given persistent barriers to accessing vaccines in these
countries.

1.3.4.3.2 Switching procurement modalities

Driven by market intelligence and/​or changes in country preferences, a
country may wish to change how it procures a given vaccine, for example,
from procuring via UNICEF SD to self-​procuring or vice versa. Market intel-
ligence on product options and variations in availability and price via different
procurement mechanisms can be used to inform whether this is a beneficial
42 Cost of finding and making vaccines

decision and present an evidence base for the switch. It may be necessary to
define new processes (e.g., in-​country tendering procedures) or update agree-
ments with the pooled procurement body to switch modality.
As vaccine markets evolve and the number of suppliers for each market in-
creases, it will be important for policymakers and decision makers to leverage
existing capacity, such as through National Immunization Technical Advisory
Groups (2021) to switch products should new, alternative options better meet
the country’s needs.

1.3.4.4 Innovative procurement mechanisms

In recent years, innovative procurement mechanisms have been designed

and implemented to encourage the development of vaccines essential for
improving public health when market mechanisms are too slow, too expen-
sive, or are simply ignoring diseases affecting citizens in lower-​income coun-
tries. These innovative mechanisms are particularly important in augmenting
demand signals and reassuring suppliers that development of potentially
expensive, high-​risk candidates could be profitable and sustainable for their
businesses.

1.3.4.4.1 Advance market commitments

• Pneumococcal Advance Market Commitment (Pneumococcal AMC).
Pneumonia is the leading vaccine-​preventable cause of child mor-
tality, and PCVs have complex development and production processes.
To accelerate vaccine introduction and achieve affordable pricing for
LMICs, the Pneumococcal AMC was set up in 2007. The AMC was
a US $1.5 billion innovative financing “pull” fund put in place to in-
centivize development of PCVs to fit a target product profile to define
suitability for use in low-​income countries. As a “pull” mechanism, the
AMC created financial incentives in the form of a “top-​up” payment
on top of the price paid (or co-​financed by Gavi) for countries to en-
courage private sector engagement by creating viable market demand.
During the 10-​year window of the AMC (from 2010 to 2020), three
PCV products were developed and received WHO prequalification,
with 60/​73 (83%) eligible Gavi countries rolling vaccines out nation-
wide (Gavi, 2021b).
• The COVAX Facility. The coronavirus disease 2019 (COVID-​19) pan-
demic in 2020 required unprecedented acceleration of vaccine
 1.3.4 Vaccine procurement 43

development to control and respond to the novel coronavirus. Due to

the speed of development required, the high risks of failure, and the
need for an immense number of doses to respond to demand, global
health partners worked to establish the COVAX Facility. The COVAX
Facility is a global risk-​sharing mechanism for pooled procurement and
equitable distribution of COVID-​19 vaccines to countries at all income
levels (Gavi, 2021a). It uses both push and pull mechanisms, including
an Advance Market Commitment for supply to lower-​income countries,
to incentivize manufacturers to develop COVID-​19 vaccines according
to a specific target product profile and to supply resulting prod-
ucts equitably at affordable prices and in adequate quantities to meet
global need.

1.3.4.4.2 Volume guarantees
In volatile or uncertain markets, a lack of clear demand signals can disin-
centivize development or production of vaccines by suppliers. In these situ-
ations, volume guarantees—​where a predetermined volume of vaccines is
guaranteed to be procured or the delta in forecast revenues will be paid to the
supplier at the end of a predetermined price window—​can be used. An ex-
ample of a volume guarantee comes from an agreement between the BMGF
and a supplier of rotavirus vaccines, secured in 2010–​2011. Through guaran-
teeing 132 million doses of sales to low-​and lower-​middle-​income countries,
the supplier was able to lower the cost of production per dose, and lower
some of the risk premium that was initially factored into the pricing—​thanks
to increased predictability and amortizing costs over a larger volume. This
allowed a 67% price reduction and a price of just $5 per immunized child
(Gavi, 2012).

1.3.4.5 Other market-​shaping levers

Alongside tools to shape procurement, other market-​shaping levers that do

not commit to procurement can be applied.

• Push funding (or product development investments). Suppliers may be able

to receive investment (e.g., in the form of grants or attractive loan terms)
to conduct clinical trials or expand manufacturing capacity for high-​
priority vaccines (e.g., from the BMGF, International Vaccine Institute,
or other large donors). By subsidizing the cost of development, donors
44 Cost of finding and making vaccines

may put in place global access terms that set a ceiling price and/​or a min-
imum production commitment to serve LMICs if successful. This form of
market shaping is the most common.
• Product development partnerships (PDPs). In addition to the Meningitis
Vaccine Project, the oral cholera vaccine was developed via a PDP.
Partnerships between the International Vaccine Institute, suppliers across
Sweden, Vietnam, India, and South Korea, alongside public and private
funding accelerated the development of the oral cholera vaccine for the
public sector (Odevall et al., 2018). PDPs are becoming less common,
with a trend towards push funding.

1.3.5 Conclusion

For public health systems to ensure access to affordable vaccines, it is impor-

tant that policymakers and decision makers understand what factors drive
vaccine pricing, and supply sustainability. This chapter has demonstrated how
firms’ costs contribute to vaccine prices, and how market conditions also af-
fect vaccine prices. Vaccine markets do not conform to the criteria for com-
petitive markets due to a high degree of consolidation on both the supply and
demand sides, high barriers to entry and exit, and imperfect information.
This chapter has also provided an overview of the vaccine supply landscape,
to provide an understanding of the range of vaccine companies and prod-
ucts that operate within vaccine markets. Suppliers often have power as price
setters rather than price takers in vaccine markets. However, the development
of pooled procurement systems—​ particularly focusing on lower-​ income
countries—​has enabled demand-​side levers to lower prices and increase ac-
cess to affordable vaccines globally.

References
Access to Medicine Foundation. (2017). Access to Vaccines Index. https://​acc​esst​omed​icin​efou​
ndat​ion.org/​acc​ess-​to-​vacci​nes-​index
André, F. E. (2002). How the research-​based industry approaches vaccine development and es-
tablishes priorities. Developments in Biologicals, 25–​29.
Batson, A. (2016, March 15). Global vaccine market. Paper presented at the Global Vaccine and
Immunization Research Forum, Johannesburg (South Africa).
Bill & Melinda Gates Foundation. (2016). Production economics for vaccines. Seattle, WA: Bill &
Melinda Gates Foundation.
Chalkidou, K., Garau, M., Silverman, R., & Towse, A. (2020). Blueprint for a market-​driven
value-​based advance commitment for tuberculosis. Center for Global Development.
 1.3.5 Conclusion 45

https://​w ww.cgdev.org/​publ​icat​ion/​bluepr​int-​mar​ket-​dri​ven-​value-​based-​adva​nce-​com​
mitm​ent-​tuber​culo​sis
Cole, C. B., & Iyer, J. K. (2016). Ensuring sustained incentives for pharma to develop medicine for
the poor. Access to Medicine Foundation. https://​acc​esst​omed​icin​efou​ndat​ion.org/​media/​
uplo​ads/​downlo​ads/​5bc​5edf​d537​2e_​2​016-​Ensur​ing-​sustai​ned-​inc​enti​ves-​for-​pha​rma-​to-​
deve​lop-​medic​ine-​for-​the-​poor.pdf
Gavi. (2012). GAVI Alliance secures lower price for rotavirus vaccine [Press release]. https://​
www.gavi.org/​gavi-​secu​res-​lower-​price-​rotavi​rus-​vacc​ine
Gavi. (2018). Eligibility and transition policy. https://​www.gavi.org/​pro​gram​mes-​imp​act/​progr​
amma​tic-​polic​ies/​elig​ibil​ity-​and-​transi​tion​ing-​pol​icy
Gavi. (2020). Market shaping. https://​www.gavi.org/​our-​allia​nce/​mar​ket-​shap​ing
Gavi. (2021a). COVAX. https://​www.gavi.org/​covax-​facil​ity
Gavi. (2021b). Pneumococcal vaccine support. https://​www.gavi.org/​types-​supp​ort/​vacc​ine-​
supp​ort/​pneum​ococ​cal
Gavi. (2021c). Supply agreements. https://​www.gavi.org/​invest​ing-​gavi/​inn​ovat​ive-​financ​ing/​
pneum​ococ​cal-​amc/​manufa​ctur​ers/​sup​ply-​agr​eeme​nts
GlaxoSmithKline. (2019). Tiered pricing and vaccines. https://​www.gsk.com/​media/​5683/​tie​
red-​pric​ing-​and-​vacci​nes-​aug19.pdf
Heaton, P. M. (2020). Challenges of developing novel vaccines with particular global health im-
portance. Frontiers in Immunology, 11, 517290. doi:10.3389/​fimmu.2020.517290
Médecins Sans Frontières. (2010). Giving developing countries the best shot: An overview of vac-
cine access and R&D. https://​msfacc​ess.org/​giv​ing-​dev​elop​ing-​countr​ies-​best-​shot-​overv​
iew-​vacc​ine-​acc​ess-​and-​rd
National Immunization Technical Advisory Groups. (2021). The unique database on National
Immunization Technical Advisory Groups. https://​www.nitag-​resou​rce.org/​
Odevall, L., Hong, D., Digilio, L., Sahastrabuddhe, S., Mogasale, V., Baik, Y., . . . Lynch, J.
(2018). The Euvichol story—​development and licensure of a safe, effective and affordable
oral cholera vaccine through global public private partnerships. Vaccine, 36(45), 6606–​6614.
doi:10.1016/​j.vaccine.2018.09.026
Pan American Health Organization. (2020). PAHO strategic fund. https://​www.paho.org/​en/​
paho-​strate​gic-​fund
PATH, & World Health Organization. (2017). Meningitis vaccine project. https://​www.mening​
vax.org/​
Pronker, E. S., Weenen, T. C., Commandeur, H., Claassen, E. H. J. H. M., & Osterhaus, A. D. M.
E. (2013). Risk in vaccine research and development quantified. PloS One, 8(3), e57755.
doi:10.1371/​journal.pone.0057755
Serum Institute of India, & PATH. (2020). Fact sheet: Pneumococcal disease, pneumococcal
conjugate vaccines, and Pneumosil®. https://​pneumo​sil.com/​wp-​cont​ent/​uplo​ads/​2020/​07/​
path-​fact-​sheet.pdf
Siripitayakunkit, U. (2018). Procurement policies—​an example for ensuring vaccine security: The
ASEAN vaccine security and self-​reliance initiative. https://​doku​men.tips/​downl​oad/​link/​
proc​urem​ent-​p olic​ies-​a-​an-​exam​ple-​for-​ensur​ing-​vacc​ine-​uncha ​lee-​sirip​itay​akun​kit.
Accessed June 26, 2022.
UNICEF. (2021). Handling fees. https://​www.uni​cef.org/​sup​ply/​handl​ing-​fees
Vidor, E., & Soubeyrand, B. (2016). Manufacturing DTaP-​ based combination vaccines:
Industrial challenges around essential public health tools. Expert Review of Vaccines, 15(12),
1575–​1582. doi:10.1080/​14760584.2016.1205492
Wolf, J., Bruno, S., Eichberg, M., Jannat, R., Rudo, S., VanRheenen, S., & Coller, B.-​A. (2020).
Applying lessons from the Ebola vaccine experience for SARS-​CoV-​2 and other epidemic
pathogens. NPJ Vaccines, 5(1), 51. doi:10.1038/​s41541-​020-​0204-​7
46 Cost of finding and making vaccines

World Health Organization. (2018). Vaccine pricing: Gavi fully self-​financing & accelerated
transition countries. https://​cdn.who.int/​media/​docs/​defa​ult-​sou​rce/​immun​izat​ion/​mi4a/​
factsheet_​vacc​_​pri​cing​_g​ av​i_​tr​ansi​tion​ing.pdf?sfv​rsn=​cc0​e556​6_​6&downl​oad=​true
World Health Organization: Immunization Vaccines and Biologicals. (2019). MI4A: Market
information for access to vaccines. https://​cdn.who.int/​media/​docs/​defa​ult-​sou​rce/​immun​
izat​ion/​mi4a/​factsheet_​vacc​_​pri​cing​_​gav​i_​tr​ansi​tion​ing.pdf?sfv​rsn=​cc0​e556​6_​6&downl​
oad=​true
1.4
Vaccination as investment in
human capital
J. P. Sevilla, David Bloom, Dan Salmon, and David Bishai

We present the economic view of vaccinations as investments in human cap-

ital. That is, they involve expenditures at a point in an individual’s life that
produce a stream of future benefits in terms of time spent alive with a high
quality of life. These benefits have both intrinsic and instrumental value.
The intrinsic value is the direct benefit of being healthy. The instrumental
value includes the broad socioeconomic value of health as an input to achieve
other goals. We discuss how conceptions of the value of vaccination have
broadened in recent years from an initial narrow focus on health impacts
on vaccinated individuals related to the averted treatment costs to the full
public health and socioeconomic value of vaccination. We discuss how the
broadened benefits from vaccine investments may not be captured by indi-
vidual rational decision-​making and how this leads to market failures. The
decision-​making challenges include the costs of rational decision-​making
encompassing misinformation, dynamic inconsistency, the use of heuristics,
motivated cognition, and vaccine hesitancy. The market failures involve ex-
ternalities, public goods, and merit goods. Important externalities include
herd immunity and macroeconomic effects. These interactions imply that
socially optimal vaccination requires both strong public sector involvement
and pro-​vaccination social norms. Public sector involvement includes public
financing, subsidies and incentives, guidance, and mandates. The social
norms involve altruism, personal and collective responsibility, and commit-
ments not to free ride. Cultivating such norms requires non-​economic policy
instruments.
We discuss how existing vaccine valuation frameworks often adopt the
health payer’s perspective and cost–​utility analysis (CUA), which can yield
to undervaluation of, and underinvestment in, vaccination. This can be

J. P. Sevilla, David Bloom, Dan Salmon, and David Bishai, Vaccination as investment in human capital In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0005
48 Vaccination as investment in human capital

remedied by wider use of societal perspective cost–​benefit analysis (CBA) and

of social welfare functions to address equity issues. Finally, we discuss optimal
investment in research and development (R&D), and how this raises the chal-
lenge of incentivizing innovation, promoting access and equity, and reducing
deadweight losses and rents, all at the same time.

1.4.1 Health as human capital, and vaccinations as

investments in human capital

Economists view health as a form of human capital (Becker, 2007), a stock

or asset embodied in the states of our bodies and minds that yields a fu-
ture stream of benefits in terms of time spent alive and free from morbidity
or disability. It is a stock whose monetary value equals the expected present
discounted monetary value of each moment within that stream. Health is a
central element of human capital, along with education and work skills. Like
any asset, health can depreciate through neglect, disease, poor living condi-
tions or habits, aging, or other factors, thereby reducing the flow of benefits.
But it can also be maintained and augmented through investments in the form
of the provision of various sorts of security (e.g., physical and socioeconomic);
good diet, exercise, and health habits; healthy environments (e.g., access to
clean water and air); and the appropriate use of healthcare and health tech-
nologies, including vaccination. Many investments, such as vaccinations, are
costly, and so appropriate levels of investment require quantifying and com-
paring such costs and benefits and undertaking those investments for which
benefits exceed costs. These costs and benefits can be computed at the level
of an individual, in which case any investment whose individual benefits ex-
ceed its individual costs can be called individually optimal. But these can also
be computed at the level of society as a whole, in which case any investment
whose societal benefits exceed its societal costs is socially optimal.
The word “investment” might suggest that vaccinations be valued for their
impact on market-​related outcomes such as wages, earnings, or gross do-
mestic product. As we discuss below, such impacts are indeed part of vaccin-
ations’ value and often ignored in vaccine evaluation. Therefore, these are an
important part of the economic view. However, it is equally important that
economic theory, and rational choice theory more generally, defines value ex-
pansively: in terms of the satisfaction of individuals’ preferences. Any situa-
tion an individual would want to occur—​for example, having a long life; being
able to spend time with friends and family; pursuing meaningful work or life
 1.4.1 Vaccination as an investment in human capital 49

projects whether paid or not; being happy, enjoying pleasurable activities and
experiences, and minimizing pain; being active, free, independent, and self-​
determining; or anything else whether market-​related or not—​is a relevant
aspect of value. Thus, the value of an investment in health is not limited to an
investment’s impact on market-​related outcomes but extends to any impact
individuals or societies care about.
The benefits of investment in health capital reflect health’s intrinsic value
(i.e., the extent to which health is valuable in and of itself to the individual
whose health is at issue) and instrumental value (i.e., the extent to which im-
proved health produces valuable socioeconomic and other non-​health out-
comes, whether for the individual or for society as a whole). The best way
within economic theory to represent and distinguish these values is through a
health-​augmented lifecycle model. Let us think of a person’s lifetime utility (we
treat utility and well-​being as synonyms, which is not uncontroversial), U, as
depending on lifetime trajectories in longevity prospects, s (where “s” stands
for survivorship), health-​related quality of life, q, consumption of goods and
services, c, and non-​market time, l. We can represent this by a function:
Equation 1.4.1. A person’s lifetime utility:

U ( s, q, c , l )

Non-​market time in turn consists of time spent on unpaid work and on lei-
sure, where unpaid work includes such activities as housework, caregiving,
and volunteering. The arguments of U are trajectories over a lifetime as op-
posed to levels at any point in that lifetime. U is a positive function of these
trajectories: the higher these trajectories, the better. U is typically assumed to
be concave in consumption and non-​market time: stable and certain trajec-
tories are superior to cyclical or risky ones. These four arguments are typically
assumed to be natural complements or mutual enhancers of each other: the
higher the level of any one quantity, the greater the value of improvements in
any of the others (the way the value of having a left glove is enhanced by also
having a right glove).
Healthier people can have higher levels of both consumption and non-​
market time. For example, higher health-​related quality of life allows a person
to be more productive during a workday and earn and consume more over
a lifetime. A longer life allows more leisure time, for example, to build rela-
tionships with one’s grandchildren. Health can also contribute to stabilizing
consumption and non-​market time such as reducing exposure to the often
significant and sometimes catastrophic out-​of-​pocket costs of illness or risks
50 Vaccination as investment in human capital

of enforced inactivity when disabled. In recognition of these mechanisms

for health’s impact on consumption and non-​market time, we can rewrite
Equation 1.4.1 thus:
Equation 1.4.2. A person’s lifetime utility (detailed):

( )
U s, q, c ( s, q ) , l ( s, q )

In this fuller statement, health matters intrinsically: s and q are direct ar-
guments in the utility function. But it also matters instrumentally: s and q
determine the level, stability, and certainty of lifetime consumption and non-​
market time. Thus, properly valuing health requires measuring its intrinsic
value (how much individuals value health in and of itself) and its instrumental
value (how it affects levels and variability over a lifetime of consumption and
non-​market time). Health’s impact on market-​related outcomes such as earn-
ings (the sick are less able to work, and often earn less from work) are therefore
part of the value of health because earnings facilitate consumption. However,
market-​related outcomes do not tell the whole story. Health’s intrinsic value,
its impact on unpaid work and leisure, and the natural complementarities
among health, consumption, and non-​market time also matter.
The lifecycle aspect of this picture requires emphasis: vaccinations yield
their benefits not only during the period immediately following their occur-
rence but potentially for the entire lifespan. A vaccine’s effectiveness often
exists for only a few years (e.g., a decade). Nevertheless, if it prevents death or
severe disability during those years, this potentially raises survival probabil-
ities, health-​related quality of life, and economic productivity for significant
and much longer chunks of the vaccinated person’s lifetime.
The above theoretical picture can and should be extended beyond the indi-
vidual being vaccinated to other individuals in society. For example, vaccina-
tion of an individual confers herd protection on the unvaccinated by reducing
the number of people from whom the latter might catch a particular infection.
The reduced mortality and morbidity risks enjoyed by the unvaccinated from
herd protections have the same intrinsic and instrumental values as those of
the vaccinated. Such protection is particularly valuable to those unable to be
vaccinated because they are too old or young, immunocompromised, or lack
access.
Finally, vaccinations can confer wholly instrumental benefits to commu-
nities and societies. Perhaps the most important and salient contemporary
example of this is the coronavirus disease 2019 (COVID-​19) pandemic,
which the world is living (and dying) through as this chapter is being written.
 1.4.2 From narrow to broad conceptions of the value of vaccination 51

Addressing infectious disease outbreaks like the global COVID-​19 pan-

demic often requires shutting down broad swathes of socioeconomic activity,
causing severe declines in economic activity, income, and employment, and
dramatic increases in poverty and economic vulnerability. Vaccination re-
duces the need for such shutdowns, thus protecting people’s livelihoods (and
therefore their consumption) over and above any health benefits such vaccin-
ations produce.

1.4.2 From narrow to broad conceptions of the value

of vaccination

Historically, the field of vaccine valuation has had an excessively narrow focus.
To the extent that health was seen as an important value element, vaccine
evaluations adopted a “therapeutic paradigm,” focusing only on the health
impacts on the vaccinated individual, and impacts related to the vaccine’s
target pathogen (e.g., measles-​related outcomes in measles vaccine recipients)
(Gessner et al., 2017). From an economic perspective, the main element of
value was seen to be the averted costs to the health system of treating disease.
From the economic perspective, health expenditures were not explicitly seen
as producing streams of economic benefits extending into the future, but im-
plicitly seen as consumption expenditures rather than investments (Bloom &
Canning, 2000).
The value focus has gratifyingly expanded in recent times. With respect
to health benefits, there is a growing appreciation of the full public health
value of vaccination (Gessner et al., 2017). A core element of this full value
are the herd protections mentioned above. Since infections often result in
hospitalization, vaccination also reduces difficult-​and expensive-​to-​treat
nosocomial (i.e., hospital-​based) infections such as Clostridium difficile,
Staphylococcus aureus, Klebsiella, and Escherichia coli. Some vaccines
have nonspecific (sometimes also called heterologous or off-​target) effects
(Saadatian-​Elahi et al., 2016). For example, measles-​containing vaccine ap-
pears to promote general immune function and reduce all-​cause mortality
(Mina et al., 2019; Mina, Metcalf, de Swart, Osterhaus, & Grenfell, 2015).
Childhood bacillus Calmette–​Guérin vaccination is associated with dimin-
ished rates of lung cancer at older ages (Usher et al., 2019). Vaccination can
reduce or slow down the progress of antimicrobial resistance in both targeted
and bystander pathogens, either directly by preventing resistant infections or
indirectly by reducing antibiotic treatments (Sevilla, Bloom, Cadarette, Jit, &
52 Vaccination as investment in human capital

Lipsitch, 2018). Reduced use of antibiotics can also protect human health by
preserving the gut microbiome (Cully, 2019). Severe disease case definitions
often do not exist—​for example, severe pneumonia is typically defined in
terms of requiring hospitalization—​so appreciating vaccinations’ full health
benefits requires appreciating the averted health burdens of severe disease
(Wilder-​Smith et al., 2017). Other elements of full public health value are di-
sease control, including their elimination and eradication (which eliminates
all downstream burdens of disease and vaccination costs; Xue & Ouellette,
2020); health system strengthening; and health equity (i.e., reductions in dif-
ferences across socioeconomic/​demographic groups in health outcomes and
access and use of health services).
There is also a growing appreciation of the full socioeconomic value of vac-
cination. Vaccination can reduce various socioeconomic burdens of the acute
stage of the disease, from the potentially catastrophic out-​of-​pocket costs paid
by patients and their families and friends and caregivers to the economic costs
(in terms of lost earnings and foregone uses of non-​market time) of time lost
seeking care, providing care, or recuperating. Over longer time horizons, vac-
cination is an enabler and driver of human capital accumulation, productivity,
labor supply, income generation, and poverty alleviation. Vaccinated children
tend to have better records of school attendance, higher levels of educational
attainment, and better cognitive function—​as evidenced, for example, in
studies focused on a variety of vaccines (e.g., for measles, tetanus, Haemophilus
influenzae type b, rotavirus, and pneumococcal disease) and different country
contexts, including Bangladesh, China, Ethiopia, India, the Philippines, South
Africa, and Vietnam (Anekwe, Newell, Tanser, Pillay, & Bärnighausen, 2015;
Bloom, Canning, & Shenoy, 2011; Canning et al., 2011; Megiddo, Klein, &
Laxminarayan, 2018; Nandi, Deolalikar, Bloom, & Laxminarayan, 2019;
Nandi, Shet, et al., 2019; Oskorouchi, Sousa-​Poza, & Bloom, 2020). Studies
have also found that vaccinated individuals have higher trajectories of pro-
ductive market and productive non-​market activities (Bloom, Khoury, Algur,
& Sevilla, 2020) than their (otherwise comparable but less healthy) unvacci-
nated counterparts (Sevilla et al., 2020; Sevilla et al., 2019).
Vaccination thereby translates into higher levels of consumption and sav-
ings at the micro level and more rapid economic growth at the macro level
(Masia, Smerling, Kapfidze, Manning, & Showalter, 2018). The parents of vac-
cinated children also tend to have better records of work attendance and pro-
ductivity, as well as greater peace of mind associated with reduced health and
related financial risks associated with infectious disease.
 1.4.3 Individual decision-making 53

The social benefits of vaccination include improvements in health and so-

cial equity and the distribution of economic well-​being. Such benefits can
arise insofar as vaccination prevents diseases that disproportionately afflict
the poor and disadvantaged racial and ethnic minorities. Vaccination can also
confer intergenerational benefits. For example, to the extent that vaccination
of adolescents against human papillomavirus affords protection against sub-
sequent development of cervical cancer, maternal survival will be improved
with concomitant benefits for the nurturing and development of children and
disruptions to intergenerational impulses regarding the transmission of pov-
erty (Bärnighausen, Bloom, Cafiero, & O’Brien, 2012; Riumallo-​Herl et al.,
2018; Verguet et al., 2015).

1.4.3 Individual decision-​making

Whether a particular vaccination is individually optimal depends on the rela-

tive size of the vaccine’s individual benefits (in terms of its individual intrinsic
and instrumental values) and individual costs. When a vaccine is part of the
national immunization program (NIP) such as an infant measles, mumps, and
rubella vaccine, individuals don’t need to pay for vaccines out of pocket when
they get vaccinated. Instead, such vaccines are financed by a NIP that is in turn
financed by taxes or social health insurance premiums. For such vaccines, the
individual costs are largely the costs in terms of time, effort, and funds re-
quired to make an appointment, travel to a clinic, or take time off work; risk
of adverse reactions to the vaccine; and discomfort and distress resulting from
muscle aches, nausea, and a fear of needles. Of course, when vaccines are not
in the NIP, such as influenza vaccines for working age adults without risk fac-
tors, the individual costs will include the out-​of-​pocket costs for the vaccine
doses and administration. Later, we will see how cognitive dissonance, and
certain moral preferences and attitudes can also make vaccination costly to
the individual. The magnitude of the expected benefits to the individual are a
function of many of the various intrinsic and instrumental values described
above that accrue to vaccinated individuals. They are also a function of the
magnitude of the risk of infection, which in turn is a function of the incidence
or prevalence of the disease and the effectiveness of the vaccine. If individual
costs are sufficiently high (e.g., because vaccination clinics are not conven-
iently located and so are costly in terms of time, effort, and funds to access)
and benefits sufficiently low (e.g., because incidence or vaccine effectiveness
54 Vaccination as investment in human capital

or severity of infection are low), then it would be individually rational not to

get vaccinated.
Recall that vaccination is socially optimal if its social benefits (i.e., indi-
vidual benefits summed across all individuals in society) exceed its social
costs. Not all vaccinations are socially optimal: if infection risk, vaccine effi-
cacy, and mortality and morbidity risks of infection are low or the vaccine is
costly, then vaccination may not be individually or socially optimal.
From society’s point of view, it would be ideal if individually optimal vac-
cination decisions coincided with socially optimal ones, in which case society
could take a hands-​off approach and simply leave vaccination choices to indi-
viduals. But there are multiple reasons why individual decisions may diverge
from socially optimal ones.
First, even if we were to make the extreme assumption that individuals are
perfectly (in the sense of being relatively well informed and capable of bearing
the cognitive burden of thinking through complex decisions) and self-​
interestedly rational, externalities would often result in vaccination incentives
faced by individuals falling short of what is socially optimal. Externalities are
various consequences of a decision maker’s choices that are borne by others
in society and that for self-​interested reasons decision makers do not factor
into their decisions. Central externalities related to vaccination are herd pro-
tections, reductions in collectively financed treatment costs, benefits to the
rest of the world and to future generations of disease eradication, and indeed
most elements of the full public health and socioeconomic value of vaccina-
tion that are realized at a macro level (e.g., at the level of a community, society,
or population).
Second, perfectly rational decision makers may, again for self-​interested
reasons, prefer to free ride. Herd protections are an example of what econo-
mists refer to as a pure public good. A pure public good is one that is both
nonexclusive and non-​rival. A nonexcludable good is a good for which there
cannot be selective access. A non-​rival good is one such that one person’s use
of the good does not deplete the stock available for others to use. Herd immu-
nity is both non-​excludable and non-​rival. If herd protections are achieved, it
is impossible to exclude anybody from enjoying them. And once the threshold
for achieving any community immunity has been achieved, the marginal cost
of letting one more person enjoy community immunity is zero.
The non-​excludability of a public good gives people an incentive to free
ride. According to economic theory, producing an efficient amount of a
public good requires each person to contribute to financing the production
of that good by an amount equal to their willingness to pay for it. However,
 1.4.3 Individual decision-making 55

non-​excludability implies that people will not have an incentive to make those
contributions. From a selfish perspective, it is better to let others contribute
since once the public good has been produced, since it will be available to
non-​contributors as much as to contributors. Thus, it may be rational for indi-
viduals to forego vaccination, thereby avoiding the inconveniences and time
and effort required, and hope that enough other people get vaccinated to pro-
vide everyone with protection. If there are sufficiently many free riders, vac-
cination uptake may fail to reach the threshold required to produce the herd
effects, jeopardizing the production of the public good.
Third, rational decision-​making about health is informationally and cogni-
tively demanding. An individual must know or research probabilities of infec-
tions and of death and various kinds of disabilities that result from infection,
the impact on quality of life and economic well-​being of various disabilities,
what actions they can take to prevent or treat disease, and the costs and effec-
tiveness of those actions. It takes a lot of time, effort, funds, and thought to
make informed rational choices. Facing those demands, we may end up being
paralyzed into inaction.
Fourth, individuals fall prey to systematic departures from rationality in
three areas: biases in the perception or understanding of information, the
use of heuristics (or rules of thumb or cognitive shortcuts), and self-​control
problems (Sassi et al., 2015). Self-​control problems (also referred to as time
inconsistency and present bias) can lead individuals to overweight today’s
costs and benefits (such as the inconvenience of getting vaccinated today)
over tomorrow’s (such as reduced infection risk) (O’Donoghue & Rabin,
2015). (Such bias is distinct from discounting since one may prefer to get vac-
cinated tomorrow rather than face a higher infection risk 2 days from now.)
Omission biases lead individuals to prefer potentially harmful inactions (such
as a vaccine-​preventable infection) to potentially harmful actions (such as
vaccine-​induced adverse effects) (Ritov & Baron, 1990). Ambiguity aversion
or dread leads us to prefer known risks (risk of infection) to unknown risks
(risks of adverse events) (Han, Reeve, Moser, & Klein, 2009). Affect heur-
istics make individuals’ risk judgments depend more on positive or negative
feelings toward some aspect of the risk than objective statistical informa-
tion about disease risks and vaccine benefits (Betsch, Ulshöfer, Renkewitz, &
Betsch, 2011). The availability heuristic makes individuals mistakenly judge
outcomes of which they are aware (e.g., from acquaintances’ experience or
media reports) as more probable (Blumenthal-​ Barby & Krieger, 2015).
Optimism bias leads people to overestimate the probability of good outcomes
(such as not getting infected) and underestimate the probability of bad ones
56 Vaccination as investment in human capital

(such as getting infected) (Sharot, 2011). A naturalness bias leads individuals

to prefer actions and outcomes perceived to be more natural (e.g., natural im-
munity) than those seen to be artificial, synthetic, or processed (e.g., vaccine-​
induced immunity) (daCosta DiBonaventura & Chapman, 2008).
Confirmation biases lead people to underweight information that conflicts
with what they already believe, and motivated cognition leads individuals to
underweight information that conflicts with what they want to believe (Stone
& Wood, 2018). Both are forms of cognitive dissonance and can lead individ-
uals to be overly sensitive to information about adverse events and resistant to
information about vaccine benefits. The Dunning–​Kruger effect can lead indi-
viduals to overestimate their own expertise (e.g., on the causes of autism) and
underestimate those of others (like professional experts) (Motta, Callaghan,
& Sylvester, 2018). Moral foundations theory suggests that moral prefer-
ences (e.g., commitments to individual liberty) can shape empirical judg-
ments (Rossen, Hurlstone, Dunlop, & Lawrence, 2019). The “attitude roots”
literature emphasizes how receptivity to scientific evidence can be shaped by
underlying attitudes such as world views (especially individualistic and lib-
ertarian ones), conspiratorial ideation, vested interests, personal identity ex-
pression, social identity needs, fears and phobias (e.g., of needles), and disgust
sensitivity (Hornsey & Fielding, 2017; Hornsey, Harris, & Fielding, 2018).
The above four decision-​making obstacles contribute to vaccine hesitancy,
which according to the World Health Organization is a top ten threat to global
health (World Health Organization, 2019a) and “as contagious and dangerous
as the diseases it helps to spread” (World Health Organization, 2019b). The
World Health Organization defines the vaccine hesitant as “those who refuse
some or all vaccines, delay some vaccine perhaps according to an ‘alterna-
tive schedule’, or accept all vaccines but remain concerned” (World Health
Organization, 2014). Vaccine hesitancy should not be considered a single
state but rather a spectrum of combinations of behaviors and attitudes (Dubé,
Ward, Verger, & MacDonald, 2021). At one extreme are individuals who (be-
haviorally) reject all vaccines and (attitudinally) are very sure of their views,
and at the other extreme are those who (behaviorally) accept all relevant vac-
cines and are equally (attitudinally) very sure of their views. In between these
extremes are many intermediate behavioral and attitudinal combinations.
Behaviorally, individuals may accept or reject some but not all vaccines,
or they may delay or be genuinely undecided regarding certain vaccines.
Attitudinally, people may vary in the certainty of their acceptance or rejection.
Only those who reject all vaccines and are highly certain in such rejection
can be thought of as hardcore anti-​vaxxers. Anyone who accepts at least one
 1.4.4 The government’s role in optimizing investment in vaccination 57

vaccine, who entertains some uncertainty about any refusal, or who merely
delays or is genuinely undecided, is potentially susceptible to pro-​vaccine
messaging. Attitudes can reflect concerns about vaccines, and even those who
get vaccinated can be concerned about such vaccinations.
A fifth set of reasons individual vaccination choices may fall short of so-
cially optimal ones relate to equity or distributional issues. Infectious disease-​
related risks and burdens are distributed unequally in society, often reflecting
and interacting with socioeconomic/​demographic inequalities, so that the
poor and vulnerable often face higher risks and burdens. Many societies have
strong commitments to reducing so-​called socioeconomic gradients in health,
that is, disparities in health outcomes and in access to and use of health serv-
ices across socioeconomic/​demographic groups. Vaccines, like many other
health technologies, are also held by many to be merit goods for which access
ought not to depend on ability to pay. Individuals’ health-​related decisions
are influenced largely by self-​interest, and the most important redistributive
mechanisms are those within the capacity of governments, so we expect gov-
ernment as opposed to individual action to be the primary mechanisms for
achieving equity-​related goals. Merit goods also inherently require redistribu-
tive funding mechanism like taxes or subsidies (from those with ability to pay
to those who do not) and cannot rely for financing exclusively or even prima-
rily on out-​of-​pocket expenditures since these are, of course, constrained by
ability to pay.

1.4.4 Optimal investment in vaccination cannot be left

to markets but requires a strong government role

For reasons given above, we cannot expect individuals to invest socially op-
timal amounts in vaccination. We therefore cannot leave vaccination deci-
sions to individuals and the market. Getting vaccinated is very different from,
say, buying a pizza. Pizza purchases and consumption are typically left to de-
centralized market decisions with only two parties: buyers and sellers dealing
directly with each other. Each person decides if they want pizza, what kind,
and at what price, and is assumed perfectly capable of making these judgments
for themselves and their children. Transactions between buyer and seller are
voluntary, and so mutually beneficial. Consumers finance pizza production
directly out of pocket. Prices are market determined, reflecting countless
decentralized decisions. Governments don’t regulate pizza prices to protect
consumers since competition among pizza makers tends to keep prices low.
58 Vaccination as investment in human capital

When markets determine aggregate outcomes—​who gets pizza and who

doesn’t, who produces it, the reigning prices, and the quantities sold—​we
seldom worry that these outcomes fall short of some social goals. Innovation
in pizza-​making technology is not a matter of profound social concern, and
we are quite happy to let market forces dictate its pace.
Vaccination is very different. Markets play much smaller roles and gov-
ernments much bigger roles as intermediaries. Few countries leave vaccine
choices entirely to people, instead relying on expert judgment as reflected
in vaccination guidelines, mandates, and advice of healthcare professionals.
Vaccine mandates compel people to act against their own preferences re-
garding their own health and that of their children, in part to solve the
free-​rider problem. This raises issues involving the balance of public health
imperatives and the greater good on the one hand and freedom and self-​
determination on the other. Vaccine financing involves government setting
and collecting taxes or social insurance premiums. Prices paid to suppliers
often reflect bilateral negotiations between manufacturers and large regional
or national vaccine purchasers. Who is eligible for vaccinations and with what
copay are determined by public health officials and government and payer
reimbursement decisions. Aggregate outcomes are scrutinized for efficiency
(e.g., are enough people being vaccinated?) and equity (do the worse off have
meaningful access?). Promoting vaccines innovation is often a significant
global priority, forcing trade-​offs between strengthening patent rights to pro-
mote innovation and weakening them to promote access.
Government policies attempt to shift individual perceptions of the net
benefits of vaccination by lowering the cost to obtain vaccines and supplying
information to raise demand. Costs of vaccine are lowered not only by pro-
viding them for zero out-​of-​pocket monetary costs, but also by offering con-
venient venues and low waiting time. Sometimes there are conditional cash
transfers to compensate people for the effort and time taken to get vaccin-
ated. Vaccine information campaigns use multiple channels including media,
medical professionals, schools, social media, and faith-​based initiatives to
communicate facts about the safety and benefits of vaccines. Policy incen-
tives to address heuristics surrounding benefit and risk perception are still an
evolving area of communications research. Setting up narratives and stories to
counter concerns that vaccines have dreaded and unknown properties needs
to be done with care. Widely disseminating messages to the public in an im-
pactful manner is challenging in many ways, including the need to individ-
ually tailor such messages. As described by the World Health Organization,
“messages need to be tailored for the specific target group, because messaging
 1.4.5 Do governments invest enough in vaccination? 59

that too strongly advocates vaccination may be counterproductive, reinfor-

cing the hesitancy of those already hesitant” (Dubé et al., 2021).

1.4.5 Do governments invest enough in vaccination?

For reasons given above, we cannot expect individuals to invest socially op-
timal amounts in vaccination. Thus, governments must play a central role in
such investments. But there are challenges to socially optimal government in-
vestment as well.
Consider three different layers of governmental decision-​making. First,
in many countries, a Department of Health (DOH) is often allocated a fixed
budget by the Department of Finance (DOF), and the DOH must in turn allo-
cate that health budget across the different health sector activities, including
vaccination. The fixed budget implies that spending more on vaccines requires
spending less on other activities. The second layer involves the DOF, which
often operates with a fixed public sector budget set by the tax-​and-​transfer
policies set by legislatures. The DOF can expand the DOH budget, allowing
the DOH to, say, spend more on vaccines without having to sacrifice other
health expenditures. However, the fixed public sector budget implies that such
expansion requires the DOF to allocate less to other non-​health departments
such as education, infrastructure, or social assistance. The third layer involves
the legislature, which has the power to set taxes and therefore determine what
share of national income gets reallocated to the DOF and what stays in private
hands. It can raise taxes, thereby facilitating higher public spending, but at the
expense of reducing households’ after-​tax income.
These three decision makers directly and indirectly affect levels of invest-
ment in vaccination. The DOH can spend more on vaccination but less on
other health technologies. The DOF can expand the DOH budget, allowing
vaccination spending to rise while protecting other health spending, but at the
cost of foregoing non-​health sector public spending. And legislatures can ex-
pand the DOF budget, relaxing health versus non-​health trade-​offs in public
spending, but reducing household disposable income.
To achieve socially optimal decisions, each of these three decision makers
should compute the full social costs and benefits of each of its options and
then prioritize those options according to which yields the largest net social
benefit per dollar spent from the relevant budget. Thus, to achieve optimal
public spending on vaccination, we should measure the net social benefit
of vaccination per dollar spent out of the DOH budget (or equivalently, the
60 Vaccination as investment in human capital

social rate of return to vaccination). The DOH should compare this social rate
of return to that of other health spending, the DOF should compare it to that
of non-​health sector public spending, and the legislature should compare it to
the social rate of return of extra household post-​tax income.
Optimal public spending at all levels thus requires quantitative estimates
of the full social benefits and costs per DOH dollar spent on vaccination.
Such estimates must include all the elements of the intrinsic and instrumental
values described earlier, including the full public health and socioeconomic
values of vaccination. It also requires the three governmental decision makers
to allow such social value estimates to drive their decisions. These are easier
said than done.
Many DOHs, when deciding whether to include a vaccine in a NIP, will
conduct an economic evaluation of that vaccine. Performing such evaluation
requires making two specification choices. The first is the choice of perspec-
tive, the most important options being the health payer and societal perspec-
tive. The second is the choice of analysis, the most important options being
CUA and CBA.
The health payer’s perspective narrowly focuses only on two aspects of the
value of vaccines: their health impacts (typically measured in quality-​adjusted
life years (QALYs) or disability-​adjusted life years (DALYs)) and conse-
quences for the DOH’s budget (often the costs of the vaccination program
offset by averted costs of treating infections and controlling outbreaks). The
goal underlying the health payer perspective is to allocate the DOH budget
to maximize the health. It therefore reimburses a vaccine only if its incre-
mental cost-​effectiveness ratio (the ratio of incremental DOH costs to the
incremental QALYs/​DALYs) falls below the incremental cost-​effectiveness
ratio of the marginal health technology (the technology likely to be displaced
to accommodate the vaccine given the fixed DOH budget, which should
be the currently reimbursed technology with the highest incremental cost-​
effectiveness ratio). This perspective assumes—​or at least does not question—​
the optimality of the size of the DOH budget and does not attempt to inform
the decisions of the DOF and of legislatures. The societal perspective, in con-
trast, considers the full public health and socioeconomic value of vaccines,
and indeed of any policy whether health related or not, being considered for
public funding out of the DOH or DOF budget.
CUA and CBA differ from each other in what each takes as the unit of value.
CUA assumes that every QALY or DALY has equal value, which is sometimes
called the “QALY is a QALY is a QALY” assumption. CBA assumes that every
dollar has equal value, which we might call the “dollar is a dollar is a dollar”
assumption. These views empirically diverge when the same size health gains
 1.4.5 Do governments invest enough in vaccination? 61

can have differential socioeconomic implications. One critical example of

such divergence comes from lifecycle variations in economic activity: since
prime-​aged adults are more likely to be working and less likely to be de-
pendent on social pensions and assistance than the elderly, raising the health
of prime-​aged adults by 1 QALY may have more instrumental economic value
for society as a whole than raising the health of the elderly by 1 QALY. Another
example of such divergence might be between diseases with and without out-
break potential. As discussed earlier, addressing the COVID-​19 pandemic in
the absence of vaccines requires shutting down economic activity, which has
negative macroeconomic effects. This implies that COVID-​19 vaccines do not
just raise population health but also facilitate a return to normal pre-​pandemic
levels of macroeconomic activity. Other vaccine-​preventable infectious dis-
eases (e.g., pneumococcal disease) may be less macroeconomically costly to
control, so vaccines that address such diseases may have less macroeconomic
value to them. CBA is better able than CUA to reflect heterogeneity in socio-
economic effects—​whether driven by lifecycle or macroeconomic factors—​of
improved health.
CBA’s equal value per dollar assumption, however, makes it give inappro-
priate priority to the health needs of the wealthy, who have greater ability to
pay. Such problematic priority leads many to prefer the health payer perspec-
tive and CUA, which for all its shortcomings values the QALY gains of the
poor as much as those of the rich. To eliminate such problematic priority,
distribution-​sensitive versions of CBA should be developed. The most prom-
ising such version involves using social welfare functions, which effectively
scale individuals’ willingness to pay by their marginal utility of income (to
neutralize differences in ability to pay) and by distributional weights which
are larger for the worse off (Adler, 2019). Combining CBA and social welfare
functions allows economic evaluation of vaccines to robustly incorporate eq-
uity considerations.
Many DOHs worldwide adopt a health payer perspective CUA as the base-
line specification for vaccine evaluation. Such adoption risks underinvest-
ment in vaccines. The health payer’s perspective, by focusing only on health
gains and health payer budget impacts, ignores important socioeconomic in-
strumental value elements of vaccines. Vaccines may have disproportionate
socioeconomic value relative to other health technologies because, for ex-
ample, they prevent outbreaks (like the COVID-​19 pandemic) that are ac-
companied by macroeconomic burdens. Because of such omission, there is
a risk of suboptimally low vaccine spending out of the health payer’s budget.
Another suboptimal aspect of the health payer’s perspective is its unques-
tioning acceptance of the size of the health payer’s budget. The fact that this
62 Vaccination as investment in human capital

perspective ignores socioeconomic values yet assumes the health payer’s

budget to be optimal implies that this perspective faces the risk of spending
too little on health in general (and therefore vaccines in particular). From a
societal perspective, public spending on health technologies should reflect
both its intrinsic and instrumental value. The fact that the health payer per-
spective ignores socioeconomic value makes it likely that from a societal per-
spective, public spending on health is suboptimally low. The solution to such a
problem is to raise the health payer’s budget in recognition of the instrumental
value of its expenditures, but the health payer perspective assumes that the ex-
isting budget is already optimal, thereby prohibiting the expansion of health
expenditures merited by their instrumental value. This suggests that simply
taking the health payer’s budget as given and optimal risks underinvestment
in health and vaccination relative to their socially optimal levels.
To facilitate setting the health payer’s budget at a socially optimal level, and
thereby facilitate optimal vaccination investment out of that budget, a shift
from a health payer perspective to a societal one is critical. More generally,
optimal spending and decision-​making by DOHs, DOFs, and legislatures will
require shifting from health payer perspective CUAs toward societal perspec-
tive CBAs. Such a shift of perspective would fully incorporate the full social
benefits and costs of vaccination into economic evaluation, helping identify
optimal levels of vaccine spending. A shift from CUA to CBA would allow
more accurate estimation of the socioeconomic value of vaccination, uncon-
strained by simplifying assumptions such as equal values per QALY.

1.4.6 Research and development, or investment

in future vaccines

We have so far discussed optimal investment in existing vaccines, but also im-
portant is optimal investment in R&D into future vaccines. R&D financing
comes from three main sources: governments which often fund basic research,
for-​profit pharmaceutical companies, and non-​government nonprofits like
the Bill & Melinda Gates Foundation.
For-​profits will invest in vaccine R&D only if they expect sufficient risk-​
adjusted return on such investment. The probable underspending on vaccina-
tion by national governments that we previously discussed therefore creates
risks of sending adverse market signals to for-​profits and risks of suboptimally
low investment in vaccine R&D, which could harm future generations.
Greater recognition of the full public health and socioeconomic value of vac-
cination, and superior quantification of such value using societal perspective
 1.4.6 Research and development, or investment in future vaccines 63

CBA and social welfare functions may therefore facilitate not just optimal in-
vestment by governments in existing vaccines but also optimal investment by
for-​profits in future vaccines. Such broader vaccine evaluation can also stim-
ulate R&D funding by governments and nongovernmental nonprofits by pro-
viding more comprehensive estimates of the social value of future vaccines.
There is a tension, however, between trying to incentivize innovation
on the one hand and trying to achieve widespread access and equity on
the other. Pharmaceutical company R&D is responsive to expected profits,
which in turn is facilitated by vaccine prices reflecting healthy markups
above marginal production costs. But such markups, in turn, can make
vaccines unaffordable, reduce access, and cause deadweight loss and ineq-
uity. There is, therefore, a potential trade-​off between present and future
generations, the former benefitting from low markups and the latter from
the innovation enabled by higher markups. Of course, there may also be
markups that are so high that they yield expected profits in excess of what
is necessary to incentivize R&D. Any profits extracted from such markups
are called rents. They are purely redistributive, causing deadweight loss, in-
equity, and lack of access without facilitating any innovation. If equity is a
central policy goal, then all else being equal, such rents should be minim-
ized. Reconciling innovation with access and equity, and minimizing rent,
all at the same time, is a challenging policy balancing act. This has provoked
research on possible responses like value-​based pricing (Danzon, Towse, &
Mestre‐Ferrandiz, 2015), fair pricing (Moon, Mariat, Kamae, & Pedersen,
2020), innovative financing mechanisms like advanced market commit-
ments (Kremer et al., 2020), and more radical proposals like eliminating
patents (Boldrin & Levine, 2013).
Ultimately, the pathway to achieving investments in vaccines that get us
closer to embracing their full value will require both private and public sector
investors who are motivated by accounts that trace that value across the mul-
tiple ways that vaccines benefit society. The work of vaccine economics and
the methods that account for the full value of vaccines are the subjects of the
remainder of this book.

References
Adler, M. (2019). Measuring social welfare: An introduction. New York, NY: Oxford University
Press.
Anekwe, T. D., Newell, M.-​L., Tanser, F., Pillay, D., & Bärnighausen, T. (2015). The causal effect
of childhood measles vaccination on educational attainment: A mother fixed-​effects study in
rural South Africa. Vaccine, 33(38), 5020–​5026. doi:10.1016/​j.vaccine.2015.04.072
64 Vaccination as investment in human capital

Bärnighausen, T., Bloom, D. E., Cafiero, E. T., & O’Brien, J. C. (2012). Economic evaluation
of vaccination: Capturing the full benefits, with an application to human papillomavirus.
Clinical Microbiology and Infection, 18(s5), 70–​76. doi.org/​10.1111/​j.1469-​0691.2012.03977.x
Becker, G. (2007). Health as human capital: Synthesis and extensions. Oxford Economic Papers,
59(3), 379–​410. doi.org/​10.1093/​oep/​gpm020
Betsch, C., Ulshöfer, C., Renkewitz, F., & Betsch, T. (2011). The influence of narrative v. statis-
tical information on perceiving vaccination risks. Medical Decision Making, 31(5), 742–​753.
doi:10.1177/​0272989x11400419
Bloom, D. E., & Canning, D. (2000). Policy forum: public health. The health and wealth of nations.
Science, 287(5456), 1207, 1209. doi:10.1126/​science.287.5456.1207
Bloom, D. E., Canning, D., & Shenoy, E. S. (2011). The effect of vaccination on children’s phys-
ical and cognitive development in the Philippines. Applied Economics, 44(21), 2777–​2783.
doi:10.1080/​00036846.2011.566203
Bloom, D. E., Khoury, A., Algur, E., & Sevilla, J. P. (2020). Valuing productive non-​market ac-
tivities of older adults in Europe and the US. De Economist, 168(2), 153–​181. doi:10.1007/​
s10645-​020-​09362-​1
Blumenthal-​Barby, J. S., & Krieger, H. (2015). Cognitive biases and heuristics in medical deci-
sion making: A critical review using a systematic search strategy. Medical Decision Making,
35(4), 539–​557. doi:10.1177/​0272989x14547740
Boldrin, M., & Levine, D. K. (2013). The case against patents. Journal of Economic Perspectives,
27(1), 3–​22. doi:10.1257/​jep.27.1.3
Canning, D., Razzaque, A., Driessen, J., Walker, D. G., Streatfield, P. K., & Yunus, M. (2011).
The effect of maternal tetanus immunization on children’s schooling attainment in Matlab,
Bangladesh: Follow-​up of a randomized trial. Social Science & Medicine, 72(9), 1429–​1436.
doi:10.1016/​j.socscimed.2011.02.043
Cully, M. (2019, June 17). Antibiotics alter the gut microbiome and host health. Nature Research.
https://​www.nat​ure.com/​artic​les/​d42​859-​019-​00019-​x
daCosta DiBonaventura, M., & Chapman, G. (2008). Do decision biases predict bad decisions?
Omission bias, naturalness bias, and influenza vaccination. Medical Decision Making, 28(4),
532–​539. doi:10.1177/​0272989X07312723
Danzon, P. M., Towse, A., & Mestre‐Ferrandiz, J. (2015). Value‐based differential pricing: Efficient
prices for drugs in a global context. Health Economics, 24(3), 294–​301. doi:10.1002/​hec.3021
Dubé, È., Ward, J. K., Verger, P., & MacDonald, N. E. (2021). Vaccine hesitancy, acceptance,
and anti-​vaccination: Trends and future prospects for public health. Annual Review of
Public Health, 42(1), 175–​191. doi:10.1146/​annurev-​publhealth-​090419-​102240
Gessner, B. D., Kaslow, D., Louis, J., Neuzil, K., O’Brien, K. L., Picot, V., . . . Nelson, C. B.
(2017). Estimating the full public health value of vaccination. Vaccine, 35(46), 6255–​6263.
doi:10.1016/​j.vaccine.2017.09.048
Han, P. K. J., Reeve, B. B., Moser, R. P., & Klein, W. M. P. (2009). Aversion to ambiguity regarding
medical tests and treatments: Measurement, prevalence, and relationship to sociodemographic
factors. Journal of Health Communication, 14(6), 556–​572. doi:10.1080/​10810730903089630
Hornsey, M. J., & Fielding, K. S. (2017). Attitude roots and Jiu Jitsu persuasion: Understanding
and overcoming the motivated rejection of science. American Psychologist, 72(5), 459–​473.
doi:10.1037/​a0040437
Hornsey, M. J., Harris, E. A., & Fielding, K. S. (2018). The psychological roots of anti-​vaccination
attitudes: A 24-​ nation investigation. Health Psychology, 37(4), 307–​ 315. doi:10.1037/​
hea0000586
Kremer, M., Levin, J., & Snyder, C. M. (2020). Advance market commitments: Insights from theory
and experience. AEA Papers and Proceedings, 110, 269–​273. doi.org/​10.1257/​pandp.20201017
 1.4.6 Research and development, or investment in future vaccines 65

Masia, N. A., Smerling, J., Kapfidze, T., Manning, R., & Showalter, M. (2018). Vaccination and GDP
growth rates: Exploring the links in a conditional convergence framework. World Development,
103(C), 88–​99. doi:10.1016/​j.worlddev.2017.1
Megiddo, I., Klein, E., & Laxminarayan, R. (2018). Potential impact of introducing the pneu-
mococcal conjugate vaccine into national immunisation programmes: An economic-​
epidemiological analysis using data from India. BMJ Global Health, 3(3), e000636.
doi:10.1136/​bmjgh-​2017-​000636
Mina, M. J., Kula, T., Leng, Y., Li, M., de Vries, R. D., Knip, M., . . . Elledge, S. J. (2019). Measles
virus infection diminishes preexisting antibodies that offer protection from other patho-
gens. Science, 366(6465), 599–​606. doi:10.1126/​science.aay6485
Mina, M. J., Metcalf, C. J., de Swart, R. L., Osterhaus, A. D., & Grenfell, B. T. (2015). Long-​term
measles-​induced immunomodulation increases overall childhood infectious disease mor-
tality. Science, 348(6235), 694–​699. doi:10.1126/​science.aaa3662
Moon, S., Mariat, S., Kamae, I., & Pedersen, H. B. (2020). Defining the concept of fair pricing
for medicines. BMJ, 368, l4726. doi:10.1136/​bmj.l4726
Motta, M., Callaghan, T., & Sylvester, S. (2018). Knowing less but presuming more:
Dunning–​Kruger effects and the endorsement of anti-​vaccine policy attitudes. Social Science
& Medicine, 211, 274–​281. doi:10.1016/​j.socscimed.2018.06.032
Nandi, A., Deolalikar, A. B., Bloom, D. E., & Laxminarayan, R. (2019). Haemophilus influenzae
type b vaccination and anthropometric, cognitive, and schooling outcomes among Indian
children. Annals of the New York Academy of Sciences, 1449(1), 70–​ 82. doi:10.1111/​
nyas.14127
Nandi, A., Shet, A., Behrman, J. R., Black, M. M., Bloom, D. E., & Laxminarayan, R. (2019).
Anthropometric, cognitive, and schooling benefits of measles vaccination: Longitudinal
cohort analysis in Ethiopia, India, and Vietnam. Vaccine, 37(31), 4336–​4343. doi:10.1016/​
j.vaccine.2019.06.025
O’Donoghue, T., & Rabin, M. (2015). Present bias: Lessons learned and to be learned. American
Economic Review, 105(5), 273–​279. doi:10.1257/​aer.p20151085
Oskorouchi, H. R., Sousa-​Poza, A., & Bloom, D. E. (2020). The long-​term cognitive and schooling
effects of childhood vaccinations in China. National Bureau of Economic Research, Inc.
https://​ideas.repec.org/​p/​nbr/​nbe​rwo/​27217.html
Ritov, I., & Baron, J. (1990). Reluctance to vaccinate: Omission bias and ambiguity. Journal of
Behavioral Decision Making, 3(4), 263–​277. doi.org/​10.1002/​bdm.3960030404
Riumallo-​Herl, C., Chang, A. Y., Clark, S., Constenla, D., Clark, A., Brenzel, L., & Verguet, S.
(2018). Poverty reduction and equity benefits of introducing or scaling up measles, rota-
virus and pneumococcal vaccines in low-​income and middle-​income countries: a modelling
study. BMJ Global Health, 3(2), e000613. doi:10.1136/​bmjgh-​2017-​000613
Rossen, I., Hurlstone, M. J., Dunlop, P. D., & Lawrence, C. (2019). Accepters, fence sitters, or
rejecters: Moral profiles of vaccination attitudes. Social Science & Medicine, 224, 23–​27.
doi:10.1016/​j.socscimed.2019.01.038
Sassi, F. et al. (2015) Introduction. In McDaid, David, Franco Sassi, and Sherry Merkur.
Promoting health, preventing disease: the economic case. Open University Press, Maidenhead,
UK. ISBN 9780335262267.
Saadatian-​Elahi, M., Aaby, P., Shann, F., Netea, M. G., Levy, O., Louis, J., . . . Warren, W. (2016).
Heterologous vaccine effects. Vaccine, 34(34), 3923–​3930. doi:10.1016/​j.vaccine.2016.06.020
Sevilla, J. P., Bloom, D. E., Cadarette, D., Jit, M., & Lipsitch, M. (2018). Toward economic evalu-
ation of the value of vaccines and other health technologies in addressing AMR. Proceedings
of the National Academy of Sciences of the United States of America, 115(51), 12911–​12919.
doi:10.1073/​pnas.1717161115
66 Vaccination as investment in human capital

Sevilla, J. P., Stawasz, A., Burnes, D., Agarwal, A., Hacibedel, B., Helvacioglu, K., . . . Bloom, D. E.
(2020). Indirect costs of adult pneumococcal disease and the productivity-​based rate of re-
turn to the 13-​valent pneumococcal conjugate vaccine for adults in Turkey. Human Vaccines
& Immunotherapeutics, 16(8), 1923–​1936. doi:10.1080/​21645515.2019.1708668
Sevilla, J. P., Stawasz, A., Burnes, D., Poulsen, P. B., Sato, R., & Bloom, D. E. (2019). Indirect costs
of adult pneumococcal disease and productivity-​based rate of return to PCV13 vaccination for
older adults and elderly diabetics in Denmark. Journal of the Economics of Ageing, 14, 100203.
doi.org/​10.1016/​j.jeoa.2019.100203
Sharot, T. (2011). The optimism bias. Current Biology, 21(23), R941–​R945. doi:10.1016/​
j.cub.2011.10.030
Stone, D., & Wood, D. (2018). Cognitive dissonance, motivated reasoning, and confirma-
tion bias: Applications in industrial organization. In V. J. Tremblay, E. Shroeder, & C.
Horton Tremblay (Eds.), Handbook of behavioral industrial organization (pp. 114–​137).
Northampton, MA: Edward Elgar Publishing.
Usher, N. T., Chang, S., Howard, R. S., Martinez, A., Harrison, L. H., Santosham, M., &
Aronson, N. E. (2019). Association of BCG vaccination in childhood with subsequent cancer
diagnoses: A 60-​year follow-​up of a clinical trial. JAMA Network Open, 2(9), e1912014.
doi:10.1001/​jamanetworkopen.2019.12014
Verguet, S., Olson, Z. D., Babigumira, J. B., Desalegn, D., Johansson, K. A., Kruk, M.
E., . . . Jamison, D. T. (2015). Health gains and financial risk protection afforded by public
financing of selected interventions in Ethiopia: An extended cost-​effectiveness analysis.
Lancet Global Health, 3(5), e288–​e296. doi:10.1016/​s2214-​109x(14)70346-​8
Wilder-​ Smith, A., Longini, I., Zuber, P. L., Bärnighausen, T., Edmunds, W. J., Dean,
N., . . . Gessner, B. D. (2017). The public health value of vaccines beyond efficacy: Methods,
measures and outcomes. BMC Medicine, 15(1), 138. doi:10.1186/​s12916-​017-​0911-​8
World Health Organization. (2014). Report of the Sage Working Group on Vaccine Hesitancy.
Retrieved from Geneva: https://​www.asset-​scien​cein​soci​ety.eu/​sites/​defa​ult/​files/​sage_​
working_​group_​revis​ed_​r​epor​t_​va​ccin​e_​he​sita​ncy.pdf
World Health Organization. (2019a). Ten threats to global health. https://​www.who.int/​news-​
room/​spotli​ght/​ten-​thre​ats-​to-​glo​bal-​hea​lth-​in-​2019
World Health Organization. (2019b). WHO Director General statement on the role of social
media platforms in health information [Press release]. https://​www.who.int/​news-​room/​
spotli​ght/​ten-​thre​ats-​to-​glo​bal-​hea​lth-​in-​2019
Xue, Q. C., & Ouellette, L. L. (2020). Innovation policy and the market for vaccines. Journal of
Law and the Biosciences, 7(1), lsaa026. doi:10.1093/​jlb/​lsaa026
1.5
Economics of vaccine delivery
George Pariyo and Onaopemipo Abiodun

1.5.1 Introduction

Since the global recognition of immunization as one of the most effective and
cost-​effective ways to promote public health, countries have adopted different
approaches to deliver vaccines. High-​income countries often rely on private
providers to deliver vaccines with financing through public and private health
insurance (Cherian & Mantel, 2020; World Health Organization, 2018). On the
other hand, most low-​and middle-​income countries (LMICs) deliver vaccin-
ations mainly through government-​funded national immunization programs
(NIPs) (Cherian & Mantel, 2020). NIPs tend to deliver vaccines at dedicated
clinics or through outreach and mobile services (Cherian & Mantel, 2020).
However, in many LMICs which have a mixed public–​private health delivery
system, and especially in fragile states where government institutions are too
weak to meet basic needs, private not-​for-​profit (PNFP) facilities are often
contracted by governments to provide vaccinations (Levin & Kaddar, 2011).
Private providers are also playing an increasingly important role in non-​fragile
LMICs. The level at which their services are integrated with those of public
providers and regulated by the government varies widely. Effective coordina-
tion between the public and private sectors is essential, but there is no single
way to achieve this since different contexts have different needs. For LMICs to
enjoy the full benefits of immunization, they must be able to commit to long-​
term financing requirements (Results for Development, 2017). The aim of this
chapter is to assess the advantages and disadvantages of various immunization
delivery systems—​public, private not-​for-​profit, private for-​profit (PFP), and
mixed delivery—​across five domains: efficiency, which refers to immunization
delivery at the lowest possible cost per dose; financial sustainability, which re-
fers to the ability to finance immunization delivery in the long term; and eq-
uity, which means that every eligible individual can be immunized with all

George Pariyo and Onaopemipo Abiodun, Economics of vaccine delivery In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0006
68 Economics of vaccine delivery

appropriate vaccines (World Health Organization, 2013). We will also look at

delivery strategies from the perspective of quality, which is delivering a service
according to standards, as well as coverage, which refers to the proportion of
those who would benefit from a service who receive it. While we are aware that
it is impossible to strike a perfect balance among these five domains, the goal of
this chapter is to support the critical assessment of immunization delivery sys-
tems. This is necessary to ensure the systems adequately serve those who rely
on them. First, we consider these domains and their implications.

1.5.2 Efficiency

Efficiency refers to the ratio of input costs to desired outputs. In the context of
immunization programs, we may measure efficiency as cost per dose of vaccine
administered or cost per fully immunized child. In this chapter, we refer to effi-
ciency in terms of cost per dose of vaccine administered, varying widely within
and between countries with mixed delivery systems. A multi-​country study con-
ducted in Benin, Ghana, Honduras, Moldova, Uganda, and Zambia found that
the average unit cost of routine immunization provided through public and non-
governmental organization facilities ranged from $2 in Benin to $18 in Moldova
(Brenzel, Young, & Walker, 2015). While these differences may represent potential
opportunities for efficiency gains, they may also be due to operational differences
that cannot be readily altered, such as wages, site location, and health systems
structure (Menzies, Suharlim, Resch, & Brenzel, 2020). In each country, high
service volume was strongly associated with lower average cost per dose. On the
other hand, higher per capita gross domestic product and outreach services were
correlated with higher average cost per dose, hinting at the effect of country-​level
differences on efficiency measures as well as likely trade-​offs between efficiency
and efforts to extend access to hard-​to-​reach population. Nongovernmental or-
ganization facilities were also associated with a higher average cost per dose than
public facilities (Menzies et al., 2017). This hints at a trade-​off between efficiency
and equity in public health policy. For instance, a country may choose to prioritize
vaccinating all the citizens regardless of the fact that some harder-​to-​reach com-
munities may significantly increase the unit costs of the program.

1.5.3 Financial sustainability

Sustainability has a wide interpretation in global health and may refer to

financial, logistical, personnel, or political aspects of a program’s ability to
 1.5.4 Equity 69

continue long term. In the context of this chapter, we are mainly referring to
financial sustainability, the ability of a program to obtain ongoing funding
from secure or predictable sources (World Health Organization, 2013). The
financial sustainability of immunization delivery is under threat in many
settings with a mixed mode of delivery. Oftentimes, this stems from a low
prioritization of immunization by governments. In some LMICs, some pol-
icymakers may not allocate sufficient national budgetary funds to immuni-
zation with the expectation that international donors will step in to bridge
the gap (Songane, 2018). In the US, private practice pediatricians and family
physicians who function as small independent businesses are the main pro-
viders of early childhood vaccines (Berman, 2008). However, these phys-
icians are at a disadvantage when negotiating vaccine prices as studies have
found that vaccine prices are related to the amount ordered. As such, larger
private practices usually receive lower prices than smaller practices. These
small practices, instead, rely on insurance reimbursements, which they ne-
gotiate with health plans to cover the vaccine costs that they incur (O’Leary
et al., 2014). Despite evidence showing the effectiveness of vaccines, insur-
ance companies use their larger negotiating power, owing to their larger
size, to negotiate reimbursements below costs (Berman, 2008; O’Leary et al.,
2014). Due to the risk of uncompensated costs, some physicians have delayed
offering vaccines or even outsourced immunizations (Beaulieu-​Volk, 2014).
In Ghana, the National Health Insurance Scheme (NHIS) only reimburses
curative care (Results for Development, 2017). Indonesia operates a decen-
tralized health system, where local governments are responsible for covering
immunization service delivery costs using funds disbursed from the national
level (Results for Development, 2017). Wide variation in managerial capa-
bility and commitment to immunization at the subnational level has resulted
in varying immunization coverage throughout the country (Maharani &
Tampubolon, 2014; Results for Development, 2017). To maintain the finan-
cial sustainability of mixed immunization delivery systems, government in-
tervention is necessary. This could take the form of legislation mandating
that immunization be funded at a rate that encourages providers to deliver
this essential service (Berman, 2008).

1.5.4 Equity

Equity implies the absence of avoidable and unnecessary disparities in access

to or utilization of a service based on individual or community characteristics
such as geography, race and ethnicity, religious affiliation, or socioeconomic
70 Economics of vaccine delivery

status. In assessing the equitability of immunization in mixed delivery systems,

it is important to acknowledge existing societal inequities, which can be dif-
ficult to surmount. For instance, all adults aged 65 years and older in the US
have access to the 13-​valent pneumococcal conjugate vaccine (PCV13), which
is universally recommended for them by the Centers for Disease Control and
Prevention with no out-​of-​pocket costs because of full coverage by Medicare
Part B. Yet, racial and socioeconomic differences in PCV13 uptake among
adults in this age group remain (McLaughlin et al., 2019). These disparities are
likely due to other healthcare barriers that minorities and low-​income indi-
viduals often face, such as poor access to reliable transportation and primary
care (O’Malley & Forrest, 2006; Silver, Blustein, & Weitzman, 2012). On the
other hand, contractual relationships between the public and private sector
with explicit equity goals can reduce disparities in immunization. This was the
case in Cambodia, where results of a large-​scale quasi-​experiment showed that
children in the poorest 50% of households in the districts served by contrac-
tors were more likely to be fully immunized than their counterparts in districts
using the typical government model for health service delivery (Schwartz &
Bhushan, 2004).

1.5.5 Quality

Quality broadly refers to performance according to standards. Quality in

health has several different dimensions including safety, effectiveness, client
centeredness, timeliness, efficiency, and equity (Mosadeghrad, 2012). All
these aspects of quality are important with respect to immunizations. People
should be vaccinated with an efficacious vaccine in a safe procedure, with little
to no adverse effects, at a place and time that is convenient to them, and by
friendly service providers who are knowledgeable and able to communicate
effectively to respond to their questions and reassure them that vaccination
is not only highly effective in preventing serious diseases but also very safe.
This requires not only the right vaccines on the schedule but also effective
cold chain and logistics facilities as well as knowledgeable and motivated per-
sonnel. Regular assessments of effective vaccine management, delivery, and
cold chain systems conducted by countries with support of the World Health
Organization and UNICEF have often found some deficiencies especially
in LMICs (Gavi, 2018). Surveillance, monitoring, and reporting of adverse
events following immunization is an important component of an immuniza-
tion program.
 1.5.7 Options for increased vaccine delivery 71

1.5.6 Coverage

Coverage refers to the percentage of a population who would benefit from

a service who can obtain the service (World Health Organization, 2010).
Tanahashi (1978) described different aspects of coverage including availa-
bility, accessibility, acceptability, contact, and effective coverage. Effective cov-
erage refers to those who eventually get in contact and receive effective care. In
the context of immunization, this implies the percentage of those who would
benefit who receive a safe and effective vaccine at the correct time and for the
correct indication. Coverage rates are regularly reported through routine ad-
ministrative systems and occasionally validated through immunization cov-
erage surveys. Cross-​country comparisons of immunization coverage rely on
the annually reported World Health Organization and UNICEF estimates of
national immunization coverage reports (Burton et al., 2009).

1.5.7 Options for increased vaccine delivery

As countries work toward sustainably financing their immunization programs,

which delivery system will best position them to achieve this goal? The answer,
as the available evidence shows, may be context specific. Some countries, like
Malaysia, have achieved impressive levels of immunization coverage with a
largely government-​run system, while others, like Ghana and Indonesia, have
adopted a mixed financing system (Results for Development, 2017).

1.5.7.1 Public delivery

Some countries have been able, using general government revenue, to provide
immunizations for free at government-​run health facilities (Box 1.5.1). As of
2015, over 90% of Malaysian states achieved three-​dose diphtheria, tetanus,
and pertussis (DTP3) vaccine coverage above 90%, while 60% achieved over
99% of coverage (Coe, Gergen, Mallow, Moi, & Phily, 2017). However, budget
constraints are a major barrier to the adoption of new vaccines that address
some of the vaccine-​preventable diseases, such as dengue (Coe et al., 2017).
Public delivery is dependent on assured funding secured through the na-
tional budget process. Having established budget lines to fund vaccine pur-
chase and delivery of immunizations through routine health service delivery
is a great way to institutionalize sustainability of funding for immunization.
72 Economics of vaccine delivery

Box 1.5.1 Case study—​Malaysia

Malaysia has achieved impressive levels of immunization coverage with a

largely government-​run system. Using general government revenue, Malaysia
provides immunizations for free at government-​run health facilities (Coe et al.,
2017). As of 2015, over 90% of its states achieved DTP3 vaccine coverage above
90%, while 60% achieved over 99% of coverage (Coe et al., 2017). Yet these high
percentages may mask concerning inequities. For instance, Malay ethnic children,
who belong to the majority ethnic group, were found to have a significantly lower
likelihood of receiving all recommended primary vaccine doses by the age of
12 months, as compared to minority Indian and other Bumiputera children (Lim
et al., 2017).
Additionally, although Malaysia has been able to achieve high immunization
coverage using general government revenue, budget constraints are a major bar-
rier to the adoption of new vaccines for major disease burdens, such as dengue
(Coe et al., 2017). The inability of public facilities to provide new vaccines has
driven many Malaysians to seek them from private facilities, where they are re-
quired to pay out of pocket. Variations in people’s ability to pay raises equity
concerns (Coe et al., 2017). Private facilities are playing an increasingly im-
portant role in the health system, as they provide an alternative to the public
system, which is struggling to keep up with a growing population (World Health
Organization: Regional Office for the Western Pacific, 2012). However, the govern-
ment has not been able to integrate and regulate the fast-​growing private health
sector. As a case in point, private providers are not mandated to follow the gov-
ernment immunization schedule (Ahmad, Jahis, Kuay, Jamaluddin, & Aris, 2017).
This puts many children at risk of missing necessary immunizations. A study
found that 10.7% of surveyed children aged 12–​23 months did not complete their
primary immunization due to lack of vaccine stock at private facilities, while an-
other 6.2% did not complete their primary immunization because private health-
care centers advised that they were not yet due to be immunized, as reported by
their mothers (Ahmad et al., 2017).
Malaysia’s NHIS is being designed to facilitate integration between the public
and private sectors, and improve access to vaccines (Coe et al., 2017; World Health
Organization: Regional Office for the Western Pacific, 2012). It is also important
the government enforces a standard immunization schedule for all children at all
facilities, public or private, so that timely vaccination is available at all points of
service delivery (Ahmad et al., 2017).
 1.5.7 Options for increased vaccine delivery 73

This means it can be a guaranteed source of funding if there is political com-

mitment and national consensus that prioritizes immunization. Global
agencies that help LMICs through funding for new vaccines and cold chain
and logistics costs cover a substantial portion of the cost of immunization
(Songane, 2018). Having immunization funding institutionalized in national
and district health budgets helps prepare for the time external funding agen-
cies phase out, ensuring better chances for sustainability of immunization
financing. On the downside, it increases government expenditure and may
suffer when the political priority of immunization and other health programs
changes with successive governments. In many LMICs, public facilities may
be the only available source of immunization services, particularly in rural
areas, where other providers may have less incentive to operate. Ministries
of health, through their NIPs, can exercise stronger supervision of standards
and have the power to take corrective action. This helps to ensure quality, for
instance, promptly investigating any reports of adverse events following im-
munization, and taking corrective action. In areas where the overall quality
of government-​run health services is low, immunization services may also be
avoided due to low credibility and a trust deficit between the communities
and public facilities. Public sector salaries are often lower than equivalently
qualified workers in the private sector. However, public facilities are often
perceived to be wasteful compared to their counterparts in PNFP or PFP fa-
cilities. The main sources of inefficiency may be due to high fixed costs for
personnel and facility maintenance and utilities, as well as high slack capacity,
especially in a context of low utilization.

1.5.7.2 Private not-​for-​profit providers

PNFP facilities are common in many LMICs and may contribute up to 40%
of health service outputs such as vaccinations (Levin & Kaddar, 2011). They
are often located in rural or other hard-​to-​reach areas of a country and thus
serve as a source of much-​needed services where public facilities do not exist
and where there may be no incentive for PFP clinics to operate. This promotes
equity in immunization delivery. The fact that they often depend on unpre-
dictable donations, either from local sources or from foreign donors, makes
their sustainability rather uncertain. Nevertheless, these facilities often ben-
efit from government subsidies when conducive policies exist. Given that
many of these PNFPs may have a faith-​based orientation, at least in their ori-
gins and often in their management, they may be perceived to be community
74 Economics of vaccine delivery

oriented and motivated by a strong sense of service and community solidarity.

However, some segments of the population may hesitate to access them if they
do not feel welcome due to being of a different faith. There have also been
concerns about PNFP facilities, for budgetary reasons, employing staff who
have fewer professional qualifications than their public facility counterparts.
General technical competence was found to be lower in private facilities than
public in one review of the literature (Berendes, Heywood, Oliver, & Garner,
2011). Health workers’ knowledge of immunization schedules, waste, and
vaccine management practices were found lacking in private facilities in a
study of Cambodia (Soeung, Grundy, Morn, & Samnang, 2008).

1.5.7.3 Private clinics and practitioners

Private sector vaccine delivery offered by PFP clinics may exacerbate health
inequities. A study of private sector Haemophilus influenzae type b vac-
cine coverage in India found that coverage was mainly limited to richer
and more urbanized states (Sharma, Kaplan, Chokshi, Hasan Farooqui, &
Zodpey, 2015).
In the US, adult immunization is mainly performed by private providers
(Hinman, Orenstein, & Rodewald, 2004). Private providers recover costs
through public and private insurance reimbursements for vaccine purchase
and administration (O’Leary et al., 2014). However, a national survey car-
ried out among private pediatricians and family physicians found that many
of them are dissatisfied with insurance payments for immunization (O’Leary
et al., 2014). Vaccine costs, which made up a minor part of the overheads of
a private pediatric practice in the 1980s, are now one of the top overhead ex-
penses largely due to new vaccines. This increases the risk of uncompensated
costs to private providers in the US (American Academy of Pediatrics, 2006).
Many physicians reported delaying offering vaccines to patients if insurance
coverage is not certain (Hurley et al., 2017; O’Leary et al., 2014). This is a con-
cerning trend, since physician recommendation has been found to play a cru-
cial role in a patient’s vaccination (Hurley et al., 2017).
Although PFP providers deliver a limited proportion of immunization
services in LMICs, their activities are garnering more attention. In addition
to providing traditional World Health Organization Expanded Program on
Immunization vaccines to those who can afford to pay, private providers also
introduce new and underutilized vaccines (Levin & Kaddar, 2011). However,
they are more susceptible to market pressures than public providers. As such,
 1.5.7 Options for increased vaccine delivery 75

they are often targeted by pharmaceutical marketing campaigns to introduce

new and costly vaccines, and sometimes deviate from national immuniza-
tion schedules (Levin & Kaddar, 2011; World Health Organization, 2018).
Deviating from the national immunization schedule can have deleterious ef-
fects, such as increasing the risk of vaccine-​preventable diseases or confusing
the population, thereby reducing trust in the health system (World Health
Organization, 2018). Nevertheless, individuals may choose private providers
over public providers for immunization for a variety of reasons, including
convenience from being closer or operating at better hours; they are the
only providers of a particular vaccine; or they are perceived to provide better
quality services than public providers (Amarasinghe, Davison, & Diorditsa,
2018; Levin & Kaddar, 2011). Yet, studies from several countries have revealed
a number of quality concerns that can limit the efficiency of PFP immuniza-
tion services, such as substandard vaccine storage practices and lack of know-
ledge on waste and vaccine management practices (Levin & Kaddar, 2011;
Soeung et al., 2008). While the financial sustainability of PFP immunization
may not be of concern, since providers mainly cater to those who can afford
it, this attribute raises equity concerns. Studies conducted in LMICs confirm
these concerns as they show that individuals who seek immunization services
from these providers tend to be of a higher socioeconomic status or have a
higher income (Levin & Kaddar, 2011).

1.5.7.4 Mixed delivery

Each financing and delivery system has its advantages and disadvantages.
Some countries have opted for a mixed approach, which includes public and
private funding and delivery of immunization.
For instance, Ghana and Indonesia provide free immunizations to their
populations through private and public providers using mostly govern-
ment budgetary funds with some contributions from their NHISs (Results
for Development, 2017). While still adopting a mixed-​financing system,
Thailand provides all health services, including immunization, through a so-
cial health insurance scheme that is funded through general taxation (Results
for Development, 2017).
Since 2016, he Malaysian government has been establishing an NHIS to fa-
cilitate integration between the public and private sectors and improve access
to vaccines (Coe et al., 2017). However, mixed financing systems pose certain
challenges. While funds from their NHISs have helped to diversify Ghana’s
76 Economics of vaccine delivery

and Indonesia’s immunization financing, there is the risk of curative services

being prioritized over immunization and other preventive care for different
reasons. In Ghana, salaries are taking up an increasing proportion of the
Ministry of Health budget, which has made public facilities more dependent
on NHIS claims for payments for immunization delivery. However, these pay-
ments are only received for curative care (Results for Development, 2017). In
Indonesia, capitation payments made by the NHIS, known as JKN, to primary
care providers for immunization are sometimes diverted to curative care due
to confusion among providers and local governments, which are responsible
for covering service delivery costs in Indonesia’s decentralized health system
(Results for Development, 2017).
As many LMICs move toward mixed financing and health insurance
scheme models for immunization, they can learn from the experiences of
high-​income countries, like the US, with similar financing models. In the
US, immunization is provided through a mix of public and private providers
via government and private financing. In 1993, the US government passed
the Vaccines for Children (VFC) Act, which sets aside funding to immunize
children who are uninsured, on Medicaid, or who are American Indians/​
Alaska Natives free of charge (Hinman et al., 2004; Robinson, Sepe, & Lin,
1993). The VFC Act also provides free vaccines at federally qualified health
centers for insured children without immunization coverage (Robinson et al.,
1993). Through the VFC program, the federal government purchases a size-
able share of pediatric vaccines. However, approximately half of pediatric
vaccines are purchased by private pediatricians and family physicians, who
vaccinate most US children (Hill, Elam-​Evans, Yankey, Singleton, & Kang,
2018; Tayloe, 2009).
While collaboration with the private sector has helped many governments
to improve their NIPs, there have also been drawbacks. For instance, public
financing is not always sufficient to meet the needs of private practices. In
the US, many private providers participating in the VFC program have ex-
pressed dissatisfaction with reimbursements they receive from public insurers
(O’Leary et al., 2014). As of 2016, it was reported that private practices lose ap-
proximately $5–​$15 for each dose of VFC vaccine administered (Diasio, 2016).

1.5.8 Discussion

We have discussed the pros and cons of using public, PNFP, PFP, and mixed
immunization delivery options from the perspectives of the five domains
 1.5.8 Discussion 77

of efficiency, sustainability, equity, quality, and coverage (immunization fi-

nancing is further discussed in Chapter 5). While there is no single best solu-
tion, some lessons emerge and can guide country immunization policymakers
in choosing the approach which works best in their setting. For LMICs to
enjoy the full benefits of immunization, they must be able to commit to long-​
term financing requirements (Results for Development, 2017). From 2000,
Gavi, the Vaccine Alliance financed the provision of vaccines that contrib-
uted to saving the lives of millions of children in over 75 of the world’s poorest
countries (Gavi, 2019). With or without external donor support, countries
need to find new ways to finance their national immunization portfolios in
the context of increasing cost of new vaccines (Saxenian et al., 2015). One
estimate suggested that African countries needed approximately US $17 bil-
lion to finance vaccine purchases and service delivery between 2016 and 2020
(Songane, 2018).
Unsurprisingly, there appears to be no universally successful approach to
immunization financing and no one vaccine delivery approach works best
in different contexts even in the same country. All countries, and especially
LMICs, need to adopt flexible vaccine delivery and financing models that best
suit their various contexts. However, country experiences show a few com-
monalities in terms of factors to consider in the design and adoption of vaccine
delivery systems. These include the need for clearly defined responsibilities in
the system, leveraging both public and private sector mechanisms, strong su-
pervision of both public and private sector providers to ensure performance
according to international and national standards, as well as approaches to
reduce inequities and ensure that immunization is prioritized in the health
system as a public good. Private providers of immunization need access to
public resources such as technical support, skills building and training op-
portunities, and financial subsidies in recognition of their contribution to the
public good.
In order to achieve equity of immunizations, countries need to deliver
vaccines through providers who are more readily accessible to communities
near to where they live regardless of whether these are public, PNFP, or PFP
providers. Achieving efficiency does not have to come at the cost of equity
if all the resources available in the community are leveraged. Involving the
private sector to deliver immunizations in countries where the public facil-
ities offer the bulk of immunizations will help alleviate the burden on these
often-​underfunded services, while leveraging the innovation and additional
infrastructure and personnel resources of the private sector. However, pri-
vate facilities will need additional support through skills-​based training
78 Economics of vaccine delivery

funded by the public sector, as well as access to guidelines and technical sup-
port available in the public sector to ensure adherence to national and inter-
national standards.

1.5.9 Conclusion

An integrated immunization delivery system which leverages the respective

strengths of the public and private sectors, delivering the life-​saving and cost-​
effective vaccines to whomever and wherever they are needed, will contribute
to socioeconomic development and help reduce inequities in health status.
For countries where immunization is traditionally largely delivered through
expensive, inefficient public delivery systems, it is time to try a new approach
which prioritizes immunization regardless of who is delivering it. Finding
ways to leverage available private sector providers in a way that does not im-
pose too much of a financial burden on them will yield better coverage rates
and healthier people, while ensuring better prospects for achieving health
goals sustainably.

References
Ahmad, N. A., Jahis, R., Kuay, L. K., Jamaluddin, R., & Aris, T. (2017). Primary immuniza-
tion among children in Malaysia: Reasons for incomplete vaccination. Journal of Vaccines &
Vaccination, 8(3), 358. doi:10.4172/​2157-​7560.1000358
Amarasinghe, A., Davison, L., & Diorditsa, S. (2018). Engagement of private providers in im-
munization in the Western Pacific region. Vaccine, 36(32, Part B), 4958–​4962. doi:10.1016/​
j.vaccine.2018.01.008
American Academy of Pediatrics. (2006). The business case for pricing vaccines and immuniza-
tion administration. Washington, DC: American Academy of Pediatrics.
Beaulieu-​Volk, D. (2014). Immunizations: How to make this vital service financially
viable. Experts offer best practices related to smart purchasing, inventory control, and
reducing waste. Medical Economics, 91(3), 68–​72. https://​pub​med.ncbi.nlm.nih.gov/​
25211​950/​
Berendes, S., Heywood, P., Oliver, S., & Garner, P. (2011). Quality of private and public ambu-
latory health care in low and middle income countries: Systematic review of comparative
studies. PLoS Medicine, 8(4), e1000433. doi:10.1371/​journal.pmed.1000433
Berman, S. (2008). Is our vaccine system at risk for a future financial “meltdown?” Pediatrics,
122(6), 1372. doi:10.1542/​peds.2008-​2881
Brenzel, L., Young, D., & Walker, D. G. (2015). Costs and financing of routine immuniza-
tion: Approach and selected findings of a multi-​country study (EPIC). Vaccine, 33, A13–​A20.
doi:10.1016/​j.vaccine.2014.12.066
Burton, A., Monasch, R., Lautenbach, B., Gacic-​Dobo, M., Neill, M., Karimov, R., . . . Birmingham,
M. (2009). WHO and UNICEF estimates of national infant immunization coverage: Methods
 1.5.9 Conclusion 79

and processes. Bulletin of the World Health Organization, 87(7), 535–​ 541. doi:10.2471/​
blt.08.053819
Cherian, T., & Mantel, C. (2020). National immunization programmes. Bundesgesundheitsblatt—​
Gesundheitsforschung—​Gesundheitsschutz, 63(1), 16–​24. doi:10.1007/​s00103-​019-​03062-​1
Coe, M., Gergen, J., Mallow, M., Moi, F., & Phily, C. (2017). Sustainable immunization financing
in Asia Pacific. Washington, DC: ThinkWell.
Diasio, C. (2016, February 9). Pediatric vaccination: Who bears the burden? HealthAffairs.
https://​www.health​affa​irs.org/​do/​10.1377/​hblog2​0160​209.053​058/​full/​
Gavi. (2018). Annual program report. https://​www.gavi.org/​sites/​defa​ult/​files/​publi​cati​ons/​
progr​ess-​repo​rts/​Gavi-​Progr​ess-​Rep​ort-​2018.pdf
Gavi. (2019). Annual program report. https://​www.gavi.org/​sites/​defa​ult/​files/​pro​gram​mes-​
imp​act/​our-​imp​act/​apr/​Gavi-​Progr​ess-​Rep​ort-​201​9_​1.pdf
Hill, H. A., Elam-​Evans, L. D., Yankey, D., Singleton, J. A., & Kang, Y. (2018). Coverage among
children aged 19–​35 months—​United States, 2017. Morbidity and Mortality Weekly Report,
67, 1123–​1128. doi:10.15585/​mmwr.mm6740a4
Hinman, A. R., Orenstein, W. A., & Rodewald, L. (2004). Financing immunizations in the
United States. Clinical Infectious Diseases, 38(10), 1440–​1446. doi:10.1086/​420748
Hurley, L. P., Lindley, M. C., Allison, M. A., Crane, L. A., Brtnikova, M., Beaty, B. L., . . . Kempe, A.
(2017). Primary care physicians’ perspective on financial issues and adult immunization in the
Era of the Affordable Care Act. Vaccine, 35(4), 647–​654. doi:10.1016/​j.vaccine.2016.12.007
Levin, A., & Kaddar, M. (2011). Role of the private sector in the provision of immunization
services in low-​and middle-​income countries. Health Policy and Planning, 26(Suppl 1),
i4–​i12. doi:10.1093/​heapol/​czr037
Lim, K. K., Chan, Y. Y., Noor Ani, A., Rohani, J., Siti Norfadhilah, Z. A., & Santhi, M. R.
(2017). Complete immunization coverage and its determinants among children in
Malaysia: Findings from the National Health and Morbidity Survey (NHMS) 2016. Public
Health, 153, 52–​57. doi:10.1016/​j.puhe.2017.08.001
Maharani, A., & Tampubolon, G. (2014). Has decentralisation affected child immunisation
status in Indonesia? Global Health Action, 7, 24913. doi:10.3402/​gha.v7.24913
McLaughlin, J. M., Swerdlow, D. L., Khan, F., Will, O., Curry, A., Snow, V., . . . Jodar, L. (2019).
Disparities in uptake of 13-​valent pneumococcal conjugate vaccine among older adults in
the United States. Human Vaccines & Immunotherapeutics, 15(4), 841–​849. doi:10.1080/​
21645515.2018.1564434
Menzies, N. A., Suharlim, C., Geng, F., Ward, Z. J., Brenzel, L., & Resch, S. C. (2017). The cost
determinants of routine infant immunization services: A meta-​regression analysis of six
country studies. BMC Medicine, 15(1), 178. doi:10.1186/​s12916-​017-​0942-​1
Menzies, N. A., Suharlim, C., Resch, S. C., & Brenzel, L. (2020). The efficiency of routine in-
fant immunization services in six countries: A comparison of methods. Health Economics
Review, 10(1), 1. doi:10.1186/​s13561-​019-​0259-​1
Mosadeghrad, A. M. (2012). A conceptual framework for quality of care. Materia Socio-​medica,
24(4), 251–​261. doi:10.5455/​msm.2012.24.251-​261
O’Leary, S. T., Allison, M. A., Lindley, M. C., Crane, L. A., Hurley, L. P., Brtnikova, M., . . . Kempe,
A. (2014). Vaccine financing from the perspective of primary care physicians. Pediatrics,
133(3), 367–​374. doi:10.1542/​peds.2013-​2637
O’Malley, A. S., & Forrest, C. B. (2006). Immunization disparities in older Americans: Determinants
and future research needs. American Journal of Preventive Medicine, 31(2), 150–​158.
doi:10.1016/​j.amepre.2006.03.021
Results for Development. (2017). Immunization financing: a resource guide for advocates, pol-
icymakers, and program managers. https://​r4d.org/​resour​ces/​immun​izat​ion-​financ​ing-​
resou​rce-​guide-​advoca​tes-​polic​ymak​ers-​prog​ram-​manag​ers/​
80 Economics of vaccine delivery

Robinson, C. A., Sepe, S. J., & Lin, K. F. (1993). The president’s child immunization initiative—​a
summary of the problem and the response. Public Health Reports (Washington, D.C.: 1974),
108(4), 419–​425. https://​www.ncbi.nlm.nih.gov/​pmc/​artic​les/​PMC​1403​402/​
Saxenian, H., Hecht, R., Kaddar, M., Schmitt, S., Ryckman, T., & Cornejo, S. (2015). Overcoming
challenges to sustainable immunization financing: Early experiences from GAVI graduating
countries. Health Policy and Planning, 30(2), 197–​205. doi:10.1093/​heapol/​czu003
Schwartz, J. B., & Bhushan, I. (2004). Improving immunization equity through a public-​
private partnership in Cambodia. Bulletin of the World Health Organization, 82(9), 661–​
667. https://​www.ncbi.nlm.nih.gov/​pmc/​artic​les/​PMC​2622​985/​
Sharma, A., Kaplan, W. A., Chokshi, M., Hasan Farooqui, H., & Zodpey, S. P. (2015). Implications
of private sector Hib vaccine coverage for the introduction of public sector Hib-​containing
pentavalent vaccine in India: Evidence from retrospective time series data. BMJ Open, 5(2),
e007038. doi:10.1136/​bmjopen-​2014-​007038
Silver, D., Blustein, J., & Weitzman, B. C. (2012). Transportation to clinic: Findings from a pilot
clinic-​based survey of low-​income suburbanites. Journal of Immigrant and Minority Health,
14(2), 350–​355. doi:10.1007/​s10903-​010-​9410-​0
Soeung, S. C., Grundy, J., Morn, C., & Samnang, C. (2008). Evaluation of immunization know-
ledge, practices, and service-​delivery in the private sector in Cambodia. Journal of Health,
Population and Nutrition, 26(1), 95–​104. https://​www.ncbi.nlm.nih.gov/​pmc/​artic​les/​PMC​
2740​687/​
Songane, M. (2018). Challenges for nationwide vaccine delivery in African countries.
International Journal of Health Economics and Management, 18(2), 197–​219. doi:10.1007/​
s10754-​017-​9229-​5
Tanahashi, T. (1978). Health service coverage and its evaluation. Bulletin of the World Health
Organization, 56(2), 295–​303. https://​www.ncbi.nlm.nih.gov/​pmc/​artic​les/​PMC​2395​571/​
Tayloe, D. T. (2009). Immunization financing: Key area for American Academy of Pediatrics
advocacy. Pediatrics, 124(Suppl 5), S455. doi:10.1542/​peds.2009-​1542C
World Health Organization. (2010). Monitoring the building blocks of health systems: A hand-
book of indicators and their measurement strategies. https://​apps.who.int/​iris/​bitstr​eam/​han​
dle/​10665/​258​734/​978924​1564​052-​eng.pdf
World Health Organization. (2013). Global vaccine action plan 2011–​2020. https://​www.who.
int/​publi​cati​ons/​i/​item/​glo​bal-​vacc​ine-​act​ion-​plan-​2011-​2020
World Health Organization. (2018). WHO guidance note: Engagement of private providers
in immunization service delivery. Considerations for national immunization programmes.
https://​www.who.int/​publi​cati​ons/​i/​item/​WHO-​IVB-​17.15
World Health Organization: Regional Office for the Western Pacific. (2012). Malaysia health
system review. https://​apps.who.int/​iris/​han​dle/​10665/​206​911
 2
ESTIMATING THE COST OF
IMMUNIZATION SERVICES
Edited by Logan Brenzel
2.0
Section introduction: estimating
the cost of immunization services
Logan Brenzel

Immunization costing is concerned with identifying, measuring, and valuing

the resources used to deliver immunization services. Resources are defined as
the inputs used to vaccinate, such as labor time, vaccines, and medical sup-
plies. Immunization services pertain to vaccinations offered through the na-
tional immunization program, at a district level, as part of a campaign or other
delivery strategy, or through a new vaccine introduction initiative.
It is useful to distinguish primary costing studies from cost modeling ex-
ercises. When resources have been (or are being) consumed to deliver an
already-​implemented immunization service, a primary costing study is car-
ried out to estimate the quantities of resources consumed and their associ-
ated monetary value. The study will involve collecting and analyzing primary
data on resource use, program output, and other program characteris-
tics. Data sources typically include a mix of sources including financial and
other administrative records, surveys, interviews, and—​sometimes—​direct
observation.
In contrast, some costing modeling exercises aim to forecast the cost of
implementing a new immunization service, scaling up existing services, or
some other change to immunization policy, prior to these changes occurring,
in order to set budgets, mobilize resources, or inform decisions about whether
or not to move forward with a proposed policy change. These cost modeling
exercises typically draw on information regarding the cost of immunization
services that have already been implemented, and thus depend, in part, on in-
formation from primary costing studies.
This section of the handbook covers methods for primary costing studies
of implemented vaccine programs as well as cost modeling exercises for
projecting the resource requirement of (future) new vaccine introduction.

Logan Brenzel, Section introduction: estimating the cost of immunization services In: Handbook of Applied Health Economics in
Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0007
84 Estimating the cost of immunization services

Chapter 2.1 motivates the section with an overview of the many uses of cost
estimates from primary studies, and helps readers navigate a large set of ex-
isting guides, tools, and other technical resources for immunization costing.
Chapter 2.2 provides an overview of key concepts and definitions. Chapter 2.3
covers key elements of costing study design. Chapter 2.4 covers the methods
for data analysis. Chapter 2.5 illustrates how to apply methods for evaluating new
vaccine introduction. This section also links to two appendices that contain exer-
cises related to costing a routine immunization program (Appendix 1, Costing
Exercise) and introducing a new vaccine (Appendix 2, Costing New Vaccine
Introduction). Additional content on (Costing Exercise and Costing New
Vaccine Introduction) is available online, 10.1093/oso/9780192896087.012.0001.
2.1
Why costing studies are needed
Ann Levin, Stephen Resch, and Logan Brenzel

2.1.1 Introduction

Information regarding the cost of immunization services can be useful for

a range of policy and program management processes. Estimating the cost
of immunization programs is important for several reasons, including: (1)
strategic planning and program budgeting; (2) assessing variation in service
delivery costs and analysis of program efficiency or quality; (3) developing
cost benchmarks; and (4) conducting cost-​effectiveness analysis (CEA), cost–​
benefit analysis (CBA), and return on investment analysis to inform resource
allocation decisions. The intended purpose of the analysis will determine sub-
sequent study design choices, including the methods and the type of data to
be collected.

2.1.2 Immunization program strategic planning

and budgeting

Program managers can use information on the cost of immunization serv-

ices to plan for implementation of their program activities, prepare immu-
nization program budgets, develop multi-​year strategic plans, and inform
policy decisions such as whether to introduce a new vaccine. Managers can
understand the relative magnitude of resource requirements for different
vaccination activities, including the inputs or activities that account for the
largest shares of total cost (cost drivers). Costing studies can offer insights
regarding the share of program cost attributable to categories of resources
(vaccines, labor), program activities (supply chain, facility-​based vaccine ad-
ministration, outreach), or organizational level (health facility, district, na-
tional). Understanding what may be driving up costs of services can lead to

Ann Levin, Stephen Resch, and Logan Brenzel, Why costing studies are needed In: Handbook of Applied Health Economics in
Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0008
86 Why costing studies are needed

better management choices in how resources can be utilized. Another output

of costing studies useful for program planning is a set of activity-​specific unit
costs such as cost per dose delivered via campaigns, or cost per dose delivered
via routine facility-​based services.
Evidence generated by costing studies can subsequently be used in cost mod-
eling exercises to project the resources required in the future under different
policy scenarios. This type of cost modeling is a common part of strategic pla-
nning used to ensure program goals are aligned with the expected available
resources and to support mobilization of additional resources when needed.
Estimates of annual or multi-​year resource requirements can be used to advo-
cate for funding for the program or new interventions such as a new vaccine
introduction. Cost projections can also factor into budget impact analysis of
adding a new vaccine or delivering services in a different manner.

2.1.3 Assessing variation in the cost

of immunization service delivery

The analyst can use information on unit costs to compare variation in the cost
of vaccination among health facilities, districts, or other administrative levels,
or for different types of service delivery.
Cost estimation is important for assessing the technical efficiency of vac-
cination, which is how effective the program is in turning a set of inputs into
a performance output (vaccinated child or dose delivered). A recent analysis
(Geng, Suharlim, Brenzel, Resch, & Menzies, 2017) compares total and unit
costs of delivering a dose of vaccine for a sample of facilities and subnational
administrative units in six countries to determine which locations provided
the greatest level services for the least amount of cost. Sites that were the most
efficient were those that had lower shares of labor costs, either because of less
time allocated or different staff complements, or both. This information can
then be used to identify ways to improve the efficiency of service delivery.
Hence, cost information can be used to determine whether specific inputs,
such as personnel time, are translating into productive outputs.

2.1.4 Developing benchmarks

Another reason to estimate immunization costs is to develop unit cost bench-

marks such as cost per vaccine dose administered or cost per fully immunized
 2.1.5 Economic evaluations 87

child. If country-​specific information is unavailable, cost benchmarks can

be helpful in filling the gap in information. A recent dataset of immuniza-
tion delivery costs was developed (Portnoy et al., 2020). The World Health
Organization (WHO) Choosing Interventions that are Cost-​ Effective
(CHOICE) program created benchmarks for a range of healthcare costs
(Stenberg, Lauer, Gkountouras, Fitzpatrick, & Stanciole, 2018). Benchmarks
can also be used to develop or evaluate grant applications or as inputs to cost
modeling exercises projecting financial requirements of a policy change.

2.1.5 Economic evaluations

Economic evaluations such as CEA, CBA, and return on investment calcu-

lations can be used to make choices among vaccination delivery strategies,
among new vaccines, and among new vaccination technologies to maximize
the use of available resources to obtain health impacts. To conduct these ana-
lyses, it is necessary to have data on both costs and effectiveness or benefits.
CEA and CBA are useful for determining whether a health intervention pro-
vides good value for money (compared to some alternative) and are used to
inform resource allocation decisions. A CEA reflects the change in costs asso-
ciated with a new intervention such as new vaccine introduction, compared
to a change in effectiveness of the program, such as number of deaths or disa-
bility adjusted life-​years (DALYs) averted. For instance, a policymaker might
be interested to know whether it is a good investment to introduce a measles–​
rubella vaccine into their program: a study comparing the incremental costs
with the incremental benefits would provide the answer to that question. There
is an extensive literature on the cost-​effectiveness of vaccines and technolo-
gies for the immunization program (Edoka et al., 2021; Jit, Brisson, Portnoy,
& Hutubessy, 2014; Le et al., 2015; Levin, Levin, Kristensen, & Matthias, 2007;
Novaes et al., 2015; Ozawa et al., 2017). A systematic review found that the
cost/​DALY of new vaccines ranges from less than $100 to less than $1,000 per
DALY averted (Ozawa, Mirelman, Stack, Walker, & Levine, 2012). Chapters 3
and 4 go into further detail on how to conduct these types of analyses.
CBAs are similar to CEAs except that the benefits of the interventions are
monetized rather than expressed as units of health such as deaths, life-​years,
or DALYs. A CBA calculates the net monetary benefit of a program (com-
pared to a next-​best alternative). This allows for the immunization program
to be compared to other health and non-​health interventions, such as com-
paring with a nutrition program to determine the best allocation of resources.
88 Why costing studies are needed

Return on investment analyses are used to estimate the ratio between the
monetary benefits of a vaccination program divided by the costs. For instance,
a recent study showed that for every $1 invested in the vaccines supported by
Gavi, countries would yield a return of $21 based on productivity losses due
to illness (Sim et al., 2020). A key driver of vaccine value is often treatment
costs that are averted. Some economic evaluations may also include health-
care costs related to vaccination or vaccine-​preventable disease, as well as pro-
ductivity impacts.

2.1.6 Resources available on immunization costing

This part of the chapter describes the key resources for readers related to
costing methodologies and application to field-​based studies.

2.1.6.1 Resource guides

A number of resource guides are available on methods for immunization

costing. Several resource guides have come out of a Community of Practice on
immunization economics, including “ How to cost immunization programs: A
practical guide for primary data collection and analysis.” This guide discusses
methods involved for cross-​sectional retrospective costing of routine im-
munization services, derived from a sample of health facilities and provides
standardized terminology and costing units, approaches to data collection
and data requirements, guidance on facility sampling, and recommendations
for analysis, write-​up, and presentation (Resch et al., 2020).
Another resource, “How to conduct an immunization campaign costing
study: Methodological guidance,” provides guidance on costing of a wide va-
riety of immunization campaigns, such as a national planned campaign, an
integrated campaign that provides multiple vaccines or vaccines and other
health services, or outbreak response campaigns (ThinkWell, 2021).
The Global Health Cost Consortium’s reference case was developed for
improving the comparability and quality of cost estimates for health interven-
tions and is also relevant to immunization programs. This guide focuses not
only on definitions of costing terms but also on principles of “good” costing
organized around a checklist (Vassall et al., 2017). For example, one prin-
ciple is to define the study design: purpose of the analysis, the perspective,
types of cost, time horizon, and the definition of units. Other principles are
 2.1.6 Resources available on immunization costing 89

resource use measurement (such as scope of costing, measuring resource use,

and sampling), pricing and valuation (sources of price data, valuing capital
inputs, and discount, inflation, and conversion rates), and analyzing and pre-
senting results (exploring cost functions and heterogeneity, uncertainty, and
transparency).
The Second US Panel on Cost-​Effectiveness in Health and Medicine also
provides useful guidance (Neumann, Ganiats, Russell, Sanders, & Siegel,
2016). This update to the first US Panel on Cost-​effectiveness in Health and
Medicine (Gold, Siegel, Russell, & Weinstein, 1996) provides recommenda-
tions on methods for costing and CEA, such that reference case analyses
should include societal as well as health sector perspectives. Regarding costs,
it makes recommendations about the components that should be included in
costing of health services, inclusion of future costs, and discounting.
The WHO also has developed some resource guides and tools for costing.
The WHO “Guideline for estimating the costs of introducing new vaccines
into the national immunization system” contains a useful framework for
identifying which cost elements will be affected by new vaccine introduction
and how to project incremental costs. For instance, new vaccines will require
additional space in the national cold chain and the guide highlights methods
for estimating additional needs and associated costs (WHO: Immunization
Vaccines and Biologicals, 2002). The WHO “Guide for standardization of ec-
onomic evaluations of immunization programs” focuses on methods associ-
ated with different types of economic evaluations, and includes approaches to
modeling, discounting, and interpretation of results (WHO: Immunization
Vaccines and Biologicals, 2019).
The Pan American Health Organization developed COSTVAC which pro-
vides a template for selecting a sample of facilities, and analyzing cost data
(Pan American Health Organization, 2019).
The WHO also has developed costing tools for estimation of incremental
costs of either introducing new vaccines or estimating retrospective costs
of single-​vaccine implementation. These costing tools focus specifically on
introduction of a single antigen, such as the Cervical Cancer Prevention
and Control Costing Tool (Hutubessy et al., 2012) for estimation of human
papillomavirus vaccination cost and the Seasonal Influenza Immunization
Costing Tool (Pallas et al., 2020) for estimation of annual costs of
introducing seasonal influenza vaccination to populations at highest risk,
such as pregnant women and health workers. Other tools developed jointly
with the International Vaccine Institute include CholTool, which can be
used to estimate projected or retrospective incremental costs of preventive
90 Why costing studies are needed

and outbreak responses with oral cholera vaccine campaigns (Morgan,

Levin, Hutubessy, & Mogasale, 2020), and the Typhoid Conjugate Vaccine
Costing Tool (TCVCT). Finally, WHO and UNICEF developed a spread-
sheet tool to estimate the resource requirements for implementing the na-
tional immunization programs (comprehensive Multi-​Year Plan Costing
Tool; WHO: Immunization Vaccines and Biologicals, 2014). This tool helps
countries to plan and identify needed financing for the program in order to
achieve programmatic objectives.
Table 2.1.1 compares and contrasts the recommended uses of the var-
ious resources and tools available. This handbook draws from much of the
work that has proceeded, and readers are encouraged to utilize this core
material.

Table 2.1.1 Comparison of guides and tools for costing immunization programs

Resource guides and costing Main emphasis Recommended use

tools

How to cost immunization Application of costing methods Methodological

programs: A practical guide to retrospective primary cost guidance, design, and
for primary data collection and data collection, analysis, and implementation of field-​
analysis policy dialogue on costs of based studies in lower-​
routine immunization programs income settings
How to conduct an Application of costing methods Design and
immunization campaign to retrospective primary cost implementation of field-​
costing study: Methodological data collection, analysis, and based studies in lower-​
guidance policy dialogue on costs of income settings
immunization campaigns
Global Health Cost Definitions and methods Reference and
Consortium reference case methodological
guidance
Second Panel on Cost-​ Methods for costing and CEA Reference and
Effectiveness in Health and methodological
Medicine guidance
WHO guideline for estimating Application of costing methods Methodological
the costs of introducing new to retrospective primary cost guidance, design, and
vaccines into the national data collection, analysis, and implementation of field-​
immunization system policy dialogue on costs of based studies in lower-​
introducing new vaccines income settings
WHO guide for standardization Methods for different types of Reference and
of economic evaluations of economic evaluations methodological
immunization programs guidance
PAHO COSTVAC Methodology and tool for Methods, sampling,
analyzing health facility and practical cost
immunization costs estimation tool
 2.1.6 Resources available on immunization costing 91

Table 2.1.1 Continued

Resource guides and costing Main emphasis Recommended use

tools

WHO Cervical Cancer Data collection and analysis, Estimating the cost of
Prevention and Control policy dialogue human papillomavirus
Costing Tool vaccination and other
interventions to control
cervical cancer
WHO vaccine-​specific costing Data collection and analysis, Estimating the cost of
tools (Seasonal Influenza policy dialogue influenza, cholera, and
Immunization Costing Tool; typhoid vaccination
CholTool; TCVCT)
Comprehensive Estimation of immunization-​ 3–​5-​year planning of the
Multi-​Year Plan specific and shared costs for resource needs for the
Costing Tool achieving national program national immunization
objectives program
National Immunization Estimation of costs and budgets Annual and up to 5-​year
Strategy Costing Tool for achieving national program planning and budgeting
objectives for the national
immunization program
Consensus Statement Review and recommendation for Reference
future costing work
Immunization Delivery Cost Up-​to-​date compendium of Reference, country
Compendium (IDCC) published and unpublished results, and comparisons
immunization costing studies
DataVerse Datasets from country Data analysis
immunization costing studies
Teaching Vaccine Economics Teaching materials related Teaching
Everywhere (TVEE) to economic evaluation of
immunization programs

2.1.6.2 Other resources

The various resources on immunization and health costing have commonal-

ities as well as differences in their methodologies. Most of these differences can
be attributed to the distinct purposes and objectives of the guides. Four dimen-
sions of immunization cost estimation have been identified: (1) retrospective
routine immunization cross-​sectional costs, (2) retrospective single-​vaccine
costs, (3) projection of new vaccine introduction costs, and (4) projection of
national immunization program costs. Recently, the WHO led the develop-
ment of a Consensus Statement in order to harmonize and better align the prin-
ciples, terminologies, and approaches used for immunization costing studies
(Levin et al., 2022). The Consensus Statement will serve as a reference point to
92 Why costing studies are needed

promote continued improvement and innovation in methods and tools, as well

as to inform the principles and definitions that will make costing results more
easily interpretable and useful.
The Immunization Delivery Cost Catalogue contains a wealth of published
and unpublished immunization costing studies (ThinkWell, 2020). This down-
loadable database provides estimates on cost per dose and cost per child, as well
as the characteristics and details on the methods used in the individual studies.
This resource contains 192 datapoints from 62 individual country studies
from 2015.
Downloadable datasets from immunization costing studies from a variety
of countries and covering a range of immunization strategies are available
from DataVerse (https://​datave​rse.harv​ard.edu/​datave​rse/​EPIC).
Finally, teaching materials, course outlines, case studies, and curricula re-
lated to economic evaluation of immunization programs can be accessed on-
line (Teaching Vaccine Economics Everywhere, 2021).

2.1.7 Conclusion

Immunization costing studies generate essential strategic information that

program managers and policymakers require for planning and budgeting,
policy decision-​making, assessing technical efficiency, developing bench-
marks, and economic evaluation to inform resource allocation decisions.
While primary costing studies measure resource use for immunization serv-
ices that have been implemented or are ongoing, the resulting cost estimates
can be used as an input to cost modeling exercises designed to inform deci-
sions about future immunization activities.

References
Edoka, I., Kohli-​Lynch, C. N., Fraser, H., Hofman, K., Tempia, S., McMorrow, M., . . . Cohen,
C. (2021). A cost-​effectiveness analysis of South Africa’s seasonal influenza vaccination pro-
gramme. Vaccine, 39(2), 412–​422. doi:10.1016/​j.vaccine.2020.11.028
Geng, F., Suharlim, C., Brenzel, L., Resch, S. C., & Menzies, N. A. (2017). The cost structure of
routine infant immunization services: A systematic analysis of six countries. Health Policy
and Planning, 32(8), 1174–​1184. doi:10.1093/​heapol/​czx067
Gold, M. R., Siegel, J., Russell, L., & Weinstein, M. (1996). Cost-​effectiveness in health and medi-
cine. New York, NY: Oxford University Press.
Hutubessy, R., Levin, A., Wang, S., Morgan, W., Ally, M., John, T., & Broutet, N. (2012). A case
study using the United Republic of Tanzania: Costing nationwide HPV vaccine delivery
using the WHO cervical cancer prevention and control costing tool. BMC Medicine, 10(1),
136. doi:10.1186/​1741-​7015-​10-​136
 2.1.7 Conclusion 93

Jit, M., Brisson, M., Portnoy, A., & Hutubessy, R. (2014). Cost-​effectiveness of female human
papillomavirus vaccination in 179 countries: A PRIME modelling study. Lancet Global
Health, 2(7), e406–​414. doi:10.1016/​s2214-​109x(14)70237-​2
Le, P., Griffiths, U. K., Anh, D. D., Franzini, L., Chan, W., & Swint, J. M. (2015). Cost-​
effectiveness of Haemophilus influenzae type b vaccine in Vietnam. Vaccine, 33(36), 4639–​
4646. doi:10.1016/​j.vaccine.2015.05.050
Levin, A., Levin, C., Kristensen, D., & Matthias, D. (2007). An economic evaluation of ther-
mostable vaccines in Cambodia, Ghana and Bangladesh. Vaccine, 25(39), 6945–​6957.
doi:10.1016/​j.vaccine.2007.06.065
Levin, A., Boonstoppel, L., Brenzel, L., Griffiths, U., Hutubessy, R., Jit, M., . . . Yeung, K.H.T.
(2022). WHO-​led consensus statement on vaccine delivery costing: Process, methods and
findings. BMC Medicine, 20(1), 88. doi:10.1186/​s12916-​022-​02278-​4
Morgan, W., Levin, A., Hutubessy, R. C., & Mogasale, V. (2020). Costing oral cholera vac-
cine delivery using a generic oral cholera vaccine delivery planning and costing tool
(CholTool). Human Vaccines & Immunotherapeutics, 16(12), 3111–​3118. doi:10.1080/​
21645515.2020.1747930
Neumann, P., Ganiats, T., Russell, L., Sanders, G., & Siegel, J. (2016). Cost-​effectiveness in health
and medicine. New York, NY: Oxford University Press.
Novaes, H. M., de Soárez, P. C., Silva, G. A., Ayres, A., Itria, A., Rama, C. H., . . . Resch, S. (2015).
Cost-​effectiveness analysis of introducing universal human papillomavirus vaccination of
girls aged 11 years into the National Immunization Program in Brazil. Vaccine, 33(Suppl 1),
A135–​A142. doi:10.1016/​j.vaccine.2014.12.031
Ozawa, S., Clark, S., Portnoy, A., Grewal, S., Stack, M. L., Sinha, A., . . . Walker, D. (2017).
Estimated economic impact of vaccinations in 73 low-​and middle-​income countries, 2001–​
2020. Bulletin of the World Health Organization, 95(9), 629–​638. doi:10.2471/​blt.16.178475
Ozawa, S., Mirelman, A., Stack, M. L., Walker, D., & Levine, O. S. (2012). Cost-​effectiveness and
economic benefits of vaccines in low-​and middle-​income countries: A systematic review.
Vaccine, 31(1), 96–​108. doi:10.1016/​j.vaccine.2012.10.103
Pallas, S. W., Ahmeti, A., Morgan, W., Preza, I., Nelaj, E., Ebama, M., . . . Bino, S. (2020). Program
cost analysis of influenza vaccination of health care workers in Albania. Vaccine, 38(2), 220–​
227. doi:10.1016/​j.vaccine.2019.10.027
Pan American Health Organization. (2019). COSTVAC. http://​immuni​zati​onec​onom​ics.org/​
rec​ent-​activ​ity/​2019/​9/​23/​upd​ate-​of-​pahos-​pro​vac-​e-​tool​kit
Portnoy, A., Vaughan, K., Clarke-​Deelder, E., Suharlim, C., Resch, S. C., Brenzel, L., & Menzies,
N. A. (2020). Producing standardized country-​level immunization delivery unit cost esti-
mates. PharmacoEconomics, 38(9), 995–​1005. doi:10.1007/​s40273-​020-​00930-​6
Resch, S. C., Menzies, N. A., Portnoy, A., Clarke-​Deelder, E., O’Keefe, L., Suharlim, C., &
Brenzel, L. (2020). How to cost immunization programs: A practical guide for primary data
collection and analysis. ImmunizationEconomics.org. http://​www.immuni​zati​onec​onom​ics.
org/​epic
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353. doi:10.1377/​hlthaff.2020.00103
Stenberg, K., Lauer, J. A., Gkountouras, G., Fitzpatrick, C., & Stanciole, A. (2018). Econometric
estimation of WHO-​CHOICE country-​specific costs for inpatient and outpatient health
service delivery. Cost Effectiveness and Resource Allocation, 16(1), 11. doi:10.1186/​
s12962-​018-​0095-​x
Teaching Vaccine Economics Everywhere. (2021). Teaching vaccine economics everywhere.
http://​www.immuni​zati​onec​onom​ics.org/​tvee
ThinkWell. (2020). Immunization delivery cost catalogue. http://​www.immuni​zati​onec​onom​
ics.org/​ican
94 Why costing studies are needed

ThinkWell. (2021). How to conduct an immunization campaign costing study: Methodological

guidance. http://​immuni​zati​onec​onom​ics.org/​ican-​idcc-​meth​odol​ogy
Vassall, A., Sweeney, S., Kahn, J., Gomez, G. B., Bollinger, L., Marseille, E., . . . Levin, C. (2017).
Reference case for estimating the costs of global health services and interventions. LSHTM
Research Online. https://​res​earc​honl​ine.lshtm.ac.uk/​id/​epr​int/​4653​001/​1/​vassall_​etal_​
2018_​reference_​case_​for_​estimating_​co ​ sts​_g​ lo​bal_​heal​th_​se​ rvi​ces.pdf
World Health Organization: Immunization Vaccines and Biologicals. (2002). Guidelines for
estimating costs of introducing new vaccines into the national immunization system. https://​
apps.who.int/​iris/​han​dle/​10665/​67342
World Health Organization: Immunization Vaccines and Biologicals. (2014). Comprehensive
multi-​ year plan costing tool. https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​128​051/​
WHO​_I​ VB​_1​ 4.06_​eng.pdf
World Health Organization: Immunization Vaccines and Biologicals. (2019). WHO guide for
standardization of economic evaluations of immunization programmes. https://​apps.who.int/​
iris/​bitstr​eam/​han​dle/​10665/​329​389/​WHO-​IVB-​19.10-​eng.pdf
2.2
Defining immunization costs
Logan Brenzel

2.2.1 Defining costs

Chapters 1.4 and 1.5 provide an overview of why immunizations and vaccines
might be undersupplied and under-​consumed because of its dual public and
private goods nature. In competitive markets, prices are equal to the marginal
cost of the inputs that go into producing a product (McCaffrey, 2018). Total
cost is a function of outputs, prices, and other factors. A cost function de-
scribes the relationship between exogenous prices and volume of outputs pro-
duced. Imperfect markets, like the one for vaccines, do not lead to a situation
where marginal costs are equal to prices.
The assumption of cost minimization applies to competitive markets
where there are many buyers and sellers all making a similar product, lim-
ited barriers to entry, and full information. However, this scenario does not
necessarily apply to health service provision in the public sector given the het-
erogeneity and uncertainty of healthcare needs and services, the imbalance
of information between the patient and the healthcare provider, as well as the
possible absence of incentives. In some countries, public healthcare workers
are paid the same wage whether they vaccinate 100 or 10 children per day, or
whether they provide a high or low quality of care. The management and use
of health sector inputs, such as medicines, vaccines, and equipment, may not
be directed toward cost minimization in environments, due to supply con-
straints, poor information, and lack of data to be able to continually value and
monitor costs.
Chapter 2.1 laid out different reasons for conducting an analysis of im-
munization programs and service delivery costs. Dedicated costing studies
of vaccines and immunization programs in lower-​ income settings have
been conducted since the 1980s, at a time when countries were focused on
achieving universal childhood immunization (Brenzel & Claquin, 1994).

Logan Brenzel, Defining immunization costs In: Handbook of Applied Health Economics in Vaccines. Edited by:
David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0009
96 Defining immunization costs

Because of weaknesses in financial and data systems in lower-​income coun-

tries, costing is usually accomplished through an analysis of the inputs re-
quired to deliver services, assigning prices, and approximate usage for the
program if inputs are shared. Presently, the population-​weighted average ec-
onomic delivery cost per dose (net of vaccine costs) is $1.87, with labor costs
accounting for the greatest proportion of costs (Portnoy et al., 2020). The total
cost of fully vaccinating a child against measles, pertussis, diphtheria, polio,
rotavirus, human papilloma virus, and pneumococcal disease is approxi-
mately $54, though this will vary depending upon the delivery context (Sim,
Watts, Constenla, Brenzel, & Patenaude, 2020).
Variation in the unit cost per dose across different geographical areas and
between different types of providers has been documented in the literature
(Brenzel, Young, & Walker, 2015; Chatterjee et al., 2018; Fox-​Rushby, Kaddar,
Levine, & Brenzel, 2004; Geng, Suharlim, Brenzel, Resch, & Menzies, 2017;
Menzies et al., 2017; ThinkWell, 2020). Variation also exists between the
costs of delivering immunization services through different strategies, such
as routine services provided through fixed facilities or a campaign approach.
A compendium of immunization delivery costs finds that most studies have
been conducted in African countries. Immunization costs also appear to vary
by socioeconomic level of the country, with a lower average cost per dose in
lower-​income countries (Sim et al., 2020).

2.2.2 Cost definitions and categorizations

Methods for evaluating the costs of immunization programs involve meas-

uring economic, financial, or fiscal costs (Resch et al., 2020). An economic
cost is the value of all inputs used to provide immunization services with
respect to trade-​offs of where else those same inputs could have been em-
ployed. Economic costs include the value of time spent by health workers,
value of donated goods and items, and the proportional value of shared in-
puts, such as transport or equipment. For each one of these inputs, choices
must be made on investments in health or other common alternatives uses.
Thus, economic costs are resources foregone by society and are viewed from
the perspective of society.
Economic costs can also be thought of as the opportunity cost of investing
in one program relative to investing in another to achieve the same set
of outcomes. If a country has a fixed level of resources to devote to saving
children’s lives, the policymaker can invest that money in a range of services
 2.2.2 Cost definitions and categorizations 97

and must decide about how to allocate those resources. If resources are in-
vested in immunization programs, those resources are not available for other
uses. However, the return on investment in immunization may far outweigh
that for other services (Vaughan et al., 2019). Evaluating trade-​offs between
programs is the domain of cost-​effectiveness analysis (see Chapter 4).
A financial cost focuses on the cash outlays incurred for the delivery of a
program. In viewing costs in this way, donated goods would be free and
therefore not included. Financial costs can be viewed from the perspective
of the payer (see Chapter 2.3), and costs are resources foregone by the payor.
Financial costs are often used in immunization costing evaluations as these
can be linked to key policy considerations of affordability, sustainability, and
budget impact.
A further refinement of financial cost is fiscal cost, which refers to actual ex-
penditures made. For instance, the entire amount spent on cold chain equip-
ment could be assigned to the time period in which the purchase occurred.
For instance, a refrigerator purchased for $1,000 in September 2020 will rep-
resent $1,000 of fiscal expenditure in 2020. While in a financial cost analysis,
the refrigerator is assumed to be used over a period of several years, and the
corresponding cost is annualized for each year.
The language around costs can be confusing. Published costing studies
may not have explicitly stated what type of costs are being evaluated (Brenzel,
2014). Readers of this handbook are encouraged to carefully classify, charac-
terize, and report the type of costs in their analysis. Furthermore, the terms
cost, budget, expenditures, and prices are often used interchangeably, but these
represent different values. A budget is planned spending which has not yet oc-
curred. In the absence of expenditure data, one might use budgeted amounts
as a proxy for spending. But actual expenditure can vary substantially from
budgets because of delays in disbursement and lower expenditure rates.
In order to reflect the societal perspective in costing studies, the analysis
should take into consideration the costs incurred by households to seek im-
munization services. These costs can include transportation and waiting
times, as well as lost productivity from missing work. Few immunization
costing studies incorporate the full societal perspective and include house-
hold costs, primary because of the additional cost for data collection through
household surveys or facility exit interviews. Nevertheless, when evaluating
the costs of different strategies, households may incur varying levels of costs.
Fixed facility services require families to visit facilities which are often far
away, entailing missed work time and often long waiting times. Campaigns
that are offered in central locations may be closer to the community and the
98 Defining immunization costs

household, though the density of the population may entail longer waiting
times. Outreach services in which providers go to remote areas are likely to
reduce the costs for households in seeking services.

2.2.3 Immunization program cost characterization

Table 2.2.1 illustrates the range of inputs that are evaluated in a typical immu-
nization costing study. Labor costs entail the time spent by a range of health
workers, such as medical doctors, nurses, nurse aides, and community health
workers, to administer the vaccine dose, and also to undertake record keeping,
travel to obtain vaccines and supplies, advocate within the community to come
for vaccination, and to repair and maintain equipment. Because vaccines need
to be maintained at lower temperatures (2–​8°C for most vaccines), the use of
vaccine carriers, cold boxes, refrigerators, and freezers is critical to successful
vaccination delivery. These appliances need to be maintained and have associ-
ated running costs. Often, immunization services are provided in fixed facilities
or through outreach and immunization campaigns. Travel to vaccination sites
(urban or remote) entails use of resources (vehicles, fuel, per diem) that need to
be included in costing studies. In economic cost studies, the economic value of
use of building space is also valued. Strategies and approaches to data collection
and analysis for costing studies are presented and discussed in Chapters 2.3–​2.5.
Costing studies require information on the prices of inputs. Historical
prices are known from records, but they are not reflective of the opportunity
cost of future resources. As such, replacement prices are preferred to histor-
ical prices, particularly for capital inputs in cost analysis. Prices for various
cadres of healthcare workers and administrative personnel include salaries
and allowances and other benefits. Prices may not be available for volunteer
labor and studies often use an average daily wage which can be obtained
from government surveys or the International Labor Organization (2021).
Because prices may reflect the quality of inputs, it is important to control
for differences in quality of care in the cost evaluation. Selecting a best prac-
tice standard for an input may help in this regard. The prices of some inputs
such as fuel can vary within a country, and it may be useful to collect price
information in the sample areas to account for this. The prices of vaccines or
cold chain equipment should also reflect costs related to shipping, interna-
tional freight, and insurance. The freight-​on-​board price reflects additional
costs for loading and unloading, taxes and other fees to the point of delivery.
 2.2.3 Immunization program cost characterization 99

Table 2.2.1 Comparison of economic and financial cost inputs

Line item Economic cost Financial cost

Paid labor Value of time allocated to Expense made for hiring new
immunization for existing or staff for immunization-​specific
new staff activities
Volunteer labor Value of time allocated to Not included
immunization benchmarked
against average daily wage
Per diem Value of per diems given to health Expenditures on per diems
workers
Vaccines Value of vaccines administered and Purchase expense for vaccines
wasted
Injection supplies Value of syringes, wastage Purchase expense for syringes,
containers used and wasted wastage containers
Fuel and Value related to use of fuel for Expense on fuel and
transportation immunization-​specific activities transportation
Vehicle Value of maintenance related to Expense for maintenance
maintenance immunization-​specific activities
Cold chain energy Estimated value of cold chain Expense for butane or other
costs energy costs energy to run the cold chain
Waste disposal Value of waste boxes and Purchase of waste boxes, and
incineration of used vaccination expense for incineration adjusted
supplies (syringes, vaccine vials, using straight line depreciation
gauze, etc.)
Printing Estimated value of printing based on Expense for printing
unit price and volume
Utilities Estimated value of utilities and Expense for utilities
communications based on
immunization share to total services
Other recurrent Estimated value of other recurrent Expenses for other inputs related
inputs inputs based on unit prices, to immunization
allocation to immunization and
quantity used
Cold chain Annualized and discounted value Expense for cold chain adjusted
equipment of cold chain equipment used using straight line depreciation
Vehicles Annualized and discounted value Expense for vehicles used for
of vehicles used immunization adjusted using
straight line depreciation
Other equipment Annualized and discounted value of Expense for other equipment
other equipment (computers) used adjusted using straight line
depreciation
Buildings Annualized and discounted value of Expense for constructing
building space used based on useful buildings related to immunization
life and discount rate services adjusted using straight
line depreciation
100 Defining immunization costs

The cost, insurance, and freight price covers insurance, customs duties, and
rerouting costs associated with the shipment before loading.
Prices may also reflect market distortions. Countries may impose trade and
exchange rate barriers or subsidize the cost of production. Prices reflect in-
clude taxes and government fees that are transfers. An input may be produced
in noncompetitive markets, such as monopolies.
Total immunization costs can be evaluated and reported by input line item
as in Table 2.2.1, or by the type of immunization activity defined in Table 2.2.2.
Activities are comprised of a combination of line items. For instance, social
mobilization and advocacy might entail the cost of television advertising, as
well as travel to communities to bring greater awareness of the benefits of vac-
cination. Appendix 3 provides an example of a crosswalk between line items
and activities for an immunization program.

Table 2.2.2 Immunization service cost activity definitions

Immunization activity Definition

Routine facility-​based Labor time and other resources required to administer vaccines
immunization delivery in a facility, including supplies, use of cold chain equipment,
waste disposal, etc.
Outreach immunization Labor time and other resources required to administer vaccines
delivery outside of the facility, including supplies, transportation, use of
vehicles, use of cold chain, per diem, waste disposal, etc.
Record keeping and data Labor time and resources required to record and manage
management immunization records
Social mobilization and Labor time and resources required to motivate, sensitize, and
advocacy mobilize the population for vaccination services, including radio
and television broadcasting, developing and printing posters,
supporting community events, etc.
Program management Time and resources required to plan, budget, and manage
the delivery of immunization services within a facility or to a
population
Vaccine collection, Time and resources required to collect distribute and store
distribution, and storage vaccines (from port to delivery point), including transportation,
per diem, use of vehicles, etc.
Cold chain maintenance Time and resources required to maintain temperature control
for vaccines
Supervision Time and resources required to undertake supervisory activities
Training Time and resources required (per diem and travel of participants,
developing and printing of materials, renting venues,
refreshments, etc.) to provide in-​service training to vaccinators
and administrators
Surveillance Time and resources required to undertake case detection, follow
up on adverse events or other surveillance activities, including
travel, per diem, and other resources
 2.2.4 Conclusion 101

If the focus of the study is to understand and evaluate the total costs of
a national immunization program or the different strategies of that pro-
gram, all inputs utilized in the program or strategy need to be included
and measured. If the purpose of the analysis is to determine the cost of
introducing a new vaccine, then an incremental cost analysis would be
conducted. An incremental analysis would only focus on the additional
labor time, or the additional use of vehicles and equipment for the new
vaccine. Chapter 2.5 presents an approach to evaluating the introduction
of new vaccines.
Other approaches to costing health services can be applied to costing
immunization programs. For instance, a cost accounting approach called
step-​down cost accounting allocates costs to different departments in
a facility in a sequential process, starting with support services (e.g. ad-
ministration) (Conteh & Walker, 2004). Activity-​based costing is another
approach that measures the relationship between resources, activities, and
performance or outputs (Baker, 1998). This approach can be widely applied
in healthcare and may be useful for estimating the cost of achieving cov-
erage targets.

2.2.4 Conclusion

This chapter has provided the basic structure and concepts for immuni-
zation program costing. For cost-​effectiveness and other economic evalu-
ations, economic costs should be estimated to reflect the full economic costs
of services. Financial costs are more relevant if the policy question relates to
the resources that need to be mobilized or the budget impact of the program.
Household costs are rarely considered in immunization costing studies, al-
though more work would be useful in this area, as service delivery modal-
ities have varying impacts on household costs. More detailed methods and
approaches are found in the following chapters.

References
Baker, J. J. (1998). Activity-​based costing and activity-​based management for health care.
Gaithersburg, MD: Aspen Publishers.
Brenzel, L. (2014). Working Paper: Common approach for the costing and financing analyses
of routine immunization and new vaccine introduction costs (EPIC). Mimeograph. Bill &
Melinda Gates Foundation. www.immuni​zati​onec​onom​ics.org/​epic-​info.
Brenzel, L., & Claquin, P. (1994). Immunization programs and their costs. Social Science &
Medicine, 39(4), 527–​536. doi:10.1016/​0277-​9536(94)90095-​7
102 Defining immunization costs

Brenzel, L., Young, D., & Walker, D. G. (2015). Costs and financing of routine immuniza-
tion: Approach and selected findings of a multi-​country study (EPIC). Vaccine, 33, A13–​
A20. doi:10.1016/​j.vaccine.2014.12.066
Chatterjee, S., Das, P., Nigam, A., Nandi, A., Brenzel, L., Ray, A., . . . Laxminarayan, R. (2018).
Variation in cost and performance of routine immunisation service delivery in India. BMJ
Global Health, 3(3), e000794. doi:10.1136/​bmjgh-​2018-​000794
Conteh, L., & Walker, D. (2004). Cost and unit cost calculations using step-​down accounting.
Health Policy and Planning, 19(2), 127–​135. doi:10.1093/​heapol/​czh015
Fox-​Rushby, J., Kaddar, M., Levine, R., & Brenzel, L. (2004). The economics of vaccination
in low-​and middle-​income countries. Bulletin of the World Health Organization, 82, 640.
doi:10.1590/​S0042-​96862004000900002
Geng, F., Suharlim, C., Brenzel, L., Resch, S. C., & Menzies, N. A. (2017). The cost structure of
routine infant immunization services: A systematic analysis of six countries. Health Policy
and Planning, 32(8), 1174–​1184. doi:10.1093/​heapol/​czx067
International Labor Organization. (2021). Global wage report 2020–​21: Wages and minimum
wages in the time of COVID-​19. https://​www.ilo.org/​wcm​sp5/​gro​ups/​pub​lic/​-​-​-​dgrepo​rts/​-​-​-​
dcomm/​--​ ​-p
​ ubl/​docume​nts/​publ​icat​ion/​wcms​_​762​534.pdf
McCaffrey, M. (2018). The economic theory of cost: Foundation and new directions. London:
Routledge.
Menzies, N. A., Suharlim, C., Geng, F., Ward, Z. J., Brenzel, L., & Resch, S. C. (2017). The cost
determinants of routine infant immunization services: A meta-​regression analysis of six
country studies. BMC Medicine, 15(1), 178. doi:10.1186/​s12916-​017-​0942-​1
Portnoy, A., Vaughan, K., Clarke-​Deelder, E., Suharlim, C., Resch, S. C., Brenzel, L., & Menzies,
N. A. (2020). Producing standardized country-​level immunization delivery unit cost esti-
mates. PharmacoEconomics, 38(9), 995–​1005. doi:10.1007/​s40273-​020-​00930-​6
Resch, S. C., Menzies, N. A., Portnoy, A., Clarke-​Deelder, E., O’Keefe, L., Suharlim, C., &
Brenzel, L. (2020). How to cost immunization programs: A practical guide for primary data
collection and analysis. ImmunizationEconomics.org. http://​www.immuni​zati​onec​onom​ics.
org/​epic
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353. doi:10.1377/​hlthaff.2020.00103
ThinkWell. (2020). Immunization delivery cost catalogue. http://​www.immuni​zati​onec​onom​
ics.org/​ican
Vaughan, K., Ozaltin, A., Mallow, M., Moi, F., Wilkason, C., Stone, J., & Brenzel, L. (2019). The
costs of delivering vaccines in low-​and middle-​income countries: Findings from a system-
atic review. Vaccine: X, 2, 100034. doi:10.1016/​j.jvacx.2019.100034
2.3
Designing a primary costing study
or analysis
Ijeoma Edoka, Stephen Resch, and Logan Brenzel

This chapter presents and highlights critical elements related to the design,
purpose, scope, and perspective of the study. In addition, approaches for data
collection, and sampling are addressed.

2.3.1 Design considerations

When planning and designing an immunization costing study, the purpose

and use of the results and by whom should be identified as a first step. This
may entail elaborating the main question(s) the study will address. This is
an important step as this will guide the subsequent design and approach for
the costing study. In discussion with policymakers and other relevant stake-
holders, it will be useful to specify the immunization service(s) to be costed,
and to understand how the results of the study are going to be used by the gov-
ernment or funder. This will help to set the expectations for the exercise.

2.3.1.1 Scope

Defining the scope of the costing study is important for identifying the types
of resources to be counted and included in the study which allows the analyst
to set boundaries for the costing exercise. The scope of the study will depend
on the purpose of the analysis, the type of program being costed, the study
perspective adopted, and the type of costs being estimated.
As mentioned previously, the inputs to be evaluated will depend upon which
aspect and delivery strategy of the immunization program will be the focus of

Ijeoma Edoka, Stephen Resch, and Logan Brenzel, Designing a primary costing study or analysis In: Handbook of Applied Health
Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0010
104 Designing a primary costing study or analysis

the study. Routine immunization programs are largely delivered within health
facilities but may sometimes be delivered through outreach campaigns within
the community. Supplementary immunization activities are often conducted
to augment coverage of routine vaccines as well as in response to infectious
disease outbreaks. Both program types have different resource use implica-
tions and, consequently, program delivery costs. For example, cost inputs
such as transportation and fuel cost or per diems and travel allowances will
not be included in a facility-​based routine immunization program but would
be counted if supervision is occurring or an outreach program is being costed.
On the other hand, building costs and overheads would be counted in a rou-
tine immunization program costing exercise but may not be a cost input in a
community-​based outreach program.

2.3.1.2 Perspective

The perspective of a costing study identifies who is bearing the costs.

Specifying the study perspective is important for setting the boundaries of
the costing exercise and determining the types of data to collect and analyze.
There are three main types of perspectives: provider, payer, and societal.
A healthcare provider perspective would evaluate the costs incurred di-
rectly by providers, such as those made by public or private hospitals or
frontline health facilities.
A payer perspective would analyze costs from the viewpoint of the agen-
cies that are currently financing services, or that might be financing them in
the future. This could include evaluating costs from the perspective of the
Ministry of Health, or that of development partners, or both.
The broadest perspective for costing is a societal perspective which includes
costs borne by providers and payers, as well as costs incurred by households,
such as transportation and waiting time costs. The societal perspective may
include resources utilized by other government entities, such as costs to the
Ministry of Education for human papillomavirus vaccination in schools.
There are some ambiguities when it comes to defining a costing perspective.
Most immunization costing studies reflect costs borne by the public health
sector (e.g., government or public payer). The provider perspective is also
sometimes ambiguous, as private providers are not always included in studies
in the same way that publicly funded providers are studied. The perspective
taken for a costing study should directly pertain to the objective of the study
and the requirements of the target audience. If the purpose of the study is to
 2.3.1 Design considerations 105

understand how much the government needs to plan and budget for immu-
nization services, the correct perspective would be the payer (public sector)
perspective. If donor support is expected to be taken up by a government after
a time period, then the perspective of donor and partner contributions also
need to be valued.
If the purpose of the study is to provide a cost estimate for a cost-​
effectiveness analysis, a societal perspective is the most appropriate. This
may require conducting a household survey or an exit interview with house-
holds to determine the direct costs incurred by patients seeking immuni-
zation services, as well as the indirect costs of foregone productivity due
to lost work time. The societal perspective broadens the scope of the cost
analysis and requires more data. The Second Panel on Cost-​effectiveness in
Health and Medicine and the Global Health Costing Consortium recom-
mend a societal perspective as well as the health sector perspective (i.e. a
hybrid perspective of the individual patient, payer, and provider) (Sanders
et al., 2016; Vassall et al., 2017).
Finally, if the interest is to be able to compare and contrast how much
it costs for different types of facilities to provide services, and to ascertain
whether any efficiencies could be obtained, then the provider perspective
would be useful.

2.3.1.3 Time horizon

Another dimension of scope is the study time horizon. Time horizon is an

important feature of a costing model since vaccination program costs vary
considerably based on the type of vaccine. For instance, consider that a tet-
anus vaccination may be given to newborns and again several times over their
lifespan, and an influenza vaccination may be an annual event. Vaccination to
protect against shingles may be taken later in life. Depending on the vaccine
program, the time horizon could be very different.
Costing studies are usually conducted retrospectively, such that the time
period pertaining to data collection is in the past, such as for the previous
year. This type of study provides an historical perspective and evaluates the
total and unit cost of what was achieved in the past. One could design a pro-
spective costing study that collects data as services are delivered. An example
of this would be a study that is coupled with a clinical trial and gathers data
related to resource use on a patient or provider questionnaire completed after
each visit, or which measures labor with time–​motion observation methods.
106 Designing a primary costing study or analysis

Cost modeling exercises that project the expected costs of future immu-
nization services consider a time horizon that is in the future. When making
projections for new program implementation, special consideration of the
time horizon is warranted, as there are often large one-​time costs at the start of
a program, and it often takes some time for a program to reach the flat part of
a learning curve and reach an efficient scale.

2.3.2 Data collection

This part of the chapter highlights useful data sources for costing studies, data
collection instruments and approaches, issues with data quality, as well as in-
corporation of shared costs and sampling procedures.

2.3.2.1 Data sources

Retrospective costing studies often rely heavily on administrative records,

such as expenditure data in cost accounting systems, claims filed by providers
to bill payers, purchase orders, payroll records or human resources personnel
records, activity logs, routine reports, inventory lists, and clinical records.
This administrative data is typically supplemented with data gathered via in-
terview or questionnaire from program managers and other key staff. A third
type of data collection is direct observation. Although there are limitations in
combining direct observation (which is necessarily prospective) with other
data collected retrospectively, it can be a useful approach to measuring certain
aspects of resource use—​especially the allocation of shared labor resources.
Guidance documents are available on how to best collect data through direct
observation, which can be done as a scientific approach (Resch et al., 2020).

2.3.2.1.1 Administrative records
Accounting systems are frequently used as a source of information on ex-
penditures such as utilities, fuel, per diems, expenses for training activities
such as venue rental, catering, and travel. In order to use information from
a cost accounting system, one must determine the extent to which financial
transactions can be mapped specifically to immunization inputs or activities.
Purchase order information can be a useful source for information on the price
paid for commodities, vehicles, and equipment. Payroll records are useful for
estimating average salaries for workers with different job positions and level of
 2.3.2 Data collection 107

experience. Activity logs such as a vehicle trip log can be useful for allocating
an appropriate portion of a health facility’s shared vehicle pool to immuniza-
tion activities. Program managers may have other logbooks with information
regarding the number of training, management, or supervision events, and de-
tails regarding the duration and number of participants at such events. Routine
reports will often be a source of information on the number of immunization
sessions held, number of outreach activities and campaigns, number of doses
used and discarded, and vaccine inventory stock levels. There may also be in-
ventory lists for cold chain equipment and vehicles indicating information
such as the make and model, location, condition, and age of capital items.
Another source of cost data is health insurance claims data and other ad-
ministrative records of healthcare providers (Riley, 2009). In the US, claims
data from Medicare and Medicaid and other insurers provide useful infor-
mation on payments made to providers for specific services. The Medical
Expenditure Panel Survey contains a wealth of data for tracking expenditures
for the most prevalent conditions (Aizcorbe et al., 2012). Records from the US
Veterans’ Health Administration and other hospital systems generally contain
information on the resource used to produce services. When using these data
sources, it is important to distinguish between payments made, billed charges,
and resource use.

2.3.2.1.2 Retrospective self-​report of program manager and staff

Some needed information may not exist in routine records. In this case, a
common solution is to conduct interviews with or administer surveys to pro-
gram managers or key staff members. Interviews or questionnaires might be
used for gathering information about the frequency or duration of outreach
sessions, training events, supervision visits, or the use of vehicles.
An important example is labor quantity, which represents a substantial
share of total immunization program cost. The share of nursing labor that is
dedicated to immunization is usually not routinely tracked. Questionnaires
can be used to gather information from a sample of healthcare workers and
administrative staff about their time spent on a range of immunization ac-
tivities. The proportion of time or absolute value of time spent by healthcare
workers can be identified through a series of questions.
Several issues arise when designing questionnaires for estimating labor
time allocation. First, is the choice of who will be interviewed. In some cases,
the senior nurse in charge of immunization at the health facility site can be
asked to estimate the number of hours (or percent of total hours) of each staff
member that was spent on immunization-​related activities. Probing questions
108 Designing a primary costing study or analysis

can be used to explore time allocation by activity category such as vaccine

administration or record keeping. A second issue is related to the period of re-
call for interviews. Respondents reporting labor allocation information may
refer to other records to aid with recall. A shorter recall period improves re-
liability of responses. A variant of this approach could involve interviewing
workers themselves about the allocation of their own time, but this would be
far more time-​consuming, and may not be much more accurate. Questions
to ask about labor time should ascertain usual working hours for each cadre
of staff, the usual hours for immunization services, and either the hours or
proportion of time a particular health worker or cadre of worker spends on a
particular activity during that day. In many countries, vaccines are given on
a single day per week in an immunization session, and this will greatly facili-
tate data collection on time allocation to immunization. In some studies, there
may be interest in allocating immunization time per staff member or cadre of
staff to specific activities, such as record keeping or collecting and distributing
vaccines. In these cases, additional questions will need to be asked.
One common issue with labor time allocation is deciding how to handle
downtime when staff are not busy due to demand-​side factors such as a lack
of patients, which can often be the case for rural, remote health facilities. It is
suggested that information on the proportion of downtime be collected. This
type of information can be used as additional explanatory factors in the anal-
ysis. Additional guidance on how to handle downtime can be found in the
reference guide by Resch et al. (2020).

2.3.2.1.3 Direct observation and other methods

Rather than collect subjective information on time allocation that may be
subject to recall bias, an alternative approach is through direct observation.
Time–​motion studies can be conducted where the vaccination session is ob-
served, and the time spent on various activities by individual staff is recorded
and analyzed. Activities like community outreach may only happen on a
single day per month. A direct observation approach should be designed to
accommodate the service delivery schedule. Observing multiple immuni-
zation sessions allows the researcher to improve the quality of the time al-
location estimates (Brenzel, Young, & Walker, 2015). Direct observation will
necessarily be prospective, and therefore not match the exact time period of
the other input data collected retrospectively for the study. An assumption
must be made that the share of labor time allocated to immunization activities
has not changed significantly between the study time period and the current
observation time period. Direct observation methods are time-​consuming
 2.3.2 Data collection 109

and more expensive to carry out. Health workers may change their workflow
if they become aware of observers.
Another approach for collecting labor time allocation is through a daily
diary of activities that records the time spent. Diaries should be kept over
a period of a month and may need to be reviewed for compliance in filling
them out and in making any course corrections in how they are being utilized.
Diaries may not be practical for facility samples where the sampled units are
not in proximity to one another.

2.3.2.1.4 Data collection instruments

Data on costs, outputs, and facility characteristics should be collected using
pretested, standardized questionnaire formats. Several tools exist for con-
structing data collection instruments. Sample questionnaires for immuniza-
tion costing studies can be found at https://​datave​rse.harv​ard.edu/​datave​rse/​
EPIC. Other costing methodologies can generate customized data collection
instruments tailored to specific study requirements (Pan American Health
Organization, 2019).

2.3.2.2 Addressing shared costs

Often, inputs into the delivery of immunization services will be shared with
other health services and interventions. For instance, a vehicle can be used to
transport a range of commodities, including vaccines. A multipurpose health
worker provides different services. If shared resources will be included in im-
munization cost estimates, the proportion of resources used by the program
will need to be estimated to correctly allocate those shared resources and their
respective costs to the program.
Determining what portion of a shared resource to allocate to immunization
may not be straightforward and usually requires making educated assump-
tions. Allocating shared resources requires developing an “allocation key” or
“tracing factor” that serves as the basis for allocation. These tracing factors
can also be used to allocate the cost of inputs within immunization to dif-
ferent program activities. Some examples of tracing factors include (1) al-
locating shared costs based on the proportion of immunization visits out of
total outpatient visits to the facility; (2) allocating the use of a vehicle based
on the share of mileage or distance traveled per week specific to immuniza-
tion as a proportion of total mileage; and (3) allocating building space based
on the proportion of square feet or meters. Additional sources of information
110 Designing a primary costing study or analysis

exist on determining the best allocation or tracing factors for immunization

programs (Brenzel, 2014; Resch et al., 2020).

2.3.2.3 Health system levels

The scope of a costing study can be framed by level of the health system. Some
immunization activities may occur at levels in the health system above the
point of service. Costs can be incurred at administrative levels (national,
regional, provincial, or district) as well as the specific facility level. Costing
studies can include “above facility costs” to determine where the balance of
resource use lies—​with delivery of services or with their management and
administration. Evaluating administrative costs is becoming more common
in the conduct of immunization costing studies (Clift, Arias, Chaitkin, &
O’Connell, 2017; Vassall et al., 2017).
To evaluate costs at different health system levels, it is important to map
specific activities to where the resources are being used. For example,
microplanning might take place at a district health office and resources used
for meetings and travel to and from that office should be mapped to that level.
Vaccine supply and distribution might be the responsibility of the district
health system level, while procurement of vaccines might be primarily a na-
tional level exercise. Information on inputs and prices used in a cost analysis
may be aggregated and maintained at particular levels of the health system,
such as information on coverage estimates or vaccine wastage rates.
Defining activities within each level of the health system is also useful
for identifying where resources may be shared between levels. For example,
the tender prices for a new vaccine may include not only procurement costs
(freight, insurance, customs duties) and storage cost but may also include
transportation costs to different administrative units of the health system,
such as from national to regional storage facilities. For this reason, it is im-
portant to know in advance the types of immunization activities that occur
at different levels of the health system. Data collection may require separate
questionnaires for different health system levels.

2.3.3 Sampling for immunization costing studies

Immunization costing studies are based, more and more, on sampling of fa-
cilities and delivery sites. Sampling is required because of the heterogeneity
 2.3.3 Sampling for immunization costing studies 111

of total and unit costs across geographical areas and between and within fa-
cility types (Resch et al., 2020). In addition, inputs into unit costing are rarely
known in their entirety. For most resource inputs to vaccine programs, data
will have to be gathered from a sample of units of the immunization system,
and the total resource use will need to be extrapolated from that sample.
The extent of sampling will be determined by the budget and time available
to do the study, and whether representativeness is required and at what level
(whether national or subnational).
Care should be taken in the sample design to ensure it is representative and
sufficiently large to generate estimates with acceptable precision. Random
sampling is preferred to other approaches when the goal is to have nationally
or sub-​nationally representative estimates. Random sampling can also facili-
tate comparisons of cost between different locations, types of sites, or service
delivery approaches. An additional benefit of random sampling is the ability
to derive unbiased estimates of sample mean cost estimates, and to calculate
confidence intervals around that estimate.
An alternative is to select units deliberately based on some features of those
units, or select units based on convenience, such as nearby or easy-​to-​reach
facilities (purposive or convenience sampling). Purposive sampling can help
to reduce the cost of the exercise and may also provide more control over con-
text and variables of interest for the study. For instance, if the study aims to
evaluate the costs of accessing hard-​to-​reach populations in rural, remote
areas, a purposive sample of facilities located in those areas will provide more
direct evidence on costs. However, the results may be biased and it will be dif-
ficult to estimate precision for these types of studies.
Standard methods for sample size determination can be used for immuni-
zation costing studies (Leslie, Laos, Cárcamo, Pérez-​Cuevas, & García, 2021;
Vassall et al., 2017). For more complex sample designs, a statistician should
be consulted. Increasing the number of units selected in a costing study
sample will improve the precision of the results, but this also becomes more
costly. Balancing precision and costs through alternative sampling frames can
be achieved with the Sample Design Optimizer (ImmunizationEconomics.
Org, 2020).
Ultimately, the sample design for an immunization costing study will de-
pend on the objective of the study and the information available prior to data
collection. The most critical piece of information is a complete list of all sites
by type. For example, if the study is going to collect information at the central/​
national level, as well as from a sample of districts and a sample of facilities
within the sampled districts, it will be important to have a list of districts and
112 Designing a primary costing study or analysis

of health facilities within those districts. This list is the sampling frame. It is
this population of sites for which we can infer costs based on statistical anal-
ysis of the data collected from the sampled sites.
Other factors can be used to stratify the sample. For instance, the type of
health facility (private or public hospital, clinic, health post), the location of
the facility (urban or rural), or the volume of services provided in that fa-
cility (high, average, or low volumes) can be used to stratify units for selec-
tion. Stratified random sampling is more and more common in immunization
costing studies since services are organized along a hierarchy of administra-
tive levels from region to district to health facility. However, random sampling
may lead to a more expensive data collection process, as facilities and health
system sites may be more widely spread out geographically. There is a poten-
tial loss of precision associated with stratification.

2.3.3.1 Illustrative sampling procedure

Typically, for immunization costing studies attempting to sample a suffi-

cient number of sites at different organizational levels, a clustered design is
recommended, in order to reduce data collection cost. Below is an outline of
a possible sampling procedure oriented toward a study of the cost of routine
facility-​based immunization delivery:

• Geographical areas are selected. If the country is small, the geographical

area may be the entire country. If the country is large, provinces may be
selected at random or purposively. For example, if the goal is to compare
more developed provinces to less developed ones, at least one of each type
should be included. This could be accomplished by purposive selection or
by random selection from strata.
• Subnational units are selected below the province level. Districts or other
subnational units can be randomly selected where the probability of their
selection is proportional to the number of health facilities. This is done so
that there is equal representation in the sample between small or larger
districts.
• Facility sampling is based on a complete list of health facilities that are
relevant for the scope of the study. For instance, a study evaluating gov-
ernment facilities should be based on a complete list of those facilities.
Incorporating private or nongovernmental organization facilities will re-
quire complete lists of those types of facilities. If the scope of the study
 2.3.3 Sampling for immunization costing studies 113

considers only certain facilities that have immunization service available,

then the list would be restricted to these immunization-​relevant sites.
• If a stratified sample of facilities will be selected, additional information
on the following is useful to collect for each district and facility in the
sample:
• The number of vaccine doses administered or the number of outpa-
tient visits in the past year.
• For each district, information on population density should be obtained.
• Facilities could be classified as urban (urban/​peri-​urban) and rural/​re-
mote and also by ownership (government, nongovernment, private).
• The final stage of the sampling procedure is to randomly select the same
number of facilities from each selected district. Healthcare costs, which
are constrained to be non-​negative, are generally right skewed rather
than normally distributed (ImmunizationEconomics.Org, 2020;
Turner, Angeles, Tsui, Wilkinson, & Magnani, 2001). Because costs
tend to be right tailed (a large number of sites have low costs, with a
few sites having much higher costs), oversampling those sites that
may have lower costs may be desirable to more accurately represent
the cost distribution in a country. These sites tend to be low-​volume,
rural, or remote sites rather than high-​volume, urban, or peri-​urban fa-
cilities. (ImmunizationEconomics.Org, 2020; Turner et al., 2001) For
this reason, simple random sampling is recommended in order not to
overly favor those facilities that administer a large number of doses.
Oversampling of low-​volume, rural, or remote facilities could be ac-
complished by stratifying on this basis and selecting a relatively larger
sample from that stratum.
• Should a facility be unavailable for survey, such as for security/​safety
reasons or respondents being absent or preoccupied during the survey
period, that facility can be replaced in the sample through random sam-
pling. It might be useful to select the replacement facilities in advance to
use as substitutes if needed while collecting data.

Each sampled unit may have a different probability of being selected into the
sample. As a result, each unit in the sample may represent a different-​sized por-
tion of the sampling frame, which must be accounted for in data analysis and
when generalizing from the sample to the larger population. Incorporating
sample weights into the analysis (inverse of the probability of a selection) will
be important. The sampling procedure will determine sample weights used in
the data analysis.
114 Designing a primary costing study or analysis

References
Aizcorbe, A., Liebman, E., Pack, S., Cutler, D. M., Chernew, M. E., & Rosen, A. B. (2012).
Measuring health care costs of individuals with employer-​sponsored health insurance in the
U.S.: A comparison of survey and claims data. Statistical Journal of the IAOS, 28(1–​2), 43–​51.
doi:10.3233/​SJI-​2012-​0743
Brenzel, L. (2014). Working Paper: Common approach for the costing and financing analyses
of routine immunization and new vaccine introduction costs (EPIC). Mimeograph. Bill &
Melinda Gates Foundation. www.immuni​zati​onec​onom​ics.org/​epic-​info.
Brenzel, L., Young, D., & Walker, D. G. (2015). Costs and financing of routine immuniza-
tion: Approach and selected findings of a multi-​country study (EPIC). Vaccine, 33, A13–​
A20. doi:10.1016/​j.vaccine.2014.12.066
Clift, J., Arias, D., Chaitkin, M., & O’Connell, M. (2017). Landscape study: The cost, impact, and
efficiency of above service delivery activities in HIV and other global health programs. Results
for Development. https://​r4d.org/​resour​ces/​landsc​ape-​study-​cost-​imp​act-​eff​i cie​ncy-​serv​ice-​
deliv​ery-​act​ivit​ies-​hiv-​glo​bal-​hea​lth-​progr​ams
ImmunizationEconomics.Org. (2020). Sample design optimizer. Immunization­Economics.org.
http://​immuni​zati​onec​onom​ics.org/​sam​ple-​des​ign-​optimi​zer
Leslie, H. H., Laos, D., Cárcamo, C., Pérez-​Cuevas, R., & García, P. J. (2021). Health care pro-
vider time in public primary care facilities in Lima, Peru: A cross-​sectional time motion
study. BMC Health Services Research, 21(1), 123. doi:10.1186/​s12913-​021-​06117-​9
Pan American Health Organization. (2019). COSTVAC. http://​immuni​zati​onec​onom​ics.org/​
rec​ent-​activ​ity/​2019/​9/​23/​upd​ate-​of-​pahos-​pro​vac-​e-​tool​kit
Resch, S. C., Menzies, N. A., Portnoy, A., Clarke-​Deelder, E., O’Keefe, L., Suharlim, C., &
Brenzel, L. (2020). How to cost immunization programs: A practical guide for primary data
collection and analysis. http://​www.immuni​zati​onec​onom​ics.org/​epic
Riley, G. F. (2009). Administrative and claims records as sources of health care cost data.
Medical Care, 47(7 Suppl 1), S51–​S55. doi:10.1097/​MLR.0b013e31819c95aa
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M., . . . Ganiats, T.
G. (2016). Recommendations for conduct, methodological practices, and reporting of cost-​
effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and Medicine. JAMA,
316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
Turner, A. G., Angeles, G., Tsui, A. O., Wilkinson, M., & Magnani, R. (2001). Sampling manual
for facility surveys for population, maternal health, child health and STD programs in devel-
oping countries. MEASURE Evaluation. https://​www.measur​eeva​luat​ion.org/​resour​ces/​
publi​cati​ons/​ms-​01-​03
Vassall, A., Sweeney, S., Kahn, J., Gomez, G. B., Bollinger, L., Marseille, E., . . . Levin, C. (2017).
Reference case for estimating the costs of global health services and interventions. LSHTM
Research Online. https://​res​earc​honl​ine.lshtm.ac.uk/​id/​epr​int/​4653​001/​1/​vassall_​etal_​
2018_​reference_​case_​for_​estimating_​co ​ sts​_g​ lo​bal_​heal​th_​s​ervi​ces.pdf
2.4
Data analysis
Stephen Resch and Logan Brenzel

2.4.1 Introduction

Data analysis is the process of preparing the collected data, performing

calculations, and applying statistical methods to produce results that an-
swer research questions. A data analysis plan should typically be developed
prior to data collection, so that one can be assured that the data collection
includes everything needed to carry out the analysis. The data analysis pro-
cess will typically start with a data cleaning step. Next, there is a standard
set of adjustments applied to account for currency conversion and inflation,
discounting streams of cost that occur over time—​annualization of capital
investments. Another analysis step pertains to allocating shared resources
to activities. This allocation can be “built in” to the design of the data col-
lection process, but is also sometimes handled in the analysis phase, after
data is collected.
When data have been collected from a representative sample in which
randomization was used to select sampled units, the data analysis step will
include statistical methods to make inferences about the population and es-
timate the precision of the results. In some costing studies, classic methods
of statistical inference from a representative sample may not be feasible due
to lack of randomization in the sample design, or in situations where there is
a policy need to extrapolate beyond the sample frame. In these cases, other
methods of extrapolation may be utilized. In these situations, sensitivity anal-
ysis explores the uncertainty of cost estimates given the number of assump-
tions used in the analysis, and because it is not possible to estimate traditional
confidence intervals.

Stephen Resch and Logan Brenzel, Data analysis In: Handbook of Applied Health Economics in Vaccines. Edited by:
David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0011
116 Data analysis

2.4.2 Discounting

In primary costing studies, it is rare to need to use discounting, as the time

horizon is usually short and the study retrospective in nature. However, in
costing exercises to model projected costs of some future immunization ac-
tivity, discounting may be needed if a longer time horizon than 1–​2 years is
being considered. Discounting is a way to account for time preference and
risk aversion. There is a large literature on economic theory and empirical ev-
idence regarding discounting as it applies to cost accounting, economic eval-
uation, and social policy. For most economic evaluations in the health sector,
guidelines recommend a discount rate of 3–​5%, though some have argued for
higher discount rates in low-​income settings. Thus, 3% would be a common
discount rate applied to high-​income countries, as recommend by the US
Panel on Cost-​Effectiveness in Health and Medicine, whereas 5% may be a
more appropriate rate specifically for low-​and middle-​income countries. The
purpose of discounting is to convert a stream of future costs over time into
their present value.
Equation 2.4.1. Discounting:

Future value
Present value =
(1 + r )t

where r is the annual discount rate and t represents the number of years when
the future value occurs.

2.4.3 Annualization

For program inputs that last more than 1 year, such as vehicles and cold chain
equipment, the annual value of that input should be proportional to the frac-
tion of the item’s expected useful life and the opportunity cost of capital being
tied up in the input and unavailable for other uses.
In an economic analysis, annualization of capital cost “smooths out” the
actual stream of fiscal outlays in order to provide a better picture of the long-​
run average cost of an immunization program. This is important if one is
considering a short time horizon such as 1 year. It would not be appropriate
to allocate the total cost of the capital inputs only to the doses delivered in
that year when those capital inputs are expected to be in use for several years
 2.4.3 Annualization 117

and contribute to the delivery of many more doses over time. By annualizing
the cost of capital resources, we can assign a fair proportion of the cost to
the program output of a given time period and avoid idiosyncratic variations
in cost estimates that might occur simply due to the timing of large capital
investments.
The useful life of an asset is the period of time during which it is expected
to be fully employed for its original purpose. A general rule of thumb
is that useful life is equivalent to the number of years until the cost of
maintaining and repairing a piece of equipment (opportunity cost of
using an outdated model/​make) outweighs the cost of buying a new piece
of equipment. It is not uncommon to find items that have exceeded their
useful life but are still in operation. However, one should not base useful
life on these examples. Nor should one base useful life on the current age
of an item in the inventory. Many countries have standard benchmark
values for useful life of capital items. Benchmarks for useful life for immu-
nization programs have been developed (Resch et al., 2020; World Health
Organization, 2021).
The calculation of an annualized cost requires information about the
purchase price (or replacement value), an annualization factor which is a
function of the useful life, and a discount rate. The annualization factor
is sometimes referred to as the “present worth of annuity” factor. A table
of these factors can be found online (Resch et al., 2020) or in textbooks
(Drummond, Sculpher, Claxton, Stoddart, & Torrance, 2015).
For financial cost analysis, the annualized cost doesn’t factor in the
opportunity cost of capital. In this case, the purchase price (or replace-
ment value) is divided only by the number of years of useful life. This
will generally give a similar result to annualization, unless useful life
is very long. The example in Box 2.4.1 shows how annualized costs are
estimated.
Equation 2.4.2. Annualization factor:

Annualization factor =
(1 + r )n − 1
(r × (1 + r )n )
where r is the discount rate (use 3% unless you have justification for using an
alternative rate) and n is the number of years of useful life. This may or may
not correspond with an item’s actual physical or economic life.
118 Data analysis

Box 2.4.1 Example of using an annualization factor

A freezer costs $20,000 to purchase and has an estimated useful life of

10 years. Using a discount rate of 3%, the annual cost of the freezer would be $2,344.61
each year. Stated differently, a stream of payments of $2,344 per year for 10 years has
a present value of $20,000.

Cost estimate: $20,000.

Discount rate: 3%.
Years of useful life: 10.
Annualization factor =​[1 − (1 +​0.03)​10]/​0.03 =​ 8.5302
Annualized cost =​$20,000/​8.5302 =​$2,344.61
Excel formula: =​ -​PMT (0.03, 10,$20000) =​$2,344.61
Note: the PMT formula command in Microsoft Excel can be used to estimate annualized value of
cost. It is important to include a minus sign ahead of that command, because Excel assumes you
are using PMT to get the value of an annual payment that will be equivalent to the present value of
some lump sum (such as a loan amount).

2.4.4 Evaluating costs

Once data on inputs, unit prices, and allocation factors have been collected,
the analyst can assemble the estimates of total immunization costs per sam-
pled unit. This will generally follow a standard formula:
Equation 2.4.3. Cost of input:

Cost of input = Quantity of input (by type) × Unit price of input

× Allocation factor

Costs of all inputs are summed up for each facility to estimate total im-
munization cost at the facility. These costs can be divided into capital and
recurrent line items or by specific immunization activity. Representation
of costs may be best illustrated through pie charts or bar graphs. The share
of total immunization cost by line item or activity is referred to as the cost
profile. Cost profiles can be compared across the sample by facility type or
location. Most cost studies evaluate total or full immunization program or
strategy costs, as well as unit costs (total costs divided by doses adminis-
tered or total costs divided by the number of children or target population
vaccinated).
 2.4.5 Evaluating costs from a sample of facility sites 119

2.4.5 Evaluating costs from a sample of facility sites

Facility-​based cost estimates are often aggregated to reflect the total cost of the
immunization program based on the sample. In doing so, it is essential to ac-
count for underlying relationships between cost and volume and to minimize
bias and maximize precision of the estimates.
As mentioned previously, total immunization costs are not normally distrib-
uted across a sample of health facilities, but are right tailed, with lots of facilities
providing a lower volume of services at lower costs. Taking the simple average
of total immunization facility costs across the sample or within the strata of the
sample will not likely represent the true value of costs, although simple aver-
ages can be used for benchmarking and for identifying and describing facilities
with lower unit costs and higher volumes (higher performing, more efficient).
A recent systematic review of estimation techniques in multisite costing
studies found the majority of them (52%) used a volume-​weighted mean, and
the remaining studies relied on simple means or medians of total and unit
costs. Each of these estimators are associated with different levels of bias and
precision. A simple mean cost calculation could be upwardly biased by 12% to
over 100% (Clarke-​Deelder, Vassall, & Menzies, 2019).
A better approach evaluates the volume-​weighted mean of costs across the
sample of sites. The volume-​weighted mean unit cost is derived from the sum
of the total costs across the sample of sites divided by the sum of doses de-
livered across the sample of sites. To estimate the volume-​weighted total cost
of the national immunization program, the mean of total costs for the sample
sites is multiplied by the ratio of the total delivery volume in the overall pro-
gram to the mean delivery volume in the sample. This approach requires ad-
ditional information about the total delivery volume in the population. While
slightly biased, this approach significantly improves precision (World Health
Organization, 2021).
The calibration estimator uses additional information to reweight the data
in the sample to match the true distribution of costs more closely in the pop-
ulation. At a minimum, auxiliary information will include the total volume
of services delivered in the overall program of interest and the total number
of sites in the overall program of interest. Information about other variables
that drive costs can be incorporated to further improve precision. The calibra-
tion estimator has improved precision relative to the volume-​weighted mean
(and has a similar upward bias in small samples). However, estimation is more
complex and requires the use of more advanced software than the volume-​
weighted mean (Rivera-​Rodriguez, Resch, & Haneuse, 2018).
120 Data analysis

A regression approach improves precision by incorporating additional in-

formation, such as information related to distance between health facilities,
quality of care, and other factors. One simple example is a log–​log regression
of costs on delivery volumes: log(Ci) =​ Qi +​ i where C is the cost and Q is the
quantity of doses. From this model, unit costs are the sum of the predicted
costs divided by the number of doses delivered in the population.

2.4.6 Sample weights

As mentioned previously, sampling weights should be incorporated into es-

timates of total costs. The sample frame will determine the weights used.
Weights are the inverse probability of selection. With stratification, there
are multiple probabilities of being selected and these probabilities should be
multiplied by each other, with weights being the inverse of these joint prob-
abilities. For example, if one district is selected out of five (1/​5), and five fa-
cilities are selected out of 20 (5/​20), then the probability of facilities being
selected in that district in the sample is (1/​5 × 5/​20 =​1/​50). The sample weight
for a facility in that district would be the inverse, or 50.

2.4.7 Aggregating cost data

Researchers may be interested in aggregating total and unit cost data from one
health administrative level to the next. The methods described above would be
appropriate for calculating summary estimates of unit and total costs at each
level of the healthcare system before aggregation. Appropriate sample weights
and methods for estimation of standard errors should be used at each level of
the analysis. For example, in order to estimate total costs across three levels of
the health system, it is recommended that the methods previously described
are utilized for each level, and the total immunization cost or unit costs for
each level are combined to result in the national level costs. One caution is that
if vaccine and syringe costs are estimated for facility and subnational levels,
it is recommended to remove that aspect of costs in the aggregation process,
and to add in the total value of vaccines and syringes obtained from the na-
tional level.
Where more complicated sampling and/​or analysis methods are chosen,
it is recommended to have these methods reviewed by a survey statistician
or other individual with expertise in the application of these techniques.
Additional guidance is available (Drummond et al., 2015).
 2.4.8 Evaluating uncertainty in cost analysis 121

2.4.8 Evaluating uncertainty in cost analysis

Cost analysis of immunization programs is undertaken to inform decision-​

making around resource allocation and use. A researcher, policymaker, or
funder who understands both the heterogeneity and level of uncertainty in
the cost results can make more informed decisions. In addition, decisions can
be taken to mitigate the level of uncertainty.
While care is taken to generate the most robust and accurate estimates
of total and unit costs of immunization services, all immunization costing
studies contain some level of uncertainty in the estimates. Uncertainty means
there is a range of values that cost estimates can take and the true estimate is
somewhere in that range, and that there is some deviation in accuracy and
precision of the estimate. Some uncertainty is derived from the study sam-
pling process. An example is when the study sample of units (facilities) is
not representative of the entire population of units. Cost estimates will have
sample error and a higher degree of uncertainty. Randomization decreases
sample errors; increasing the sample size will improve precision.
Sampling doesn’t represent the full picture of uncertainty in cost estimates.
Many inputs into the cost calculations are based on subjective assumptions
which may lead to greater errors and level of uncertainty. We often don’t di-
rectly observe these inputs and need to make assumptions. Measurement
error may also stem from poor record keeping and recall bias. For instance,
records on number of doses of vaccines administered may be sketchy, missing,
or have reporting errors. Measurement of labor time allocation is often done
by asking individuals how much time they spend on immunization activities.
Time allocation data may be a major substantial source of measurement error
in cost studies. Researchers are encouraged to verify and validate those esti-
mates during data collection and analysis.
Uncertainty of immunization cost estimates should be estimated and re-
ported, but this is often not undertaken as it is an additional analytical step
in immunization costing studies. This is an area of methodological develop-
ment that needs to be addressed. Usually, policymakers and funding agencies
do not request this type of analysis so it has been conveniently avoided. Yet,
given that point estimates of total or unit costs may drive policy and program-
matic decisions, it is important to present the level of uncertainty to improve
understanding of the estimates and quality of these decisions. The Second
Panel on Cost-​Effectiveness in Health Care and Medicine, the International
Decision Support Initiative reference case, as well as the Global Health
Costing Consortium recommend that uncertainty be appropriately charac-
terized (Sanders et al., 2016; Vassall et al., 2017).
122 Data analysis

There are several ways that uncertainty in cost estimates can be evaluated:

1. Calculating a confidence interval is one way to represent the level of un-

certainty of an immunization cost estimate. A 95% confidence interval
means that we are 95% confident that the true value is found within that
range of estimates.
2. Identify and document major assumptions and potential sources of
measurement error in the cost estimation and conduct a one-​way sensi-
tivity analysis, testing each of those assumptions and sources. This might
include re-​estimating the number of doses delivered using an alternative
assumption, utilizing other options for tracing factors to allocate shared
costs, and examining the impact of different prices of inputs, among
others. A table or tornado diagram (Fig. 2.4.1) that reports the original
estimate and the new cost estimates with each change, and the percent
difference in the value, highlights how easily the baseline estimates are
influenced by assumptions.
3. Other statistical techniques, such as a Monte Carlo analysis, scenario
analysis, and threshold analysis can be used for economic evaluation in-
cluding costs. Please see Chapter 4.2 for further details.

Labor cost

Vaccine price

Training cost

Target population

Vaccine wastage

Transport mileage

Useful life

$(500) $(400) $(300) $(200) $(100) $- $100 $200 $300 $400 $500

Low vs base High vs base

Fig. 2.4.1 Sample Tornado diagram evaluating uncertainty in immunization

program costs.
 2.4.9 Conclusion 123

2.4.9 Conclusion

This chapter describes various techniques related to immunization program

delivery cost study design, data collection, and analysis. Discounting and
annualization are important for evaluating the costs of capital assets that have
a long useful life. Using the sampling weights to evaluate total and unit costs
is another recommended approach. Finally, all cost studies should evaluate
sources of uncertainty in the baseline estimates.

References
Clarke-​Deelder, E., Vassall, A., & Menzies, N. A. (2019). Estimators used in multisite healthcare
costing studies in low-​and middle-​income countries: A systematic review and simulation
study. Value in Health, 22(10), 1146–​1153. doi:10.1016/​j.jval.2019.05.007
Drummond, M., Sculpher, M., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015).
Methods for the economic evaluation of health care programmes (4th ed). Oxford: Oxford
University Press.
Resch, S. C., Menzies, N. A., Portnoy, A., Clarke-​Deelder, E., O’Keefe, L., Suharlim, C., &
Brenzel, L. (2020). How to cost immunization programs: A practical guide for primary data
collection and analysis. ImmunizationEconomics.org. http://​www.immuni​zati​onec​onom​ics.
org/​epic
Rivera-​Rodriguez, C. L., Resch, S., & Haneuse, S. (2018). Quantifying and reducing statistical
uncertainty in sample-​based health program costing studies in low-​and middle-​income
countries. SAGE Open Medicine, 6, 2050312118765602. doi:10.1177/​2050312118765602
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M., . . . Ganiats, T.
G. (2016). Recommendations for conduct, methodological practices, and reporting of cost-​
effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and Medicine. JAMA,
316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
Vassall, A., Sweeney, S., Kahn, J., Gomez, G. B., Bollinger, L., Marseille, E., . . . Levin, C. (2017).
Reference case for estimating the costs of global health services and interventions. LSHTM
Research Online. https://​res​earc​honl​ine.lshtm.ac.uk/​id/​epr​int/​4653​001/​1/​vassall_​etal_​
2018_​reference_​case_​for_​estimating_​co​ sts​_g​ lo​bal_​heal​th_​s​ervi​ces.pdf
World Health Organization. (2021). CHOosing Interventions that are Cost-​Effective (CHOICE).
https://​www.who.int/​news-​room/​feat​ure-​stor​ies/​det​ail/​new-​cost-​effect​iven​ess-​upda​tes-​
from-​who-​cho​ice
2.5
Costing new vaccine introduction
Susmita Chatterjee, Siriporn Pooripussarakul, and Logan Brenzel

2.5.1 Introduction

Since 2000, lower-​income countries have been able to introduce new and
underused vaccines at subsidized prices through Gavi, the Vaccine Alliance.
More than 70 Gavi-​eligible countries have introduced pentavalent and inject-
able polio vaccines. Gavi supported 60 vaccine introductions and campaigns in
2019, and the coronavirus disease 2019 (COVID-​19) Vaccines Global Access
(COVAX) initiative is supporting introduction of COVID-​19 vaccines in over
100 countries (Gavi, 2021). Country decision-​making around whether to in-
troduce a new vaccine or not depends upon a range of factors. The introduc-
tion of new vaccines depends on a range of criteria, including clinical efficacy,
safety, feasibility, policy and regulatory and economic considerations, among
others (Pooripussarakul, Riewpaiboon, Bishai, Tantivess, & Muangchana,
2018). When introducing a new vaccine, there are major concerns such as the
magnitude of disease burden, vaccine safety and effectiveness, and the viability
and sustainability of the immunization program, among other factors (Loze
et al., 2017). New and underutilized vaccine introduction (NUVI) requires sub-
stantial investment, not only in terms of vaccines and supplies but also health
systems and other infrastructure inputs such as cold chain, training, and advo-
cacy. This chapter presents methodological considerations for costing NUVI.
The costs of NUVI can be higher per dose than those for routine immuni-
zation services (Griffiths et al., 2016; Le Gargasson, Nyonator, Adibo, Gessner,
& Colombini, 2015; Levin, Wang, Levin, Tsu, & Hutubessy, 2014; ThinkWell,
2020; World Health Organization: Immunization Vaccines and Biologicals,
2002). This is largely because the highest fixed costs associated with intro-
duction are spread over fewer children being reached in the early phases. The
main drivers of NUVI costs tend to be training, social mobilization, and ac-
tual service delivery.

Susmita Chatterjee, Siriporn Pooripussarakul, and Logan Brenzel, Costing new vaccine introduction In:
Handbook of Applied Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula,
Oxford University Press. © Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0012
 2.5.2 Overall approach and rationale for NUVI costing 125

$0 $5 $10 $15 $20

Facility-based Vietnam

Routine monthly India

Integrated outreach Uganda

Pulse campaign India

School-based Peru

School-based Vietnam

School-based Uganda

School-based Tanzania

Recurrent Start-up

Fig. 2.5.1 Cost of delivering HPV vaccine per fully immunized girl.
Source: Levin et al. (2014).

Fig. 2.5.1 shows the variation in start-​up and recurrent costs for introduc-
tion of human papillomavirus vaccine in eight countries, depending upon the
strategies used. Start-​up costs were generally higher than recurring costs, and
school-​based delivery was also larger per fully immunized girl than facility-​
based delivery on the whole (Griffiths et al., 2016; Le Gargasson et al., 2015;
Levin et al., 2014; ThinkWell, 2020).

2.5.2 Overall approach and rationale

for NUVI costing

Costing of NUVI is different from costing the routine immunization pro-

gram. The main difference is that unlike in routine immunization costing
which examines the total cost of the program, in NUVI costing, the focus is
on incremental inputs and costs above and beyond the routine baseline costs.
This approach has implications for the types of inputs evaluated and how they
are measured.
The incremental cost of NUVI may be measured as a fiscal cost, a financial
cost, or an economic cost, depending on the objective of the cost estimation.
If the objective is to plan for the budget of the program, a fiscal cost analysis is
appropriate as it measures the additional expenditures required on goods and
services related to NUVI.
126 Costing new vaccine introduction

Economic costs are the opportunity costs, defined as resources that have
been forgone for alternative investments and uses. A NUVI may require ad-
ditional injections and time spent delivering a vaccine that could be used to
provide other productive interventions. If the purpose is to evaluate the cost-​
effectiveness of a new vaccine, then the focus should be on incremental eco-
nomic costs of vaccine introduction. If the result of the costing exercise is to
incorporate the findings into a sustainability plan or ascertaining budget im-
pact, the focus needs to be on evaluating incremental financial costs.
The World Health Organization generated a guideline for estimating NUVI
costs into the existing program (World Health Organization: Immunization
Vaccines and Biologicals, 2002). Other guidelines provide a step-​by-​step
overview of how to estimate NUVI costs (Brenzel, 2014).
Costing of NUVI poses several additional requirements on new immuni-
zation programs (Pooripussarakul et al., 2018). The first step is to identify the
quantities and prices of the additional inputs above and beyond the routine
program required for NUVI and delivery. The type of new vaccine introduced
will determine storage volumes, cold chain requirements, types of syringes
to be used, and additional time allocated for various activities such as ad-
ministration of the vaccine or record keeping. Depending upon the delivery
strategy used, there may other inputs to include. Table 2.5.1 shows the various

Table 2.5.1 Costs of NUVI related to type of vaccine

New vaccine formulation Likely additional resources required

Combination vaccines Vaccines, syringes

Additional labor time for administration, record keeping,
and other activities
Cold storage and distribution, including vehicles
Social mobilization and advocacy
Training for health workers
Vaccination cards and reporting systems
Disease surveillance
Monovalent vaccines Vaccines, syringes
Additional labor time for administration, record keeping,
and other activities
Supplies (reconstituting syringes, alcohol wipes, etc.)
Waste management
Additional time costs
Cold storage and distribution
Social mobilization and advocacy
Training for health workers
Vaccination cards and reporting systems
Disease surveillance
 2.5.3 NUVI cost estimation 127

additional inputs that may be required for different types of vaccines. Adding
a monovalent vaccine may entail additional inputs such as syringes, training,
reprinting of vaccination cards, advocacy and social mobilization, disease sur-
veillance, and additional cold storage and waste management supplies. If the
vaccine needs to be reconstituted, additional storage space and syringes are
required. Combination vaccines, such as the pentavalent vaccine, may even-
tually save on cold storage space if replacing more than one vaccine. A com-
bination vaccine that doesn’t require additional syringes or supplies may not
pose additional costs for the cold chain, inventory management, and distri-
bution system. On the other hand, a monovalent vaccine or a combination
vaccine with a different dosing schedule will require additional inputs into the
existing immunization program (World Health Organization: Immunization
Vaccines and Biologicals, 2002).
NUVI may require additional cold chain management, for example, the
replacement of ten-​dose vials of whole-​cell diphtheria–​tetanus–​pertussis
(DTwP) vaccine with single-​dose vials of pentavalent (DTwP–​hepatitis
B–​Haemophilus influenzae type b) vaccine. If the transportation of the new
vaccine exceeds the capacity of the existing transportation system, the inclu-
sion of additional vehicles or shorter supply intervals may be required. The
increase in vaccine and supply chain volume also requires additional storage
and cold chain management. These result in increasing financial investment
to accommodate the larger volume of vaccines in the distribution system
(Mvundura et al., 2014).
Costing of NUVI can be focused retrospectively based on historical infor-
mation or can be conducted as a projection of future costs. Cost projections
will require making assumptions based on expert interview and anticipated
changes in the routine program.

2.5.3 NUVI cost estimation

NUVI costs should be divided into initial or start-​up costs (needed to set
up the program) and ongoing costs (needed to keep the program going).
Investments related to NUVI can start long before the actual introduction
of the vaccine and may continue after the introduction. Therefore, the time
period for considering costs related to NUVI needs to be carefully demar-
cated to set the boundaries for NUVI costing, for example, when the intro-
duction period begins, for start-​up cost estimation, and when it ends, so costs
are mapped into recurrent costs. In large countries where the NUVI will be
128 Costing new vaccine introduction

in a phased manner, the approach to classify investment costs may require

modification.

2.5.3.1 Initial or start-​up costs

Investments that occur during the initial phase of the introduction or are one-​
off costs in the first year of the introduction are referred to as initial or start-​up
costs (Hidle et al., 2018; Ngabo et al., 2015). Initial costs of a new vaccine will
focus on the additional activities and resources that occur when a new vaccine
has been introduced, for example, microplanning, additional procurement
of supplies, additional cold chain and logistics, additional transportation and
per diem related to meetings and supervision, disease surveillance, develop-
ment of new training materials, advocacy, and social mobilization materials
(Hidle et al., 2018; Le Gargasson et al., 2015). Cost data include additional
costs of capital, labor, and material of those activities. For example, the cold
chain of a new vaccine may require extra cold containers and other equip-
ment, such as refrigerators, computers, and printers. Logistic costs may in-
crease in purchasing, storage and inventory management, and transportation
of a new vaccine. A buffer stock, which is usually set at an additional 25%
of doses, should be included in the estimation of initial vaccine costs (World
Health Organization: Immunization Vaccines and Biologicals, 2002).
The initial costs are not only from new vaccine procurement, but also the
costs of expanding the cold chain as well as increasing distribution costs, staff
training costs, and expanding social mobilization and surveillance systems.
UNICEF estimates that the introduction of pneumococcal conjugate vac-
cine, rotavirus, and human papillomavirus vaccines needs additional storage
capacity of 100 cm3 and adds a cost up to $20 per child (Chauke-​Moagi &
Mumba, 2012). In Thailand, 8.3% of healthcare centers required one addi-
tional refrigerator, 50% or 54.2% of the district warehouses for an additional
one or two vaccines, respectively (Riewpaiboon et al., 2015). Different vac-
cines have consequences for cold chain storage and waste management. For
instance, the cold chain volumes for pneumococcal conjugate vaccine, ro-
tavirus, and human papillomavirus vaccines were estimated to be 55.9 cm3,
46.3 cm3, and 15 cm3, respectively (Ngabo et al., 2015).
An important tool for estimating vaccine volume is the World Health
Organization vaccine volume calculator (World Health Organization, 2021).
This tool can help estimate the net storage volume per vaccine per child as well
as the net storage volume per injection supplies and diluents per child. This
 2.5.3 NUVI cost estimation 129

will assist in the estimation of whether new cold chain equipment is required.
The ratio of the share of vaccine volume for the new vaccine compared to ex-
isting storage volumes can be used to allocate cold chain cost to NUVI costs.
Capital investments made during the start-​up period are usually annual-
ized (Walker & Kumaranayake, 2002), following the same approaches out-
lined in Chapter 2.4. Capital investments include cold chain equipment, as
well as initial training and information, education, and communication ma-
terials for social mobilization and advocacy. One approach is to allocate cap-
ital item costs that occur 6 months prior to and 6 months after introduction to
NUVI costs.

2.5.3.2 Ongoing recurrent costs of NUVI

Ongoing recurrent costs for NUVI pertain to those inputs that will be con-
tinued past the initial introduction phase, and include vaccines and supplies,
personnel, per diems, transportation operation and maintenance of cold
chain equipment and vehicles, monitoring and evaluation, and supervision
(Brenzel, Young, & Walker, 2015; Hidle et al., 2018; Levin et al., 2014; World
Health Organization: Immunization Vaccines and Biologicals, 2002).
A combination vaccine with fewer injections requires fewer needles and
syringes, and thus has lower administrative costs and vaccine storage costs
compared to a monovalent vaccine (Hidle et al., 2018). For economic costs,
donated vaccines, freight, customs, transport of vaccine and supplies, vac-
cine carriers, and cold packs should be factored into the estimates (Hyde
et al., 2012).
For the freeze-​dried vaccines, a reconstitution syringe will be required.
The number of reconstitution syringes can be estimated using the number of
freeze-​dried vaccine doses administered per year, number of doses per vial,
wastage rate, reserve stock, and price per syringe including freight rate. In case
of syringes, wastage rate is generally fixed at 10%. Similar to vaccines, a re-
serve stock of 25% needs to be added in the first-​year demand calculation at
the district and state level.
NUVI will require the involvement of a wide range of staff members for
planning, training, social mobilization, surveillance, monitoring and super-
vision, administration of the vaccine, waste management, and reporting.
However, most of these staff members will not be exclusively involved in
NUVI; therefore, personnel cost needs to be calculated based on time alloca-
tion for the new vaccine. The hours spent in each NUVI-​related activity can
130 Costing new vaccine introduction

be multiplied by the hourly wage (obtained from gross salary and work hours
per week) to obtain the personnel cost related to NUVI.
NUVI will require extensive advocacy and social mobilization activities to
inform the general public and healthcare workers about the new vaccine. It
may involve displaying flyers and posters, distributing leaflets, conveying the
message regarding the new vaccine through celebrities and actors, and advert-
isements through the media (radio/​television/​print). The incremental costs
related to advocacy and social mobilization need to be calculated by gathering
information on all such activities occurring before and during the introduc-
tion period. The majority of the social mobilization costs may be start-​up
costs which can be annualized.
Disease surveillance and monitoring of the new vaccine will be integrated
into the existing system. The incremental costs of surveillance and evaluation
include additional staff, training programs, transport, and laboratory find-
ings. Recurrent costs associated with the new vaccine will revolve around ad-
ditional personnel, transportation, and per diems (Erondu, Ferland, Haile, &
Abimbola, 2019).
Table 2.5.2 provides a framework for evaluating the costs of vaccination ac-
tivities related to NUVI. These are divided into start-​up and recurrent costs,
split into their financial or economic costs.

2.5.4 Data sources

The main data sources for calculating the incremental cost of NUVI will be
immunization program managers at different levels. Information about dif-
ferent activities related to NUVI can be gathered through discussions with
officials and the related staff using a standardized questionnaire. Financial
records and other key documents related to the number of doses delivered
can be obtained and reviewed at national or subnational health authority
offices. To determine the time spent on new vaccine administration and
other activities such as record keeping, waste management, and sensitiza-
tion of the community, facility staff should be interviewed using a stand-
ardized questionnaire. Other important sources of information include
the country’s new vaccine application to Gavi which provides a detailed
budget for NUVI. The Comprehensive Multi-​Year Plan or the National
Immunization Strategy may also provide useful information for NUVI
costing. Records from UNICEF and World Health Organization offices in
the country will be other sources of data.
 2.5.4 Data sources 131

Table 2.5.2 Activities and cost components of NUVI

Cost and description Financial cost Economic cost

Start-​up costs
Micro-​planning related to new Room rental Financial cost +​value of
vaccine introduction, including Refreshments staff time
meetings and events at different Per diem
health system levels Printing
Training Printing of training materials Financial cost +​value
Training sessions and Supplies of staff time +​value of
development of materials for Venue rental volunteer worker time
a range of healthcare workers Refreshments
and administrators Per diem
Social mobilization and Venue rental Financial cost +​value of
advocacy Refreshments staff time +​value of
Community level and health Supplies volunteer worker time
worker sensitization Printing of IEC materials Financial cost of
to the new vaccine, Development of radio, materials to be
including meetings, events, television, social media annualized
development and production, messages, and materials
and materials and media Payment for securing radio/​TV
spots, and newspaper,
internet, and social media
advertising
Cold chain Cold chain equipment Annualized financial cost
Purchase of new cold chain +​value of donated items
equipment and/​or use of +​value of existing space
existing cold storage space at used for new vaccine
various levels of the health
system for storing and
distributing the new vaccine
Transport Vehicle Annualized financial cost
Purchase of additional vehicles, +​value of donated items
additional donated vehicles, +​value of usage for new
and/​or use of existing vehicles vaccine
to transport vaccines and
vaccinators.
Waste management Incinerators Annualized financial cost
Purchase of additional +​value of existing space
incinerators and/​or use of used for new vaccine
existing incinerators
Recurrent costs
Vaccines and injection supply Cost of vaccine (excluding Financial cost +​cost
donor-​supported vaccines) of vaccine (less co-​
Cost of syringes financing fees)
Cost related to customs, receiving In case of donated
transport, and storage vaccine: financial
Co-​financing fees for cost +​cost of donor-​
Gavi-​supported vaccines supported vaccine
(continued)
132 Costing new vaccine introduction

Table 2.5.2 Continued

Cost and description Financial cost Economic cost

Service delivery Per diem Financial cost +​value

Personnel time spent on Travel allowance of staff time +​value of
administering the new Fuel and maintenance volunteer worker time
vaccine, incremental per
diem, and transportation costs
Supervision, monitoring, and Per diem Financial cost +​value of
evaluation Travel allowance staff time
Ongoing supervision and Fuel and maintenance
record keeping, reporting Printing vaccination cards and
activities related to the new reporting formats
vaccines
Disease surveillance Refreshments Financial cost +​value of
Activities specific to identifying Per diem staff time
adverse events following Travel allowance
immunization, and cases of Laboratory supplies
disease Fuel and maintenance
Waste management Safety boxes Financial cost +​value of
Management of waste at the Hub cutters staff time
vaccination site Containers
Per diem
Refreshments
Vaccine transport Vehicle Financial cost +​value of
Vehicle use related to donated items +​value of
administration of the new existing space used for
vaccine new vaccine
Cold chain Ongoing running costs related Financial cost
to storage of the new vaccine
on a recurrent basis (energy)
Ongoing training Ongoing costs associated with Financial cost
refresher training related to the
new vaccine
Social mobilization and Ongoing social mobilization and Financial cost
advocacy advocacy costs

2.5.5 Conclusion

This chapter highlights a general approach and framing for evaluating the
costs of NUIV. Costs for NUVI are incremental to the existing system and
can be evaluated as fiscal, financial, or economic costs depending upon the
use of the results by decision makers. In addition, costs should be divided into
initial/​start-​up costs or ongoing, recurrent costs for NUVI. Guidance docu-
ments are available to provide additional recommended approaches for data
collection and analysis.
 2.5.5 Conclusion 133

References
Brenzel, L. (2014). Working Paper: Common approach for the costing and financing analyses
of routine immunization and new vaccine introduction costs (EPIC). Mimeograph. Bill &
Melinda Gates Foundation. www.immuni​zati​onec​onom​ics.org/​epic-​info.
Brenzel, L., Young, D., & Walker, D. G. (2015). Costs and financing of routine immuniza-
tion: Approach and selected findings of a multi-​country study (EPIC). Vaccine, 33, A13–​
A20. doi:10.1016/​j.vaccine.2014.12.066
Chauke-​Moagi, B. E., & Mumba, M. (2012). New vaccine introduction in the East and Southern
African sub-​region of the WHO African region in the context of GIVS and MDGs. Vaccine,
30(Suppl 3), C3–​C8. doi:10.1016/​j.vaccine.2012.05.086
Erondu, N. A., Ferland, L., Haile, B. H., & Abimbola, T. (2019). A systematic review of vac-
cine preventable disease surveillance cost studies. Vaccine, 37(17), 2311–​2321. doi:10.1016/​
j.vaccine.2019.02.026
Gavi. (2021). Homepage. http://​www.gavi.org/​
Griffiths, U. K., Bozzani, F. M., Chansa, C., Kinghorn, A., Kalesha-​Masumbu, P., Rudd,
C., . . . Schutte, C. (2016). Costs of introducing pneumococcal, rotavirus and a second dose of
measles vaccine into the Zambian immunisation programme: Are expansions sustainable?
Vaccine, 34(35), 4213–​4220. doi:10.1016/​j.vaccine.2016.06.050
Hidle, A., Gwati, G., Abimbola, T., Pallas, S. W., Hyde, T., Petu, A., . . . Manangazira, P. (2018).
Cost of a human papillomavirus vaccination project, Zimbabwe. Bulletin of the World
Health Organization, 96(12), 834–​842. doi:10.2471/​BLT.18.211904
Hyde, T. B., Dentz, H., Wang, S. A., Burchett, H. E., Mounier-​Jack, S., & Mantel, C. F. (2012).
The impact of new vaccine introduction on immunization and health systems: A review of
the published literature. Vaccine, 30(45), 6347–​6358. doi:10.1016/​j.vaccine.2012.08.029
Le Gargasson, J. B., Nyonator, F. K., Adibo, M., Gessner, B. D., & Colombini, A. (2015). Costs
of routine immunization and the introduction of new and underutilized vaccines in Ghana.
Vaccine, 33(Suppl 1), A40–​A46. doi:10.1016/​j.vaccine.2014.12.081
Levin, A., Wang, S. A., Levin, C., Tsu, V., & Hutubessy, R. (2014). Costs of introducing and
delivering HPV vaccines in low and lower middle income countries: Inputs for GAVI policy
on introduction grant support to countries. PLoS One, 9(6), e101114. doi:10.1371/​journal.
pone.0101114
Loze, P. M., Nasciben, L. B., Sartori, A. M. C., Itria, A., Novaes, H. M. D., & de Soárez, P.
C. (2017). Vaccines are different: A systematic review of budget impact analyses of vac-
cines. Vaccine, 35(21), 2781–​2793. doi:10.1016/​j.vaccine.2017.03.088
Mvundura, M., Kien, V. D., Nga, N. T., Robertson, J., Cuong, N. V., Tung, H. T., . . . Levin, C.
(2014). How much does it cost to get a dose of vaccine to the service delivery location?
Empirical evidence from Vietnam’s Expanded Program on Immunization. Vaccine, 32(7),
834–​838. doi:10.1016/​j.vaccine.2013.12.029
Ngabo, F., Levin, A., Wang, S. A., Gatera, M., Rugambwa, C., Kayonga, C., . . . Hutubessy, R.
(2015). A cost comparison of introducing and delivering pneumococcal, rotavirus and
human papillomavirus vaccines in Rwanda. Vaccine, 33(51), 7357–​ 7363. doi:10.1016/​
j.vaccine.2015.10.022
Pooripussarakul, S., Riewpaiboon, A., Bishai, D., Tantivess, S., & Muangchana, C. (2018).
Identifying important factors impacting the adoption of new vaccines in Thailand by using a
best-​worst scaling. Journal of Health Systems Research, 12(1), 137–​149. https://​kb.hsri.or.th/​
dsp​ace/​bitstr​eam/​han​dle/​11228/​4857/​hsri​_​jou​rnal​_​v12​n1_​p​137.pdf
Riewpaiboon, A., Sooksriwong, C., Chaiyakunapruk, N., Tharmaphornpilas, P., Techathawat,
S., Rookkapan, K., . . . Suraratdecha, C. (2015). Optimizing national immunization program
134 Costing new vaccine introduction

supply chain management in Thailand: An economic analysis. Public Health, 129(7), 899–​
906. doi:10.1016/​j.puhe.2015.04.016
ThinkWell. (2020). Immunization delivery cost catalogue. http://​www.immuni​zati​onec​onom​
ics.org/​ican
Walker, D., & Kumaranayake, L. (2002). Allowing for differential timing in cost ana-
lyses: Discounting and annualization. Health Policy and Planning, 17(1), 112–​118.
doi:10.1093/​heapol/​17.1.112
World Health Organization. (2021). Vaccine volume calculator. https://​www.who.int/​publi​cati​
ons/​m/​item/​vacc​ine-​vol​ume-​cal​cula​tor
World Health Organization: Immunization Vaccines and Biologicals. (2002). Guidelines for
estimating costs of introducing new vaccines into the national immunization system. https://​
apps.who.int/​iris/​han​dle/​10665/​67342
 3
ECONOMIC EVALUATION OF VACCINES
AND VACCINE PROGRAMS

Edited by William V. Padula, Emmanuel F. Drabo, and Ijeoma Edoka

3.0
Section introduction: economic
evaluation of vaccines and
vaccine programs
William V. Padula, Emmanuel F. Drabo, and Ijeoma Edoka

Economic evaluation is increasingly used to inform decision makers about

the value that various interventions deliver to health systems. Sometimes the
greatest value in healthcare is achieved not by what is provided to the patient
in terms of services and technology, but what is avoided. We know for a fact
that prevention of infectious diseases is a win–​win for patients, providers, and
society as a whole. When we invest in the prevention of infectious diseases
through vaccines, the relatively small costs of vaccines achieve enormous value.
Providers and payers do not have to care for as many patients facing the symp-
toms of infectious disease. And patients do not have to deal with the personal
costs of treatment, or deal with the acute and long-​term fall out of infectious
disease. We know that many diseases ranging from influenza to polio have im-
pacts on well-​being in ways that are detrimental to the patient and the produc-
tivity of society as a whole. Methods in economic evaluation measure impacts of
these outcomes and their value.
All these aspects of value added through vaccines might be implicit, but we
need to use empirical science in order to quantify value so that decision makers
can make informed choices about the adoption of vaccines for entire popula-
tions of at-​risk individuals. These are not small investments. Investments in
vaccines that represent whole percentages of a nation’s gross domestic product
rely on accurate assessments of value to solidify decisions. Approaches in eco-
nomic evaluation, ranging from cost-​effectiveness to budget impact analysis,
are tools that can be used to inform decision makers. The methods to conduct
these analyses are not always easy, but we recommend starting with simple,

William V. Padula, Emmanuel F. Drabo, and Ijeoma Edoka, Section introduction: economic evaluation of vaccines and vaccine
programs In: Handbook of Applied Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula,
Oxford University Press. © Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0013
138 Economic evaluation of vaccines and vaccine programs

intuitive models to calculate the resulting measures of value that decision

makers can rely on.
This section is broken down into several chapters which are necessary
steps in a progression toward achieving economic evaluation. Here, we take
you through a synopsis of each chapter in order to understand what data are
needed to build a value assessment.
Chapter 3.1 is an overview of decision analysis and cost-​effectiveness. Here
we want to understand value by defining it. Many academics have often ex-
pressed value as a function of cost and quality, but measuring vaccine value
and expressing it quantitatively is not so straightforward. By focusing on
cohort-​level analyses of value, we can more accurately determine the impact
that a certain vaccine will have for a specific population that is at risk for in-
fectious disease. Thus, decision makers can make local, replicable decisions
about investment in vaccines that protect different populations based on
regional variation, age, sex, race, or ethnicity. The chapter introduces basic
concepts such as the difference between cost-​effectiveness and cost–​benefit
analysis, as well as how to derive measures of budget impact and return on in-
vestment from a single economic model.
Chapter 3.2 offers guidance on defining the scope and study design of a
cost-​effectiveness analysis. As mentioned earlier, healthcare involves many
stakeholders ranging from payers, to providers, to integrated health systems,
to patients, and ultimately to society as a whole. Economic models can range
in their measurement of costs and effectiveness depending on the perspec-
tive of these stakeholders. We will explore how these perspectives impact the
design and calculation of different economic model parameters. We will also
highlight basic types of economic models (i.e., decision trees and Markov
chains) which are used to simulate the impacts of vaccines on different types
of static or time-​dependent health outcomes, as well as accumulate costs and
effectiveness measures.
Chapter 3.3 focuses on parameter estimation. Economic models that are
used to measure vaccine value are only as strong as the data used in the model.
Health economists depend on the availability of data from a number of dif-
ferent reliable sources in order to populate a model with parameters. These
sources include some of the typical study designs that one might encounter
from data on vaccine effectiveness, such as randomized controlled trials,
observational studies, and so on. The chapter explores how to extract high-​
quality information about costs, effectiveness, and probability parameters
that can be used to make an economic model functional.
 Economic evaluation of vaccines and vaccine programs 139

Chapter 3.4 reviews the methods used to measure and evaluate health out-
comes. There are a number of different measures that economists have devel-
oped over the years to quantify the effectiveness of vaccines. Two of the most
commonly used in cost-​effectiveness analysis are quality-​adjusted life years
and disability-​adjusted life years, in addition to more intuitive measures such
as mortality and incidence rates in an at-​risk population. The chapter explores
these effectiveness measurements and the types of psychometric approaches
that have been developed to derive estimates of effectiveness.
Chapter 3.5 explains some of the important considerations for reporting
and interpreting the results of economic evaluations. Once a model has been
developed and populated with parameter information, an investigator should
be prepared to understand what the results are telling them about value from
the perspective of the key stakeholders. Incremental cost-​effectiveness ratios,
budget impacts, and return on investment results are all key statistics that one
should be able to know and interpret to make recommendations about vac-
cine value. Furthermore, there are number of data visualization techniques
that can go along with these statistics to make the results clearer to decision
makers about value.
Chapter 3.6 examines other measures of economic efficiency that may be
important to decision makers. Not all determinations of value end with cost-​
effectiveness analysis. Often, decision makers depend on budget impact anal-
ysis (BIA) to determine affordability with respect to budget constraints. In
addition, return on investment (ROI) calculations give decision makers re-
assurance that vaccine programs are not sunken costs, but rather investments
in the public’s health that pay dividends over time. By becoming familiar with
multiple forms of value estimation, economic modelers can be prepared to
produce results that gain the attention of decision makers, whatever their
preference may be for information.
Chapter 3.7 consolidates the material on economic evaluation for vaccines
throughout this section with an applied case. By following along with the con-
struction of a simple decision tree and data parameter inputs, one should be
able to accurately derive and interpret key resulting statistics of value for a
common example in vaccine decision-​making.
3.1
Overview of decision analysis and
cost-​effectiveness
Emmanuel F. Drabo, Ijeoma Edoka, and William V. Padula

3.1.1 Understanding “value”

Resource allocation decisions within the health sector can be made based
on the benefits or value of the vaccine program and its associated costs.
Economic evaluations typically compare the benefits and costs of competing
alternatives to assess whether a vaccine technology represents good value for
money. Vaccines are considered one of the most effective health technologies
for preventing mortality and morbidity due to infectious disease. In addition
to their direct health benefits in vaccinated individuals, vaccines can lead to
indirect health benefits in unvaccinated individuals due to a reduction in the
transmission of disease (e.g., the herd effect) as well as broader microeco-
nomic and macroeconomic benefits that extend beyond health.
In the economic evaluation of new vaccines, the value of vaccines is typi-
cally expressed in terms of health gains (both direct and indirect) measured
using different outcomes such as the number of infections averted, deaths
averted, life years gained, disability-​adjusted life years (DALYs), or quality-​
adjusted life years (QALYs). The value of vaccines also extends beyond health
benefits to include direct and indirect economic benefits to individuals and
healthcare systems. Direct economic benefits include costs savings due to re-
ductions in direct cost incurred by individuals when seeking healthcare or
direct costs incurred by healthcare systems in delivering healthcare for in-
fectious diseases. Indirect economic benefits, on the other hand, may include
productivity gains to vaccinated individuals accruing from reductions in
days lost from work due to illness or premature death as well as productivity
gains to caregivers, also accruing from reductions in days lost from work to
care for an ill individual.

Emmanuel F. Drabo, Ijeoma Edoka, and William V. Padula, Overview of decision analysis and cost-effectiveness In: Handbook
of Applied Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0014
 3.1.2 The importance of economic evaluation of vaccines 141

The microeconomic benefits of vaccines have traditionally been used for

assessing the value for money of vaccines when making funding decisions.
However, these benefits may underestimate the true value of vaccines and a
growing body of literature has highlighted the broader value of vaccination
that goes beyond the direct/​indirect health and health-​related economic
benefits to individuals and health systems (Bärnighausen, Bloom, Cafiero-​
Fonseca, & O’Brien, 2014; Bärnighausen, Bloom, Canning, & O’Brien, 2008;
Deogaonkar, Hutubessy, van der Putten, Evers, & Jit, 2012; Jit et al., 2015;
Ozawa, Mirelman, Stack, Walker, & Levine, 2012). These broader benefits can
be more important in low-​resource settings where funding decisions are often
made alongside broader development goals that fall outside the health sector
(Deogaonkar et al., 2012). Bärnighausen et al. (2008) classify benefits of vac-
cines as narrow or broad. Narrow benefits include health gains, healthcare
cost savings, and care-​related productivity gain and broad benefits include
outcome-​related productivity gains, behavior-​related productivity gains, and
community externalities.
This framework has since been extended to incorporate other broad
values that are potentially important for making resource allocation deci-
sions on introducing new vaccines. These include broader community and
health system externalities such as ecological effects, equity, financial sus-
tainability, and household security as well as broader economic indicators
such as public sector budget impact and macroeconomic impact. While the
link between vaccination and the so-​called narrow benefits of vaccines are
well established, the causal link between vaccines and several broad bene-
fits remains less certain (Jit et al., 2015), and may limit their incorporation
into assessments of the value for money of vaccines.

3.1.2 The importance of economic evaluation

of vaccines

Economic evaluations, such as cost-​effectiveness analyses (CEAs), cost–​

utility analyses (CUAs), and cost–​benefit analyses (CBAs), compare the costs
and benefits of competing alternatives (e.g., preventive vaccines versus treat-
ment of disease) to assess trade-​offs that arise when inevitable choices must be
made between alternative courses of action and budgets are fixed. Comparing
costs to valued outcomes is useful for decision makers in order to choose
health interventions that produce the highest benefits for a given cost. Valuing
outcomes is simplest when health maximization is the sole policy objective.
142 Overview of decision analysis and cost-effectiveness

Given the high value of vaccines in terms of their direct and indirect health
and economic benefits to vaccinated and unvaccinated individuals and the
relatively low costs of first-​generation vaccines, vaccines have largely been
shown to be a cost-​effective health intervention for preventing mortality and
morbidity associated with infectious diseases.
With growing innovation and technological advancements in the devel-
opment of new vaccines, the costs of new and next-​generation vaccines are
expected to be substantially higher (Delany, Rappuoli, & De Gregorio, 2014;
Tahamtan, Charostad, Hoseini Shokouh, & Barati, 2017). For example, human
papillomavirus (HPV) and pneumococcal conjugate vaccines have a signif-
icantly higher price per dose compared to first-​generation vaccines such as
yellow fever and oral polio vaccines. The high costs of new-​generation vac-
cines have significant budget implications for many countries, particularly for
middle-​income countries with limited access to external funding mechan-
isms. Given that decisions to fund new vaccination programs must be made in
the wider context of other competing goals both within and outside the health
sector, economic evaluations that capture the value of new-​generation vac-
cines in addition to their costs are important for assessing the value for money
of these vaccines and motivating to increase allocation of resources to new
vaccination programs.
In addition to informing resource allocation decisions on the introduc-
tion of new vaccination programs, economic evaluations can be useful for
informing the prioritization of vaccination programmatic strategies. For ex-
ample, the World Health Organization recommends various programmatic
strategies for eliminating rubella infection and congenital rubella syndrome
within a 10–​30-​year timeframe (World Health Organization, 2011). These
strategies range from vaccinating all children from 9 months to 4 years old
to vaccinating only women of childbearing age or vaccinating all adolescents
and adults between the ages of 15 and 39 years (World Health Organization,
2011). Economic evaluation has been applied in several countries, particu-
larly in high-​income countries to inform decisions on rubella elimination
strategies (Babigumira, Morgan, & Levin, 2013). Other examples of the ap-
plication of economic evaluation include the prioritization of HPV vaccina-
tion strategies such as decisions about when to add adolescent boys into the
existing female-​only HPV vaccination programs. CEAs showing higher cost-​
effectiveness of the HPV vaccine in girls than in in boys (Datta et al., 2019;
Seto, Marra, Raymakers, & Marra, 2012) have contributed to informing this
decision.
 3.1.3 Introduction to measures of value 143

Economic evaluations can also be useful as price negotiation tools through

the value-​based pricing approach. This approach allows for the determination
of the price of new health technologies based on the maximum price at which
the technology is considered cost-​effective using predefined cost-​effectiveness
thresholds of decision makers (see Chapter 3.5). This approach required a
consensus on the type(s) of value to be measured as well as an appropriate
threshold or benchmark against which this value is assessed (Claxton, 2007;
Drummond & Towse, 2019). Value-​based pricing approaches have been
widely adopted in several countries that use CEAs to inform reimbursement
decisions such as in the UK (Claxton, 2007; Drummond & Towse, 2019) and
Thailand (Teerawattananon, Tritasavit, Suchonwanich, & Kingkaew, 2014),
and have been instrumental in these countries for negotiating reductions
in the price of new health technologies including vaccines. For example, in
Thailand, data from CEA informed price negotiations resulting in a 55% re-
duction in the original asking price of a HPV vaccine (Teerawattananon &
Tritasavit, 2015). Although value-​based pricing is an instrumental tool for
price negotiation of vaccines, it may not always be appropriate for setting the
price of some health technologies, particularly interventions against rare con-
ditions or interventions that cost significantly less than an existing interven-
tion (Drummond & Towse, 2019).

3.1.3 Introduction to measures of value

Vaccines may generate both positive and negative externalities beyond the
healthcare sector. These externalities can significantly influence disease pat-
terns, as well as the impacts of disease on healthcare and society. At the same
time, the procurement and the distribution of vaccines require significant re-
sources. However, with scarce healthcare resources and multiple competing
needs over these resources, cost containment is becoming important. It is crit-
ical that whenever a decision is to be made about adopting a new vaccine, the
additional benefit produced by the intervention be sufficiently larger than the
cost. In other words, we need to ensure that the new intervention or policy
provides good value. To more accurately assess the value generated by a vac-
cination program or the introduction of a new vaccine, it is therefore critical
that we capture the value of all its relevant economic costs and benefits. Given
that vaccines may have value to a variety of non-​healthcare stakeholders such
as governments and households, vaccine value measurement must be broad
144 Overview of decision analysis and cost-effectiveness

to capture the downstream consequences of vaccine introduction, such as po-

tential displacements in healthcare resources.
Economic evaluation can systematically determine the “value for money”
of healthcare interventions, such as immunization programs. By “value for
money,” we mean the monetary value of the benefits of the program, that is,
the maximum amount of money a payer such as a government or an insur-
ance company, would be willing to pay to enjoy these benefits, also known as
the willingness to pay. Measures of value can incorporate better outcomes as
well as ideals such as increased quality, better access, and equity (Goldman &
McGlynn, 2005; Padula, Lee, & Pronovost, 2017). Decisions require a compar-
ison of value to costs, often using the ratio of value to cost (Nelson, Batalden,
Godfrey, & Lazar, 2011; Porter & Advantage, 1985). By economic cost, we
mean the opportunity cost of adopting the program, that is, the cost of the
next best alternative use of the resources devoted to the program. Hence, ec-
onomic value can be more broadly defined as the comparative analysis of the
costs and consequences of an intervention, policy, or strategy against similar
outcomes of its next best alternative(s).
Measurement of the cost component of the value equation is done by as-
signing monetary values to resources used. In contrast, measurement of out-
comes, quality access, and equity is more challenging, because these goals
are can be less tangible than cost, and involve multidimensional constructs
(Garrison, Jansen, Devlin, & Griffin, 2019; Padula et al., 2017). Economic
evaluation methods vary in their approaches to assessing the value of new and
existing interventions, mainly along the types of health outcomes used. The
most common types of economic evaluation analyses include cost of illness
(COI) analysis, budget impact analysis (BIA), return on investment (ROI)
analysis, cost comparison analysis, cost-​minimization analysis (CMA), cost–​
consequence analysis (CCA), CEA, CUA, and CBA.
The first two methods (i.e., COI and BIA) assess the burden of disease and
affordability of a new intervention. More concretely, COI analysis aims to de-
termine the economic impact of an illness or condition, such as hepatitis B, on
a given population (e.g., people with diabetes) or geographic area (e.g., region,
country), including its associated treatment costs. BIA estimates the afforda-
bility, that is, the impact of adopting a new intervention, such as introducing a
new vaccine, on a payer’s budget (Mauskopf, Earnshaw, Brogan, Wolowacz, &
Brodtkorb, 2017; Sullivan et al., 2014). The value of an intervention can there-
fore be determined in COI analysis and BIA by examining the opportunity cost
of implementing that intervention.
 3.1.3 Introduction to measures of value 145

The remaining economic evaluation methods listed above are more ex-
plicitly designed to assess value for money spent. Specifically, ROI analysis
evaluates the efficiency of an intervention, by measuring the amount of re-
turn (i.e., net benefit) on the intervention in monetary terms, relative to its
cost. Hence, ROI measures the value of investing in the intervention. CMA
determines the least costly intervention among competing alternatives that
are assumed to produce equivalent health outcomes (Berger, Bingefors,
Hedblom, Pashos, & Torrance, 2003). CEA compares two or more alterna-
tive interventions in terms of their costs, expressed in monetary units, and
health outcomes, expressed in “natural” health units such as mortality or
morbidity. CUA is a particular case of CEA, in which the health outcomes
of interventions are expressed, unlike CEA, in terms of utility and mortality,
which is expressed in QALYs. CCA is a form of CEA in which costs and
health outcomes are presented in disaggregated form, thus allowing the de-
cision maker to place preference weights on the costs and health outcomes.
Finally, CBA compares the costs and health benefits of an intervention, in
terms of monetary units.
Because CBA, CCA, CEA, CUA, and CMA measure and compare the costs
and outcomes of at least two interventions, they are referred to as full eco-
nomic evaluations (Table 3.1.1). Analyses that only evaluate a single interven-
tion or consider only one domain (e.g., cost only or effectiveness only) are
known as partial economic evaluations.
Outcomes in health can be multidimensional constructs, leading to several
measures, including QALYs gained, net costs, insurance value, value of hope,
and so on (Garrison et al., 2019). Costing is the approach used to evaluate
the “next best” use of scarce resources that would be needed to produce a
certain health effect. Depending on the context, QALYs and DALYs are the
most commonly used dimensions of value assessment in healthcare decision-​
making. As a composite measure which combines the length of life with pref-
erences (health-​related quality of life), QALYs provide a measure of value of
health outcomes which is advantageous for evaluating health technologies
and interventions.
While value assessment for vaccines can be conducted from various per-
spectives, the fact that vaccines generate significant externalities that can span
multiple sectors implies that vaccine value assessment should be conducted
at the population level rather than at the individual level. This requires using
measures of value that make sense at the population level and a broader evalu-
ation beyond the clinical focus.
Table 3.1.1 Types of economic evaluations based on costs and outcomes typologies

Type of analysis Costs Outcomes Comparison of

interventions
Inclusion Valuation Inclusion Valuation

Partial economic evaluation

Cost of illness analysis (COI) Yes Monetary units No –​ None (disease-​level
analysis)
Budget impact analysis (BIA) Yes Monetary units Yes None or maximize various Yes
Return on investment (ROI) analysis Yes Monetary units Yes Monetary units Yes, ROI
Cost comparison analysis Yes Monetary units No –​ Yes

Full economic evaluation

Cost-​minimization analysis (CMA) Yes Monetary units Yes None (assume unchanged) Yes
Cost–​consequence analysis (CCA) Yes Monetary units Yes Natural units Yes
Cost-​effectiveness analysis (CEA) Yes Monetary units Yes Natural units Yes, ICER or NMB
Cost–​utility analysis (CUA) Yes Monetary units Yes Utilities: QALY, DALY, healthy Yes, ICER or NMB
year equivalent
Cost–​benefit analysis (CBA) Yes Monetary units Yes Monetary units Yes, cost–​benefit
ratio or net benefit
 3.1.4 Define methods of value analysis 147

3.1.4 Define methods of value analysis

Health economics provides tools for defining and measuring value for
money. Tools of economic evaluation enable identification of the most effi-
cient course of action from a set of alternative options, that is, the alternative
that provides the highest value for money. CEA is one of the most widely
used methods of economic evaluation; other methods include cost-​finding
or cost-​identification analysis (CFA/​ CIA), CMA, COI analysis, CCA,
CBA, BIA, and ROI analysis (Drummond, Sculpher, Claxton, Stoddart, &
Torrance, 2015).
CFA/​CIA is the process of identifying all costs and their relative impor-
tance (Berger et al., 2003). The goal of a CFA/​CIA is to identify resources used
to produce a particular program or intervention, and to estimate the mon-
etary value of these resources. The cost data generated can then be used as
standalone information for decision-​making or integrated into decision sup-
port frameworks to inform policymakers.
Cost comparison analysis compares only the costs associated with two or
more alternative healthcare interventions. Costs are typically identified
through a CFA/​CIA. The method consists of comparing the costs of two dif-
ferent treatment mixes (including medical and non-​medical products and
services). Cost comparison analysis is particularly useful in situations when
the interventions are recurrent and complex (e.g., multiple dosage regimens),
because it can help quantify and compare the differences in costs between al-
ternative strategies (e.g., differences in costs of vaccination programs with
different intensities or coverage levels). As it is essential to identify all costs
related to an intervention, including the treatment costs, and costs of drugs,
devices, nutritional supplements, and professional care, policymakers usually
use COI analysis (described below) instead of a cost comparison analysis.
CMA aims to identify the least costly alternative among a set of options
with similar expected health outcomes (Berger et al., 2003). CMA achieves
this by identifying resources consumed in the provision of alternative inter-
ventions, to estimate and compare the monetary value of these resources
across alternatives. In rare situations when health outcomes are extremely
similar across alternative strategies, the least expensive (least costly) inter-
vention should be prioritized. In the context of vaccine economics, one may
use CMA to decide, for example, between a combination measles, mumps,
rubella, and varicella vaccine to separate injections of its equivalent compo-
nent vaccines (i.e., a measles, mumps, and rubella vaccine plus a varicella vac-
cine), provided that the health outcomes of the two options are equal. CMA
148 Overview of decision analysis and cost-effectiveness

is criticized for too often assuming identical health outcomes across inter-
ventions when this is hardly ever the case (Briggs & O’Brien, 2001). For ex-
ample, all vaccination campaigns are not created equal, and could result in
different effectiveness outcomes. Despite this criticism, CMA can be very
useful in resource-​constrained settings to inform cost-​cutting measures when
outcomes are similar across interventions of varying costs. Even in developed
countries, CMA can be useful, as resources are scarce and must be efficiently
allocated among interventions that yield the best value for money. For ex-
ample, generic versions or cheaper bioequivalent of drugs are now more com-
monly preferred over costly branded drugs, to minimize costs, as these price
variations are sometimes very substantial.
COI analysis, also called “burden of illness” analysis, focuses on the health
status of a target population without an intervention to estimate both health-
care resources consumed and associated costs. The approach consists of
measuring the medical and other costs resulting from a specific disease or
condition. Unlike cost comparison analysis, which only measures the costs
of different interventions, COI measures all relevant cost of the disease, in-
cluding its treatment. “Burden of illness” is quantified with metrics such as the
direct costs of the disease (e.g., medical costs, and non-​medical costs such as
travel costs), mortality (e.g., years of life lost), intangible costs of a disease or
injury (e.g., disabilities, limitations on daily activity), or productivity losses
such as absenteeism (being physically absent from work due to illness or care-
giving) and presenteeism (being physically present at work but working at re-
duced capacity) costs.
CCA identifies resources consumed in the provision of alternative interven-
tions, to estimate the cost of each alternative intervention. In addition, CCA
identifies and quantifies the effects of each intervention. CCA methodology
does not indicate the relative importance of the components enumerated. The
weighting of these components is left at the decision maker’s discretion. For
complex interventions with complex outcomes, CCA can be a relatively diffi-
cult exercise to conduct, as each outcome and its costs must be assessed sepa-
rately, and then evaluated.
CBA is a method derived from the economic theory that compares alter-
native interventions in terms of their net social cost, defined as the difference
of the social cost and social benefit, expressed in monetary value. All relevant
consequences for each intervention are assigned a monetary value, including
survival gains, and days of disability. CBA can be used to evaluate the wor-
thiness of adopting a single program (i.e., whether the program’s net social
benefits are positive) or to compare alternative interventions (i.e., determine
 3.1.4 Define methods of value analysis 149

which program yields the greatest net social benefit). The net monetary ben-
efit (NMB) of the intervention is then calculated as follows:
Equation 3.1.1. NMB:

NMB = Benefit of intervention − Cost of intervention

Sometime, this value assessment is done by calculating the benefit–​cost ratio:

Equation 3.1.2. Benefit–​cost ratio:

Benefit of intervention
Benefit − cost ratio =
Cost of intervention

When the NMB is positive, or equivalently, when the benefit–​cost ratio ex-
ceeds 1, the intervention is said to be cost beneficial, and to represent good
value for money. In the case where two interventions are being compared, the
incremental NMB can be calculated as the difference of the NMBs. The corre-
sponding incremental benefit–​cost ratio can also be calculated to aid decision-​
making. The CBA approach has been successfully used to estimate the benefits
of Haemophilus influenzae type b vaccine (Bärnighausen et al., 2011).
Social ROI analysis is a pragmatic form of CBA that accounts for effects on
all community stakeholders to include their perspectives on social value where
appropriate. Social ROI analyses appeal to what is known as the “triple bottom
line” to take account of the intervention’s effects on the economy, environ-
ment, and people (Drabo et al., 2021; Then, Schober, Rauscher, & Kehl, 2018).
While CBA methodology is commonly used by economists to value
programs in many sectors (e.g., environmental, transportation, education,
healthcare, etc.) and is a widely accepted valuation method in public projects,
it is often difficult to apply CBA in clinical decision-​making, due to reluctance
by many to assign a monetary value to clinical and health outcomes. Instead,
alternative economic evaluation methods such as CEA and CUA (described
below) are more commonly used in healthcare.
CEA is an economic evaluation methodology used to determine whether
a healthcare intervention such as the introduction of a vaccine produces suf-
ficient improvements in health to justify its costs. A CEA compares two or
more alternative interventions—​including the status quo—​in terms of their
efficiency, by relating their cost and outcomes. CEA does this by identifying
and measuring the costs and health benefits of these alternative interventions.
In CEA, “benefits” are measured in “natural units,” such as survival or life
years gained, clinical outcomes (e.g., reduction in body mass index), number
150 Overview of decision analysis and cost-effectiveness

of cases detected, or number of cases averted. The efficiency of each alterna-

tive is calculated as a cost per unit of benefit produced, relative to the compar-
ator intervention. This is known as the incremental cost-​effectiveness ratio
(ICER), more explicitly defined as the ratio of the incremental cost and incre-
mental health benefits of an alternative intervention, relative to its compar-
ator intervention (see Chapter 3.5). Interventions that yield the largest units
of incremental benefits per cost are more efficient. To determine whether a
given intervention represents good value for money to the payer, the ICER is
compared to the decision maker’s willingness-​to-​pay threshold per unit of the
outcome.
CUA is a specific type of CEA in which benefits are calculated in terms of
marginal gains in both number and quality (health-​related quality of life)
of life years lived. These marginal gains are measured in QALYs, DALYs, or
healthy year equivalent.
CEA and CUA have the promise of improving health outcomes within the
constraints of available funding, given their focus on efficiency. However,
efficiency may not be the only outcome that is important to a decision
maker. For example, a decision maker may also care about reducing social
inequalities or disparities in health that are considered unfair or unjust.
Unfortunately, traditional CEA and CUA provide no information about the
equity impacts of health interventions such as vaccination programs, on so-
cial inequalities. More recent and emerging methods such as extended CEA
and distributional CEA extend the CEA and CUA approaches to explicitly
incorporate equity impacts of healthcare interventions (Asaria, Griffin, &
Cookson, 2016; Cookson, Griffin, Norheim, & Culyer, 2020; Verguet, Kim, &
Jamison, 2016). For more detailed discussion of general approaches to
incorporating equity the equity impacts of social investments in economic
evaluation, the reader should refer to the work by Cookson, Drummond,
and Weatherly (2009).
Other important methods of economic evaluation which deserve a brief
discussion in this chapter and are further discussed in subsequent chapters of
this textbook are BIA and ROI analysis.
BIA is an economic assessment methodology that estimates the financial
consequences of adopting a new intervention. BIA methodology is concerned
with the affordability of adopting a new intervention or healthcare technology
to a payer (Mauskopf et al., 2017; Sullivan et al., 2014). The approach consists
of measuring the cost of the intervention for all those in need in the popula-
tion, on a “per capita” basis, per unit of time. BIA can be used by governments
 3.1.5 Understanding the incremental cost-effectiveness ratio 151

or vaccine payers to assess whether they can afford to introduce a new vaccine
for a particular population subgroup or the entire population.
ROI analysis measures the value of an investment, in terms of its efficiency in
use of resources to produce benefits. The approach consists of determining the
net benefits (i.e., benefits minus costs) of a vaccine program in the long term
over a fixed amount of time (e.g., 1 year, 5 years, or 10 years). This is done by cal-
culating the ROI metric, which is the ratio of the net benefits of an intervention
to its costs, expressed as a percentage:
Equation 3.1.3. Return on investment:

Benefits of investment − Cost of investment

ROI = × 100%
Cost of investment

The ROI captures, therefore, the amount of return (profitability) on a project,

relative to its cost. ROI estimates across different projects can be compared
to assess their relative efficiency. It is worthwhile highlighting here that ROI
analysis differs from CBA in one important way: CBA is an evaluation exer-
cise as its results help identify a better intervention strategy in relation to alter-
native competing strategies, whereas ROI analysis is a valuation exercise that
measures the value of an investment.

3.1.5 Understanding the incremental

cost-​effectiveness ratio

The ICER is the single most important interpretation of a CEA. The ICER can
be represented by a formula:
Equation 3.1.4. Incremental cost-​effectiveness ratio:

Cost A − Cost B ∆Cost

ICER = =
Effectiveness A − Effectiveness B ∆Effectiveness

The ICER provides a fundamental interpretation of the value of a vaccine pro-

gram compared to its next best alternative(s), and this can be interpreted in
multiple directions. For instance, the ICER can be reflective of value as a result
of CEA or CUA, so it is common to see this reported with either approach.
Resulting ICERs can be positive or negative from these analyses, and each
sign change implies different meanings.
152 Overview of decision analysis and cost-effectiveness

First, let’s explore the meaning of an ICER value. If a negative ICER occurs
with a negative numerator it implies that the cost of the vaccine program (A) is
less than the cost of an alternative (B), even though a positive denominator
implies that A has greater clinical effectiveness than B. A negative ICER in a
case with a negative numerator offers more effectiveness at less cost. The op-
posite case where a negative ICER has a positive numerator would imply that
the vaccine program is not in fact a cost-​effective solution at the given price.
Second, ICERs can also be positive, where perhaps the vaccine program
(A) is more costly than the alternative, but also arrives at a greater clin-
ical effectiveness. In such cases, it is important to examine a community’s
willingness-​to-​pay threshold so that we can determine whether or not the vac-
cine program has an opportunity cost which exceeds its alternative (Padula &
Sculpher, 2021).

3.1.6 Translating value into resource allocation

A typical problem is deciding if a vaccine should be allocated to the general

population.
Measuring an ICER relative to the community’s willingness-​ to-​
pay
threshold, also termed “cost-​effectiveness threshold” and symbolized by
λ, is key to examining its priority to be allocated to the general population.
Threshold parameter λ is defined as the maximum value of the ratio of Δ
Cost/​Δ Effectiveness at which an intervention is acceptable. An intervention
that is better than the threshold is said to offer positive net monetary benefit
which can alternatively be expressed as a positive net health benefit as shown
in the equations below.
Note that in the following equations, NMB is different when interpreted
from the results of a CEA or CUA, compared to the interpretation of a CBA:
Equation 3.1.5. NMB:

NMB = ∆Effectiveness × λ − ∆Cost

Equation 3.1.6. Net health benefit:

∆Cost
NHB = ∆Effectiveness −
λ
In either case, net benefits that are positive imply that the vaccine program
would generate value for the community of interest. Likewise, negative net
 3.1.6 Translating value into resource allocation 153

benefits imply that the community would pay more money for less health
benefit than current alternatives at a given willingness to pay.
When considering the value of investments, one must also consider op-
portunity costs. That is, the costs of forgone benefits relating to the next best
option when financial resources are committed to a particular activity. For
example, spending resources on expensive vaccine delivery programs can
withdraw finances from other priorities, such as primary and preventive care,
or treatment of infectious disease. Opportunity costs create differential con-
sequences for stakeholders that will affect their support for an intervention
that might be in the overall public interest. Payers investing in covering vac-
cines for some patients may have limited budgets. In the long run, payers can
overcome budget constraints by increasing premiums or sharing costs with
patients and providers (McGuire, 2000). Providers might not be compen-
sated for all the costs of a new vaccine policy and they might fail to support
it. Overall, these tactics place other higher-​value, preventive programs such
as vaccines in greater jeopardy, and therefore necessitate mindful inclusion of
the various stakeholders’ considerations into models of value and cost.

References
Asaria, M., Griffin, S., & Cookson, R. (2016). Distributional cost-​ effectiveness analysis:
A tutorial. Medical Decision Making, 36(1), 8–​19. doi: 10.1177/​0272989X15583266
Babigumira, J. B., Morgan, I., & Levin, A. (2013). Health economics of rubella: A systematic re-
view to assess the value of rubella vaccination. BMC Public Health, 13(1), 406. doi:10.1186/​
1471-​2458-​13-​406
Bärnighausen, T., Bloom, D. E., Cafiero-​Fonseca, E. T., & O’Brien, J. C. (2014). Valuing vaccina-
tion. Proceedings of the National Academy of Sciences of the United States of America, 111(34),
12313–​12319. doi:10.1073/​pnas.1400475111
Bärnighausen, T., Bloom, D. E., Canning, D., Friedman, A., Levine, O. S., O’Brien, J., . . . Walker,
D. (2011). Rethinking the benefits and costs of childhood vaccination: The example of
the Haemophilus influenzae type b vaccine. Vaccine, 29(13), 2371–​2380. doi:10.1016/​
j.vaccine.2010.11.090
Bärnighausen, T., Bloom, D. E., Canning, D., & O’Brien, J. (2008). Accounting for the full bene-
fits of childhood vaccination in South Africa. South African Medical Journal, 98(11), 842,
844–​846. https://​pub​med.ncbi.nlm.nih.gov/​19177​886/​
Berger, M. L., Bingefors, K., Hedblom, E. C., Pashos, C., & Torrance, G. W. (2003). Health care
cost, quality, and outcomes: ISPOR book of terms. Lawrenceville, NJ: International Society for
Pharmacoeconomics and Outcomes Research.
Briggs, A. H., & O’Brien, B. J. (2001). The death of cost-​minimization analysis? Health
Economics, 10(2), 179–​184. doi:10.1002/​hec.584
Claxton, K. (2007). OFT, VBP: QED? Health Economics, 16(6), 545–​558. doi:10.1002/​hec.1249
Cookson, R., Drummond, M., & Weatherly, H. (2009). Explicit incorporation of equity consid-
erations into economic evaluation of public health interventions. Health Economics, Policy,
and Law, 4(2), 231–​245.
154 Overview of decision analysis and cost-effectiveness

Cookson, R., Griffin, S., Norheim, O. F., & Culyer, A. J. (2020). Distributional cost-​effectiveness
analysis: Quantifying health equity impacts and trade-​offs. Oxford: Oxford University Press.
Datta, S., Pink, J., Medley, G. F., Petrou, S., Staniszewska, S., Underwood, M., . . . Keeling, M.
J. (2019). Assessing the cost-​effectiveness of HPV vaccination strategies for adolescent girls
and boys in the UK. BMC Infectious Diseases, 19(1), 552. doi:10.1186/​s12879-​019-​4108-​y
Delany, I., Rappuoli, R., & De Gregorio, E. (2014). Vaccines for the 21st century. EMBO
Molecular Medicine, 6(6), 708–​720. doi:10.1002/​emmm.201403876
Deogaonkar, R., Hutubessy, R., van der Putten, I., Evers, S., & Jit, M. (2012). Systematic review
of studies evaluating the broader economic impact of vaccination in low and middle income
countries. BMC Public Health, 12(1), 878. doi:10.1186/​1471-​2458-​12-​878
Drabo, E. F., Eckel, G., Ross, S. L., Brozic, M., Carlton, C. G., Warren, T. Y., . . . Pollack, C.
E. (2021). A social-​return-​on-​investment analysis of Bon Secours Hospital’s “Housing For
Health” affordable housing program. Health Affairs (Millwood), 40(3), 513–​520. doi:10.1377/​
hlthaff.2020.00998
Drummond, M., Sculpher, M., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015).
Methods for the economic evaluation of health care programmes (4th ed). Oxford: Oxford
University Press.
Drummond, M., & Towse, A. (2019). Is rate of return pricing a useful approach when value-​
based pricing is not appropriate? European Journal of Health Economics, 20(7), 945–​948.
doi:10.1007/​s10198-​019-​01032-​7
Garrison, L. P., Jansen, J. P., Devlin, N. J., & Griffin, S. (2019). Novel approaches to value assess-
ment within the cost-​effectiveness framework. Value in Health, 22(6), S12–​S17. doi:10.1016/​
j.jval.2019.04.1915
Goldman, D. P., & McGlynn, E. A. (2005). US health care: Facts about cost, access, and quality.
Santa Monica, CA: Rand Corporation.
Jit, M., Hutubessy, R., Png, M. E., Sundaram, N., Audimulam, J., Salim, S., & Yoong, J. (2015).
The broader economic impact of vaccination: Reviewing and appraising the strength of evi-
dence. BMC Medicine, 13(1), 209. doi:10.1186/​s12916-​015-​0446-​9
Mauskopf, J., Earnshaw, S. R., Brogan, A., Wolowacz, S., & Brodtkorb, T.-​H. (2017). Budget-​
impact analysis of health care interventions: A practical guide. Cham: Springer International
Publishing AG.
McGuire, T. G. (2000). Physician agency. In A. J. Culyer & J. P. Newhouse (Eds.), Handbook of
health economics (Vol. 1, pp. 461–​536). Amsterdam: Elsevier.
Nelson, E. C., Batalden, P. B., Godfrey, M. M., & Lazar, J. S. (2011). Value by design:
Developing clinical microsystems to achieve organizational excellence. Hoboken, NJ: John
Wiley & Sons.
Ozawa, S., Mirelman, A., Stack, M. L., Walker, D., & Levine, O. S. (2012). Cost-​effectiveness
and economic benefits of vaccines in low-​and middle-​income countries: A systematic review.
Vaccine, 31(1), 96–​108. doi:10.1016/​j.vaccine.2012.10.103
Padula, W. V., Lee, K. K. H., & Pronovost, P. J. (2017). Using economic evaluation to illustrate value
of care for improving patient safety and quality: Choosing the right method. Journal of Patient
Safety, 17(6), e568–​e574. doi:10.1097/​PTS.0000000000000410
Padula, W. V., & Sculpher, M. (2021). Ideas about resourcing health care in the United
States: Can economic evaluation achieve meaningful use? Annals of Internal Medicine,
174(1), 80–​85. doi:10.7326/​m20-​1234
Porter, M. E., & Advantage, C. (1985). Creating and sustaining superior performance.
Competitive Advantage, 167, 167–​206.
Seto, K., Marra, F., Raymakers, A., & Marra, C. A. (2012). The cost effectiveness of human
papillomavirus vaccines: A systematic review. Drugs, 72(5), 715–​ 743. doi:10.2165/​
11599470-​000000000-​00000
 3.1.6 Translating value into resource allocation 155

Sullivan, S. D., Mauskopf, J. A., Augustovski, F., Caro, J. J., Lee, K. M., Minchin, M., . . . Shau, W.-​
Y. (2014). Budget impact analysis—​principles of good practice: Report of the ISPOR 2012
Budget Impact Analysis Good Practice II Task Force. Value in Health, 17(1), 5–​14.
Tahamtan, A., Charostad, J., Hoseini Shokouh, S. J., & Barati, M. (2017). An overview of his-
tory, evolution, and manufacturing of various generations of vaccines. Journal of Archives in
Military Medicine, 5(3), e12315. doi:10.5812/​jamm.12315
Teerawattananon, Y., & Tritasavit, N. (2015). A learning experience from price negotiations for vac-
cines. Vaccine, 33(Suppl 1), A11–​A12. doi:10.1016/​j.vaccine.2014.12.050
Teerawattananon, Y., Tritasavit, N., Suchonwanich, N., & Kingkaew, P. (2014). The use of eco-
nomic evaluation for guiding the pharmaceutical reimbursement list in Thailand. Zeitschrift
fur Evidenz, Fortbildung und Qualitat im Gesundheitswesen, 108(7), 397–​404. doi:10.1016/​
j.zefq.2014.06.017
Then, V., Schober, C., Rauscher, O., & Kehl, K. (2018). Social return on investment anal-
ysis: Measuring the impact of social Investment. Cham: Springer.
Verguet, S., Kim, J. J., & Jamison, D. T. (2016). Extended cost-​effectiveness analysis for health
policy assessment: A tutorial. PharmacoEconomics, 34(9), 913–​923.
World Health Organization. (2011). Rubella vaccines: WHO position paper. Weekly
Epidemiological Record, 86(29), 301–​316. https://​pub​med.ncbi.nlm.nih.gov/​21766​537/​
3.2
Defining the scope and study design
of cost-​effectiveness analysis
Joseph F. Levy and Charles E. Phelps

3.2.1 Defining the scope of an economic

evaluation study

3.2.1.1 Building from basics

The scope of any analysis of the value of a vaccine begins with the incentives
for individuals in a defined population to want to be vaccinated. In the purest
economic terms, wanting to be vaccinated means that one’s willingness to pay
for the vaccine exceeds the cost that the person confronts, both monetary and
personal (time, pain, fear of vaccines, etc.). Here, standard economic analysis
of a “representative single consumer” comes into play, using standard models
of human choices in healthcare (Phelps, 2017).
However, as Chapter 1.2 emphasized, vaccines (unlike almost all other
healthcare interventions) have “spillover” benefits to others in the society who
surround each “representative individual” (Phelps, 2017). While there are some
“private” benefits to vaccination (reduction of disease risk, with attendant re-
ductions in lost earnings, and even lower risk of death), there are also “public”
benefits—​the most important of which is increased protection for others in the
community (“herd immunity”). Therefore, analysis of vaccination programs
must include “everybody” in a defined population, not just those who receive the
vaccine. This forms the first basis for establishing the proper scope of analysis.

3.2.1.2 The “adoption” question

Now consider a typical public health decision maker with responsibility over
some specific defined population—​a county, state, or nation, or a subdivision,

Joseph F. Levy and Charles E. Phelps, Defining the scope and study design of cost-effectiveness analysis In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0015
 3.2.1 Defining the scope of an economic evaluation study 157

district, division, state, or nation. The relevant “local” decision focuses on the
following question:

If a vaccine against disease X existed that had certain specific attributes (efficacy,
number of injections required, side effects, cold chain storage demands, fit with
existing vaccine schedules in target communities, and others), would it be cost-​
effective to deploy that vaccine in “my community”?

This analysis, if properly done, should account for the public good aspects of
the vaccine among the community’s citizens. If the disease is tetanus or rabies,
there is essentially no public good issue, since the disease does not transmit
from person to person. But for diseases that are spread from person to person,
spillover effects are essential components of the analysis.
Just as local officials must consider spillover effects of vaccination programs
among individuals in the local community, so too must national officials con-
sider spillovers between local settings (counties, districts, shires). The “private”
incentives of the director of public health in a single county or district will not
fully consider the benefits conferred on surrounding counties or districts.
The same is true even as we aggregate from regional to national levels:
vaccination levels in country X provide spillover benefits to nations around
the world, particularly in a world with regular international travel by air
and sea. The coronavirus disease 2019 (COVID-​19) pandemic of 2020 high-
lighted this issue, since international travel rapidly spread the virus from
country to country, creating a worldwide pandemic in the space of weeks.
When modeling the benefits of vaccines, local health officers will likely
make decisions based on the interests and financial capacity of their own de-
fined populations, and that is necessary, but not sufficient, in understanding
the full value of vaccines for contagious diseases. Somebody—​the public
health officer at a higher level of government—​must also evaluate the bene-
fits from a wider perspective. Eventually, for highly contagious diseases, in-
ternational modeling of the disease, and then international coordination and
cooperation, are necessary to fully capture the benefits of vaccines on a world-
wide basis.
An extensive software program is available for free to assist in dealing with
these types of questions—​the SMART Vaccines software from the National
Academy of Medicine.1 While originally designed to prioritize vaccine

1 See https://​nap.nation​alac​adem​ies.org/​smartv​acci​nes/​to access the software and download all three

reports about the project.

158 Defining the scope and study design of cost-effectiveness analysis

development choices, it readily helps guide decisions about adoption, with the
ability to define various populations at all levels of analysis from “local” to “na-
tional” or even “super state” combinations of nations. Based on multi-​attribute
utility theory, it allows users to choose among 29 possible attributes to describe
vaccines (and to add their own if desired) and to specify value weights on the
attributes chosen for inclusion in the analysis.

3.2.1.3 The investment decision

We now turn to the scope of analysis that logically comes first: is it worth
trying to develop the vaccine in the first place? Here, the investment deci-
sion basically adds up the potential benefit across all potential target popu-
lations, and then asks if it is worth taking the risk of investing in a vaccine,
taking all the potential worldwide benefits into account as summarized by
answers to all the “community” questions.
This is in some sense a business planning decision for vaccine manufac-
turers, adding up the presumptive demands for their product from around the
world. Indeed, manufacturers can use the SMART Vaccines software to help
assemble this information in a coherent fashion, and to model how different
vaccine specifications (attributes) might alter the demand. However, financial
return to developing vaccines can be more complicated than other pharma-
ceuticals as various nongovernmental organizations (NGOs) often participate
in collective purchase agreements for vaccines (assuring a large quantity of
sales in return for lower prices). This is discussed in Chapter 1.3.

3.2.1.4 Alternatives to vaccination

When thinking about the value of developing a new vaccine, one must al-
ways consider the alternatives available to deal with the disease in question.
These fall into two main categories: treatments of the disease and non-​vaccine
methods to reduce disease spread. These issues affect both the desired scope
of analysis and measures of value of vaccines.

3.2.1.5 Herd immunity and disease eradication

Many alternatives exist to deal with contagious diseases, but one important
issue differentiates contagious diseases from other health conditions, and
 3.2.1 Defining the scope of an economic evaluation study 159

another related issue differentiates vaccines from other alternatives. These

two concepts are herd immunity and disease eradication, respectively.
Standard economic analyses of vaccine programs invariably emphasize the
difference between private gain from immunization and the public good that
also comes with it. The private gain is the reduction in or elimination of the
chances of getting the disease in question, saving the person the aches and
pains of the disease, the lost time from work or school, and the medical costs
that might be needed to treat the condition. But getting the vaccination also
confers a small benefit to others in the community—​their chances of getting
the contagious disease fall, perhaps only by a bit, but this applies to many
people. When enough people in a community become immune to the di-
sease, any new introduction of an imported case will be unable to lead to a
sustained epidemic. They have achieved “herd immunity.” This issue affects
the proper scope of analysis of vaccines.
Eradication of a disease takes place when the disease has no carriers in the
population. Eradication has almost been completed for polio and the last
known case of smallpox occurred in 1977. In some cases, vaccination alone
can achieve total disease eradication.2

3.2.1.6 Treatment

The most obvious alternative to vaccination is medical treatment, but often,

treatment and prevention go hand in hand as public health policies. They are
not in opposition to one another. Some diseases have wonderful vaccines but
no effective treatments, for example, polio. Others have highly effective ther-
apies, but no viable vaccines, including HIV and gonorrhea. Some diseases
have both, and disease control strategies use them in concert to combat the di-
sease. Examples include tuberculosis, influenza, pneumococcal pneumonia,
and others.
Comparing treatments with vaccines is a relatively straightforward task,
with vaccines reducing the probability of disease (and hence the risk of death
in some cases) and treatments reducing the consequences (severity of illness,
costs of treatment, and, in some cases, the risk of death). Such comparisons
must include not only the issues of medical costs but the value of reducing the
illness burden on people, either by reducing the probability of disease or the
consequences of contracting it.

2 A counterexample is hepatitis C, where the reproduction rate, R , is so low that eradication through
0
treatment of infected individuals can in concept eradicate the disease.
160 Defining the scope and study design of cost-effectiveness analysis

3.2.1.7 Social interventions

Social interventions are the earliest known methods to control infectious

diseases. The ancient book of Leviticus has numerous passages devoted to
the isolation (and then possible readmission to society) of lepers. When a
novel pathogen appears in the modern world, often neither vaccines nor
treatments are available, as evidenced (for example) in the Medieval Black
Plague, the 1918 “Spanish Flu” pandemic, and the 2020 COVID-​19 pan-
demic. These methods include forced quarantine,3 self-​isolation, physical
distancing, face-​mask wearing,4 swimming pool closures (for polio), pro-
hibitions on public gatherings, and similar restrictions on behavior. These
inevitably cause economic disruption, sometimes of major magnitude, and
these economic losses highlight the potential value of vaccines. In general,
social interventions provide only stop-​gap protections until vaccines and/​or
effective treatments emerge.

3.2.1.8 Vector control

Vector control will commonly enter consideration for any disease that has
other animal species involved, including mosquitos, snails, and ticks, and
zoonotic diseases that are transmitted directly from animals to humans (in-
cluding rabies, hantavirus, anthrax, and numerous diseases transmitted
through “wet” markets for wild animals that are desired for various alleged
exotic properties). In all such cases, control of the vector (or zoonotic carrier)
is an important alternative to vaccination.

3.2.1.9 Public sanitation

Simple measures of public sanitation are often important (and sometimes,

sufficient) to control infectious diseases. Clean drinking water is essential
for public health, often requiring investment in waste disposal systems that
are distinctly separated from water supplies. Better housing, sanitation, and

3 “Typhoid” Mary Mallon, an American woman, was forcibly quarantined twice over decades after she

was identified as an asymptomatic carrier of typhoid fever. A cook by occupation, she was known to have
infected 53 people, three of whom died. After release from quarantine, she worked as a cook using false
names, and some believe that up to 50 deaths were caused by her behavior.
4 As in the previous 1918 “Spanish Flu” outbreak, mandatory mask-​wearing became highly politicized in

some countries in 2020.

 3.2.1 Defining the scope of an economic evaluation study 161

solid waste management are all benefits of modern public health practice.
Investments in basic public health strategies should always be considered as
complements if not alternatives to vaccination programs where applicable.

3.2.1.10 Comparators in cost-​effectiveness analysis

We now come to the final step in this introduction to evaluation of vaccines.

Every cost-​effectiveness analysis model compares the incremental value of the
intervention to the incremental cost. The key question, then, is “Compared
to what?” In some diseases, the answer is straightforward. For many can-
cers, the answer will be “state-​of-​the-​art” combinations of chemotherapy,
radiation therapy, and surgery. For a badly arthritic knee, the comparison is
continued pain medication (perhaps augmented by physical therapy) versus
joint replacement. In some cases, with no effective treatments or vaccines,
the comparison may be to “doing nothing.” Finally, as we have seen earlier,
for vaccines, the issues of choosing the proper alternative to vaccines can be
much more complicated, particularly in cases of vector-​borne diseases.

3.2.1.10.1 For adoption

For decisions about adopting a new vaccine, the obvious choice is to use the
most cost-​effective of existing and available alternatives. The comparison
should not be something that doesn’t exist yet, perhaps such as another vac-
cine in phase I trials or a new medication touted by the company that makes it
(but without clinical trial data to support it).
The comparator must deal with the same disease. Comparing a poten-
tial zika vaccine with existing polio vaccines has no meaning. In some cases
(such as zika), no meaningful preventive intervention may exist, which means
that the ongoing costs of treatment in the relevant community are the only
issue. Of course, in diseases with potential fatalities, their possible preven-
tion also has major economic value, and the quality-​adjusted life years saved
from preventing premature deaths may be the most important part of a cost-​
effectiveness analysis for many infectious diseases.
As appropriate, non-​medical interventions such as vector control or be-
havioral control are appropriate comparators for vaccines. Malaria offers the
useful example where home screening, bed netting, mosquito repellants, re-
duction of standing water in the community, and widespread application of
162 Defining the scope and study design of cost-effectiveness analysis

pesticides are viable alternatives to vaccines, and in some cases, would be the
proper comparator.

3.2.1.10.2 For research and development decisions

Decisions about investing in research and development for new vaccines have
similar dimensions, but innovators will likely also want to consider what they
know of other innovations underway in parallel, how far along they are in de-
velopment, and (as far as is known) their likelihood of scientific success. These
are relatively straightforward business activities that will rely not only on the
progress of the vaccine in question but also “corporate intelligence” on other
alternatives.

3.2.2 Define the perspective impacted by a

vaccine program

As introduced in Chapter 2.3, all decisions depend on the perspective of the

decision maker. Therefore, presenting an economic model with respect to the
primary perspective of interest is required to undertake an economic evalua-
tion for a vaccine program. Here, we review perspectives again with respect
to how these options impact models. The choice of a third-​party payer’s per-
spective, for instance, should align with the incentives of the intended deci-
sion maker who will utilize the results of the model. This is a methodological
choice that establishes the viewpoint of the analysis —​and answers the ques-
tion regarding which costs and outcomes should be considered in the model.
Consider first a narrow viewpoint, that of a single insurer. For a vaccine
program, an insurer will pay the costs of the vaccine, and the costs (or avoided
costs) of future disease impacted by the decision. At the same time, the in-
surer will accumulate the health benefits (life years, quality-​adjusted life years,
disability-​adjusted life years, burden avoided, etc.) of their insurance plan
members. To that insurer, these are the only relevant costs and effects to ex-
amine, as these are relevant to their covered population. So, using this payer’s
perspective, if a patient were to switch insurance, the cost and benefits after
switching would no longer impact the initial insurer. For this reason, narrow
perspectives like a single insurer, or even a single NGO or municipality, are
less frequently considered in the vaccine space, as vaccines programs tend to
have wide and long-​lasting impacts on entire populations.
 3.2.2 Define the perspective impacted by a vaccine program 163

The most common payers’ perspectives considered for vaccine economic

evaluations are those of (1) the healthcare sector, an amalgam of all the payers
who remunerate providers for formal healthcare services and includes all rel-
evant health effects to an entire population; and (2) the societal payers’ perspec-
tive, which includes the healthcare sector costs and effects as well as broader
costs beyond formal health services, informal healthcare services, and even
beyond the healthcare sector that may be impacted by treatment.

3.2.2.1 The healthcare sector payers’ perspective

The healthcare sector payers’ perspective considers costs for direct health-
care that would emanate from the decision to vaccinate (or alternatives). This
would include the costs of the vaccine itself and vaccine administration costs,
costs of potential side effects, and costs associated with treatment of the di-
sease that patients may incur or that may be avoided as a result of the vaccine
program. The healthcare sector payers’ perspective is agnostic about who is
paying the costs; thus, it would include patients’ out-​of-​pocket costs, costs to
an NGO of running the vaccine distribution, costs to a private insurance firm
to pay for treatment in the future, and costs to the government if it provides
state-​funded health medical services, and so on.

3.2.2.2 The societal perspective

The societal perspective is an even broader perspective that will consider non-​
healthcare-​related costs and effects that may emanate from the decision of
interest. Under this perspective, costs of any kind and to anyone that will em-
anate from the treatment decisions are considered. Generally, these include
the informal healthcare sector costs related to treatment, such as patients’ lost
work time, caregivers’ lost leisure/​work time, and transportation costs associ-
ated with treatment, and non-​healthcare sector costs, such as a patient losing
work and productivity due to illness, or loss of consumption of goods (which
are a benefit to society) due to mortality differences across treatment strat-
egies. Other non-​healthcare sector costs relevant for vaccine programs would
be the positive spillovers effects, such as improved educational attainment due
to pediatric vaccine programs. The positive societal costs emanating from that
additional education attainment, for example, should be considered as costs
in a societal perspective model.
164 Defining the scope and study design of cost-effectiveness analysis

3.2.2.3 Which perspective should we use?

The Second Panel on Cost-​Effectiveness in Health and Medicine advocates

conducting cost-​effectiveness analysis from both the healthcare sector payers’
and societal payers’ perspective (Sanders et al., 2016). As societal perspective
encompasses all healthcare sector costs and more, redoing the model with
societal-​only costs removed is not complicated. Further, highlighting the po-
tential difference in results from the model using one approach may have par-
ticular interest for decision makers who weigh non-​healthcare sector costs
differently from others.

3.2.3 Framing the time horizon for a vaccine program

As first discussed in Chapter 2.3, time horizons of economic evaluations

should be of sufficient length to capture all costs and effects that can be attrib-
uted to the decision being considered. For certain vaccines, such as seasonal
flu, the effect (and thus the impact) of the disease and or immunity is short, so
the benefits and costs possibly emanating from the decision will also be short
too, perhaps 1 year after the decision about the program. So, a time horizon of
1 year would be reasonable to accumulate costs and effects post treatment, as
we would not expect the strategies under consideration (e.g., to provide a flu
vaccine or not) to produce additional differential impacts far beyond 1 year.
However, many vaccines impact mortality and/​or health outcomes long
in the future, sometimes over the person’s remaining life. It would be pref-
erable to track costs and health outcomes over this remaining lifetime for all
individuals. In practice, however, we only observe costs and outcomes for
individuals over a short time window. To extrapolate this information over
the individual’s lifetime, it is common to resort to economic modeling. In
modeling studies, this extrapolation of costs and health effects is done by al-
lowing the model to run until all patients (e.g., individuals or a cohort of
individuals) expire. This hypothetical ideal is frequently juxtaposed with the
realities of data available to inform decision-​making; for instance, data for
a new vaccine cannot, with certainty, predict the long-​term costs and out-
comes. Balancing the need for long-​term results to make efficient allocation
decisions with a lack of, or unreliable, data to inform these outcomes pre-
sents a challenge.
It is becoming more common to apply modeling approaches that use ob-
served data to project beyond the period of data collection. Extrapolation
 3.2.4 Role of modeling and applicability of data inputs 165

of survival or time-​to-​event data is an important and growing field in eco-

nomic evaluations. It will often be necessary to extrapolate treatment efficacy
estimates beyond what has been observed (Latimer, 2013). Indeed, there is
growing consensus that analyses should consider various extrapolation tech-
niques. A standard approach is to present the model results using different ex-
trapolation techniques to assess robustness of model results to these different
assumptions.

3.2.4 Role of modeling and applicability of data inputs

Decision-​analytic models are to decision makers what animal models are to

a bench scientist. They can be used to construct counterfactual worlds and
test the impacts of alternative interventions. They can also help overcome
several challenges with the estimation of the impacts of health interven-
tions, by systematically combining the best available data (e.g., epidemic,
clinical, economic) from multiple sources (e.g., individual clinical trials,
pooled results of multiple clinical trials, i.e., meta-​analysis, observational
studies, or primary data collected), to permit comparisons of the current
and future epidemiological, clinical, and economic impacts of multiple al-
ternative interventions or strategies, and for different population subgroups.
They are particularly useful in situations where there are multiple strategies
that need to be evaluated, when strategies vary in “intensity” of implemen-
tation, when strategies combine multiple interventions, and, importantly,
when it is impractical and costly to evaluate strategies within a single ran-
domized controlled trial. Even in situations where randomized controlled
trials can be conducted, models can be useful in extrapolating the effect
sizes observed the short time frame of the randomized controlled trial over
a longer time horizon, such as a lifetime.
Decision models are particularly relevant in the evaluation of vaccine
interventions, where the benefits can be long-​lasting, and where the external-
ities (mostly positive) are considerable. A common critical input parameter
to a vaccine economic evaluation is the hazard ratio of infection rate among
those vaccinated (compared to the unvaccinated). Ideally, estimates from
multiple phase III clinical trial studies would inform this hazard ratio and
the estimates would be weighted using meta-​analytic methods. Nonetheless,
many countries may only necessitate one large individual study of a popula-
tion for targeted vaccine use in order to obtain regulatory approval; however,
there remains the question as to whether information conducted between
166 Defining the scope and study design of cost-effectiveness analysis

countries is exchangeable. Thus, external validity remains an important con-

sideration for the interpretation of results between countries. Overall, the in-
fluence of results from alternate communities should be explicitly considered
for decision making.
For example, using results from a different context, say, the hazard ratio of
the efficacy of a vaccine in a phase III trial conducted in a developed country,
may not be the perfect analogue to a similar trial conducted in a low-​or
middle-​income setting. This could be due to several reasons, including differ-
ences in adherence or the prevalence of comorbid conditions across these set-
tings. Transparency about the limitations of applying data from one context
into another is therefore critical. This is why conducting sensitivity analyses
(discussed in detail in Chapter 3.5) is a critical topic in economic evaluation,
as it allows the modeler and decision maker to probe how one assumption
(that may not be the perfect analogue) can potentially impact the recom-
mended decisions by the model. Careful consideration about the inputs from
other studies and context applies to cost and effects estimates as well.

References
Latimer, N. R. (2013). Survival analysis for economic evaluations alongside clinical trials—​
extrapolation with patient-​level data: Inconsistencies, limitations, and a practical guide.
Medical Decision Making, 33(6), 743–​754. doi:10.1177/​0272989x12472398
Mutapi, F., Billingsley, P. F., & Secor, W. E. (2013). Infection and treatment immuniza-
tions for successful parasite vaccines. Trends in Parasitology, 29(3), 135–​141. doi:10.1016/​
j.pt.2013.01.003
Phelps, C. E. (2017). The demand for medical care: Conceptual framework. In Health economics
(6th ed., pp. 82–​111). New York, NY: Routledge Press.
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M., . . .
Ganiats, T. G. (2016). Recommendations for conduct, methodological practices, and re-
porting of cost-​effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and
Medicine. JAMA, 316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
3.3
Parameter estimation
Emmanuel F. Drabo and David W. Dowdy

3.3.1 Introduction

The quality of an economic evaluation is only as good as the data that com-
prises it. Therefore, it is important to accurately translate data from primary
and secondary resources in order to supply a vaccine model with appropriate
estimates of probability, cost, and effectiveness. Investigators must typically
estimate both a point estimate and an uncertainty distribution for each pa-
rameter. In order of quality, these estimates generally derive from empirical
data, the scientific literature, or assumptions (e.g., “expert opinion”). Most ec-
onomic evaluations will additionally require some parameter estimates based
on assumption (i.e., where empiric data or scientific literature do not exist); it
is therefore important to evaluate the influence of all parameter estimates on
the primary results and conclusions of any economic evaluation.
The materials covered in this chapter are intended to provide a strong
foundation for students interested in conducting basic, yet highly impactful
work in vaccine economics. There are more advanced methods available to
handle the topics covered in this chapter that are beyond the scope of this text.
Readers interested in these more advanced methodologies are encouraged to
undertake more advanced study or collaborate with peers with such advanced
training. Several reference textbooks in the Handbooks in Health Economic
Evaluation series (Briggs, Sculpher, & Claxton, 2006; Glick, Doshi, Sonnad, &
Polsky, 2014) are good starting points.

3.3.2 Ways to derive model input parameters

Parameters for vaccine economics applications should be estimates from high-​

quality data wherever possible. The “gold standard” for parameter estimation

Emmanuel F. Drabo and David W. Dowdy, Parameter estimation In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0016
168 Parameter estimation

is to collect all data empirically in the context of a combined effectiveness–​

economics study of primary or secondary data, such as a clinical trial or ob-
servational study. Ideally, parameters used in vaccine economics model—​or
any health economics model—​would be estimated from carefully designed
randomized controlled trials (RCTs).
For example, in estimating the cost-​effectiveness of Haemophilus influenzae
type b vaccine, one could theoretically perform a cluster-​randomized trial of
vaccine effectiveness with long-​term follow-​up while also directly collecting
data on costs of implementation, delivery, and sustainability. Unfortunately,
such studies are typically large and resource intensive. In addition, it is also
not always possible to track all necessary parameters within a single reported
RCT. Simply put, the authors of RCT reports seldom think, “I bet that an
economist is going to want to put together a health economic model of this
intervention.” Even if they did, clinical trial reports would still differ ac-
cording to specific research questions, types of health states, study popula-
tion, effectiveness measures, time step, and so on.
Furthermore, the costs of delivering a vaccine in the context of a clinical
trial may be very different from those of programmatic delivery. As such, this
“gold standard” is rarely—​if ever—​fully achieved. Hence, it is often the re-
sponsibility of the economist to find data in the literature and translate those
data into meaningful parameters to inform vaccine economics models.

3.3.2.1 Types of parameters that may need to be estimated

Depending on the type and complexity of model being considered, dif-

ferent types of parameters may need to be estimated. For example, for a de-
cision tree model such as the one in our hepatitis B vaccination example (see
Chapter 3.7), we would need to estimate probabilities of changes in the risk
of contracting H. influenzae type b with or without a vaccine, which relies
heavily upon the efficacy of the vaccine. For a Markov model, parameters that
are typically estimated for each health state are:

• Initial distributions of the cohort, costs, and health outcomes.

• Costs and effectiveness per cycle.
• Transition probabilities.
• Transition costs.
• Fixed probabilities in non-​Markov nodes of a semi-​Markov model.
• Fixed costs (e.g., infrastructure costs) and fixed effectiveness measures
(e.g., payoffs).
 3.3.2 Ways to derive model input parameters 169

Consider the example of hepatitis B vaccination (more details are provided

in Chapter 4.6), the relevant parameters include disease state transition prob-
abilities, disease state utilities, costs, and pharmacological effectiveness meas-
ures. It is also important to consider, as noted above, that not all costs and
effectiveness outcomes are time dependent, and not all health state transition
models are full Markov models (e.g., semi-​Markov models).

3.3.2.2 Sources of data for parameter estimation

Parameters should be estimated from empiric, scientific data, whenever pos-

sible. These empiric data can come from a single source (albeit rarely) or mul-
tiple sources. Indeed, data from multiple sources are typically combined to
inform the estimation of model parameters for a particular decision problem
such as the decision whether to introduce a new vaccine.
Hence, when relying on existing empirical data or the scientific literature
to estimate parameters, it is important to understand and appreciate whether
the evidence generated emanates from a single study or is synthesized from
multiple studies.

3.3.2.2.1 E  stimating parameters using existing empirical data from a

single study
On very rare occasions, and most likely for simpler decision problems, one
may identify a high-​quality peer-​reviewed journal article reporting precisely
the type of information needed to estimate one, multiple, or all parameters in
a health economic model. In this simplest scenario, the challenge is to utilize
that single source of data to estimate the model’s parameters.
Suppose all the evidence is available in the form of summary results of a
single published study. In that case, we can use these data “as they are” to esti-
mate the relevant parameter(s) for the model. For example, a single published
study of a single (three-​arm) trial can provide head-​to-​head effectiveness and
cost data for a decision model assessing the value of three coronavirus disease
2019 (COVID-​19) vaccines. There are currently single-​shot and double-​shot
vaccines, as well as vaccines structured using antibodies or messenger RNA
(mRNA) delivery. Because the resulting measures of relative effectiveness
(e.g., odds ratios, log-​odds ratios, or relative risk ratios) between the arms of
the trial are inherently correlated, parameters estimated from these data will
also be correlated. So, it is critical to explicitly model these correlations when-
ever possible, in order to best assess uncertainty in the model, and correctly
estimate each vaccine’s expected costs and benefits.
170 Parameter estimation

Occasionally, individual patient data may be available to aid in estimating

single or multiple parameters in a model. A reanalysis of these data can help
derive appropriate model input parameters and explore and answer clinical
and economic research questions in the most flexible manner. For example,
we can use the individual-​level data from a trial comparing the efficacy, safety,
costs, and health outcomes of two vaccines to estimate the parameters for a
vaccine economics model (e.g., cost, effectiveness) by using a range of statis-
tical models. It is also easy to construct the correlation structure between the
correlated input model parameters.
The availability of individual-​level data can also aid in the analysis, or re-
analysis, of time-​to-​event data, thereby allowing the extrapolation of a trial’s
results (e.g., fatal and non-​fatal events) beyond its short observation period. It
is common to use parametric distributions to model the outcomes of interest,
which are typically governed by a combination of two or more correlated an-
cillary parameters, such as the mean and the standard deviation in the case of
the Weibull distribution. Other popular examples of parametric distributions
include the log-​logistic, and generalized gamma distributions for the analysis
of time-​to-​event outcomes (Collett, 2015).

3.3.2.2.2 E stimating parameters by synthesizing existing data

from multiple studies
In most situations, the evidence base is represented by multiple studies,
which may report results on the same parameter. In this case, it is common
to combine these into a single quantitative estimate for use in health eco-
nomic models. The exercise of combining multiple values for a parameter,
derived from different sources, must be conducted with great care in order
to minimize bias. Approaches for synthesizing and combining results from
multiple studies rely on meta-​analytic methods such as meta-​analysis, meta-​
regression, and mixed treatment comparison. These approaches are intended
to help eliminate, or at least minimize, bias.
Bias can be defined in multiple ways but can be more generally character-
ized as a systematic tendency that prevents unprejudiced consideration of
the available evidence about a question (Pannucci & Wilkins, 2010). There
are different types of bias, including publication bias (i.e., the tendency for
studies with similar results to be published or unpublished, incomplete or se-
lective reporting of outcomes, etc.), truncation bias (i.e., tendency for studies
to be published in a briefer form with less details), time-​lag bias (i.e., tendency
for delayed publication of findings), language bias (i.e., tendency for studies
being more likely to be published in a particular language such as English),
 3.3.2 Ways to derive model input parameters 171

citation bias (i.e., tendency for certain research findings to be cited more or
less), selective outcome reporting bias (i.e., tendency for selective report or
non-​report of certain study outcomes), location bias (i.e., tendency for cer-
tain studies to be published in journals with different ease of access or levels of
indexing in standard databases), duplication bias (i.e., tendency for multiple
or duplicate publications), or database bias (i.e., tendency for some databases
to be are more likely to index certain languages or journals).

Meta-​analysis
It is recommended and common, in economic evaluation studies such as cost-​
effectiveness analysis, to use estimates of relative treatment effects derived
from a meta-​analysis of multiple studies (Gold, Siegel, Russell, & Weinstein,
1996; Neumann, Ganiats, Russell, Sanders, & Siegel, 2016). For example, in
their cost-​effectiveness analysis of routine and campaign use of typhoid Vi-​
conjugate vaccine in Gavi-​eligible countries, Bilcke et al. (2019) used meta-​
analytic methods to synthesize and combine the available evidence on typhoid
disease and its burden (e.g., hospital admission rates, typhoid treatment costs,
vaccine delivery costs, length of stay in hospital, duration of illness).
There are at least two commonly used types of meta-​analyses. The first is
the fixed-​effect meta-​analysis in which it is assumed that every study included
in the analysis is estimating the same statistic (most often an odds ratio), and
that a single common and fixed effect characterizes every study included in
the analysis, so that only within-​study variations can influence the uncer-
tainty in the results (Higgins et al., 2021). However, it is not uncommon to
also observe between-​study variations in the estimates of treatment effects
(e.g., heterogeneous treatment effects). This gives rise to a second type of
model, the random-​effects meta-​analysis. The random-​effects model assumes
that individual studies are estimating different treatment effects, but that
these effects derive from a common distribution with some measure of cen-
tral tendency such as the mean or median and some measure of dispersion,
such as the variance (Higgins et al., 2021). Bilcke et al. (2019) used random-​
effects models to estimate the probability that an infected patient is admitted
to the hospital, the length of stay (in days) in the hospital, the probability of
death after admission to the hospital for typhoid infection, and the duration
of illness in inpatient and outpatient settings. For a more in-​depth review of
meta-​analytic methods, the reader can refer to the articles by Rosenthal and
DiMatteo (2001) and Sutton and Higgins (2008).
While one of the critical tenets of good meta-​analysis is the consideration
of all the available evidence, it is clear that all evidence is not created equal.
172 Parameter estimation

Some studies are stronger than others. Doing a perfect meta-​analysis based
only on badly designed biased studies will still produce a biased combined
statistic. Analysts must therefore face the difficult task of assessing the quality
of each study, and account for it in the meta-​analysis. In addition to needing
to account for the quality of the evidence, analysts must also deal with other
related biases that may be present in each study, some of which are detectable
and addressable (Sutton, Song, Gilbody, & Abrams, 2000).
For example, study-​level features, such as the characteristics of partici-
pants, may contribute to between-​study variability in effects. In the meta-​
analytic framework, this can be accommodated through a meta-​regression
approach in which study-​level covariates are included in the meta-​analysis.
Rather than excluding weaker and low-​quality studies from a meta-​analysis,
it may be preferable to use the meta-​regression approach with all studies
included while controlling for study-​level covariates which reflect the meth-
odological quality of the studies. This approach can then permit the assess-
ment of the impact of study quality on the pooled effect size. Unfortunately,
it is often impossible to control for all biases, as there are various sources
and manifestations of bias across studies. For example, meta-​regression ap-
proaches a number of weaknesses that have been discussed elsewhere (see
Thompson & Higgins, 2002), including yielding significantly lower statis-
tical power over individual patient data methods, due to the use of mean
study-​level covariate values, and, importantly, potentially leading to “eco-
logical fallacy,” that is, an incorrect inference at the individual level of the
relationships observed at the aggregate variable level (Berlin, Santanna,
Schmid, Szczech, & Feldman, 2002; Lambert, Sutton, Abrams, & Jones,
2002; Piantadosi, Byar, & Green, 1988). However, when one has access to
individual-​level data from each study, the relationships between the param-
eter of interest and the individual covariates can be more rigorously exam-
ined across the pooled studies (Wakefield, 2008). When individual-​level
data are available for each study to permit a meta-​regression, such analysis
is called a mega-​analysis and is considered the “gold-​standard” in evidence
synthesis (Higgins, Whitehead, Turner, Omar, & Thompson, 2001; Sutton,
Abrams, Jones, Sheldon, & Song, 2000).
It is possible to incorporate between-​study variability in decision-​analytic
models, such as those for vaccination decision, whenever possible, and strat-
egies for doing so have been previously discussed extensively (Ades, Lu, &
Higgins, 2005). Intuitively, the approach consists of using the heteroge-
neity parameter from the meta-​regression to derive the parameters for the
decision-​analytic model. Hence, three types of parameters are derived: the
 3.3.2 Ways to derive model input parameters 173

random-​effect estimate of the treatment effect, the variability of the random-​

effect estimate, and a measure of heterogeneity.

Multivariate meta-​analysis model

Sometimes, multiple outcomes are of interest, such as costs and quality of life
measures. A multivariate meta-​analysis model can help with the joint estimation
of these outcomes, as well as any potential correlation between them. Similar
to the univariate meta-​analytic model, multivariate meta-​analysis models can
be fixed-​effects or random-​effects models. The random-​effects meta-​analysis
model is often used in applications due to its ability to capture both the within-​
and between-​study correlation structures of the multiple outcomes of interest.

Mixed treatment comparison

When comparing two alternative vaccines or vaccination strategies, say, in
term of their efficacy in preventing infections, we would ideally want data
on their head-​to-​head comparison for that efficacy outcome. Unfortunately,
RCTs rarely compare all relevant intervention strategies, head to head, and
most only compare the new intervention to placebo or standard of care.
Indeed, a common feature of the evidence base used to inform health eco-
nomic models is the absence of head-​to-​head trials comparing all relevant
comparators. This can pause various statistical challenges to estimation of, for
example, the relative efficacy of the two interventions.
When patient-​level data are available, these can be pooled across all studies
to permit the estimation of the relative effects of comparator interventions,
through a mega-​regression. However, mega-​regressions are rarely feasible.
Alternatively, one might naively be tempted to use the differences in effect
sizes across studies. That approach would only be valid if the studies included
in the meta-​analysis involved similar numbers and types of participants in
terms of characteristics and behaviors, and if the endpoints (clinical or ec-
onomic) were measured at the same time intervals. This wishful scenario is
also hardly the case in the real world. We must therefore resort to alternative
methods.
When multiple comparators are involved, more complex forms of meta-​
analytic evidence synthesis (indirect and mixed treatment comparison) are
needed to handle the comparisons and combinations of evidence on multiple
surrogate or intermediary endpoints (Baker, 2006). These complex methods,
sometimes referred to as multiparameter evidence synthesis, include indirect
treatment comparisons and network meta-​analysis, which are simple exten-
sions of the pairwise meta-​analysis method, in which parameters are related
174 Parameter estimation

to one another by a definable structure. These models have been primarily

developed within the Bayesian framework, both for computational reasons,
and for its natural connection to decision making (see Ades et al., 2006; Dias,
Ades, Welton, Jansen, & Sutton, 2018; Hasselblad & McCrory, 1995; Lu &
Ades, 2004; Lumley, 2002).
It should be noted that the multiparameter evidence synthesis framework
relies on exactly the same assumption made under the pairwise meta-​analytic
method. Hence, depending on the analyst’s assumptions about the between-​
study heterogeneity, a fixed-​or random-​effects analysis may be conducted.

3.3.2.3 Types of input data available from the literature

Many types of inputs data may be available from the literature, including
probabilities or risk (e.g., probability of infection), rates (e.g., mortality rate
from chronic hepatitis B infection), relative risks (e.g., relative risk of infection
among vaccinated individuals, compared to unvaccinated susceptible indi-
viduals), odds ratios, risk differences, means, medians, and so on. Depending
on the specific vaccine economic modeling context, any of these input data
types may be relevant. Before we proceed with a more in-​depth overview of
each type of input data, it is worthwhile settling on some definitions.

3.3.2.3.1 Definitions
Probability is the numerical likelihood of occurrence of an event or outcome
for a single individual in a given time period, and ranges from a value of 0
to 1. The risk of occurrence of an event (e.g., infection after vaccination) is
the probability that the outcome will differ from an expected outcome. In that
sense, risk and probability refer to the same concept.
Proportion represents the number of individuals who experienced an event,
out of a population of interest; it is the number of people who had an event, di-
vided by the total number of individuals at risk of the event in the population.
Typically, the experience of the event is cumulative and the timing of when the
event occurred is any time in the past. Prevalence of a long-​duration disease
like hepatitis B is often expressed as a proportion. The denominator of a pro-
portion is always a count of a number of people.
In contrast, a rate is an instantaneous (or velocity) measure of the number
of events that occur per unit time (or person-​time), and ranges from a value
of 0 to infinity. So, hepatitis B incidence is a rate, that is, the number of new in-
dividuals who acquire hepatitis B infection (hepatitis B cases) during a given
 3.3.2 Ways to derive model input parameters 175

time period, divided by the total person-​time at risk of infection over the ob-
servation period. Similarly, the annual vaccination rate is the number of per-
sons vaccinated over the course of a year, divided by the number of persons at
risk of infection during that year.
Rates are typically scaled and can be expressed, for example, per 100
person-​years. They differ from probabilities in many ways, the most impor-
tant of which relates to the role of time. Specifically, in the calculation of a rate,
time is directly included in the denominator. In contrast, time is not included
in the denominator in the calculation of a probability. In the hepatitis B ex-
ample, an estimate of the probability would be the number of hepatitis B cases
divided by the total number of individuals at risk of infection over the relevant
observation period. The denominator of a rate is always expressed in people
units × time units most often person-​years.
To further illustrate the concept of rate, suppose that after following 36
people with diagnosed type 2 diabetes and acute hepatitis B infection for
3 years to observe new cases of chronic hepatitis B infection, we found that
exactly three, two, and one person(s) developed a chronic hepatitis B infec-
tion, exactly at the end of year 1, year 2, and year 3, respectively (so six people
in total). Hence exactly 30 people (36 − 6) spent the entire 3 years without
infection, and thus contributed 90 (30 × 3) person-​years. One person spent
3 years without infection and became infected exactly at the end of year 3; this
person thus contributed 3 (1 × 3) person-​years. Two persons spent 2 years
without infection and became infected exactly at the end of year 2, thus con-
tributing collectively 4 (2 × 2) person-​years. Finally, three persons spent ex-
actly a year without infection and became infected exactly at the end of year
1, thus contributing collectively 3 (3 × 1) person-​years. In total, individuals
in this sample contributed 100 (90 +​3 +​4 +​3) person-​years at risk of chronic
hepatitis B infection (i.e., the event). The incidence rate of compensated cir-
rhosis in this population is 6/​100 person-​years =​0.06 per person-​year, or 6 per
100 person-​years.
It is worthwhile noticing that, in this particular example, the annual inci-
dence rate is decreasing over time, as the rate in year 1 is 8 per 100 person-​years
(3/​36 person-​years); the rate in year 2 is 6 per 100 person-​years (2/​33 person-​
years), and the rate in year 3 is 3 per 100 person-​years (1/​31 person-​years).
A cumulative incidence is the fraction of individuals in a population who
experience an event over a specified time in a closed population, that is,
one in which no new individual enters, no individual is lost to follow-​up,
and there are no competing events. In our example above, the 1-​year cu-
mulative incidence of acute hepatitis B infection is 8.3% (3/​36 =​0.083); the
176 Parameter estimation

2-​year and 3-​year cumulative incidences are 13.9% (5/​36 =​0.139%) and
16.7% (6/​36 =​0.167), respectively. These calculated cumulative incidences
are proportions, not rates. Notice here that these proportions increase at a
decreasing rate, since the annual incidence rates are also declining in time.
Rates additionally possess several convenient mathematical properties that
probabilities do not. They can be added and subtracted for the same time in-
terval. They can also be multiplied or divided by a scalar, which may reflect,
for example, risk factors. Rates can additionally be divided by the number of
patients, or by time (Hunink, Glasziou, Siegel, & Weeks, 2001).
Furthermore, the rate varies with the distribution (i.e., the time) of the
event over the observation period; the probability does not. For example, sup-
pose that in a population of 100 individuals observed over a period of 4 years,
one person died of a vaccine-​preventable disease after the first year, another
person died after the second year, and another died after the third year, and
the other persons survived. So, a total of three persons died over this period.
The probability of death from the vaccine-​preventable disease over the 4-​year
period in this population is 0.03 (3/​100), and the death rate is 0.76 per 100
person-​years (3/​[100 +​99 +​98 +​97] =​3/​394, since 97 persons contributed
4 person-​years each, and three persons contributed 1, 2, and 3 person-​years
each, respectively). Now, suppose instead, that the same population was ob-
served over the same period of 4 years, but that all three deaths occurred at
the end of the first year. Then the probability of death is still 0.03, but the death
rate is now 0.77 per 100 person-​years (3/​[100 +​3 × 97] =​3/​391, since 97 per-
sons contributed 4 person-​years each, and the three deceased persons con-
tributed 1 person-​year each).
The odds of an event are the ratio of the probability that the event will
occur to the probability that it will not occur. If we denote by p the probability
of the event occurring, then the probability that the event will not occur is
1–p, since the event can either occur or not occur, and the occurrence and
non-​occurrence of the event are two mutually exclusive events, that is, they
cannot both occur at the same time for the same subject. The mathematical
expression of the odds of an event that occurs with probability p is:
Equation 3.3.1. Odds:

p
Odds =
1− p

For example, if the probability of progression from an acute hepatitis B infec-

tion to a chronic hepatitis B infection is 0.06 (6%), then the probability of not
 3.3.2 Ways to derive model input parameters 177

progressing to a chronic infection is 1 − 0.06 =​0.94 or 94%. Hence, the odds of

progressing to a chronic hepatitis B infection are 0.06/​0.94 =​0.064.
The odds can take values ranging from 0 to infinity. For example, when the
probability of occurrence of the event is 0, its odds are also 0 (0/​[1 − 0] =​0).
Similarly, when the event occurs with certainty (i.e., probability 1), its odds
are infinity. When the odds equal 1, it is referred to as even odds, since both
the occurrence and non-​occurrence of the event have an equal probability of
0.5. When the odds are larger than 1, say 5, it is common to hear people re-
ferring to it as 5 to 1. It is also worthwhile noting that when the probability of
occurrence of an event is low, as is the case in our hepatitis B example above,
the odds will be very similar to the probability.
The odds ratio is a statistic used to compare the odds of occurrence of an
event among the exposed and unexposed groups. It is particularly useful in
the case–​control design in which there is no follow-​up period, and the odds
ratio is the only measure of association that can be computed.
In contrast, in cohort-​type studies characterized by following exposure
groups to compare the incidence of an outcome, it possible to calculate both
a relative risk and an odds ratio. The relative risk is much easier to interpret
than the odds ratio. It is a direct comparison of risks across exposure groups in
terms of their ratio and is used to quantify the strength of association between
an observed outcome and the exposure.
Using the hypothetical data in Table 3.3.1, we can calculate both the odds
ratio and relative risk of the outcome for the two exposure groups. To calcu-
late the odds ratio, we need the odds for the outcome among the two exposure
groups (exposed, e, and unexposed, u). The odds for the outcome among the
exposed are:
Equation 3.3.2. Odds for exposed individuals:

pe
Oddse =
1 − pe

Table 3.3.1 Illustrative 2 × 2 table from a hypothetical cohort study

Outcome No outcome Total

Exposed a b a +​b
Unexposed c d c +​d
Total a +​c b +​d a +​b +​c +​d
178 Parameter estimation

a
where pe = a + b denotes the probability of the outcome among the exposed.
Similarly, the odds for the outcome among the unexposed are:
Equation 3.3.3. Odds for unexposed individuals:

pu
Oddsu =
1 − pu
c
where pu = c + d denotes the probability of the outcome among the unexposed.
The formula for the odds ratio is simply:
Equation 3.3.4. Odds ratio:

Oddse a × d
OR = =
Oddsu b × c

Notice that the denominator is the odds for the outcome in the unexposed
group, consistent with the convention.
The calculation of the relative risk ratio is also straightforward from the
data in Table 3.3.1. As the relative risk is the ratio of the probability for the
outcome among the two exposure groups, its formula is simple:
Equation 3.3.5. Relative risk ratio:

pe a × (c + d )
RRR = =
pu c × (a + b)

When the odds or risk of occurrence of the outcome is lower in the exposed
group compared to the unexposed group, the odds ratio and relative risk ratio
are less than 1. When the odds ratio or relative risk ratio is 1, the odds or risk
of the outcome do not differ across exposure groups. Finally, when the odds
or risk of the outcome are higher in the exposed group compared to the unex-
posed group, the ratios are above 1.
We can also calculate confidence intervals around these statistics. As neither
the odds ratio nor the relative risk ratio follows a normal distribution, they
need to be transformed before any confidence interval can be constructed.
The odds and relative risk ratios are more characterized by log-​normal dis-
tribution, so a logarithmic transformation is required to promote normality.
Because of this, the confidence interval for an odds or relative risk ratio in-
volves two steps. First, a confidence interval of the natural logarithm of the
statistic is calculated. Second the antilog of the upper and lower limits of the
 3.3.2 Ways to derive model input parameters 179

constructed confidence interval for the natural logarithm of the statistic are
calculated by taking their exponents, in order to derive the upper and lower
limits of the confidence interval for the statistic.
Formulas for calculating the confidence interval for the natural logarithm
of the odds ratio and relative risk ratios are given in the equations below:
Equation 3.3.6. Confidence interval of the natural logarithm of the
odds ratio:

) ± z 1 + 1 + 1 + 1
Ln(OR
a b c d

Equation 3.3.7. Confidence interval of the natural logarithm of the relative risk:

) ± z b / a + d / c
Ln(RR
a+b c +d

where Ln denotes the natural logarithm, and z denotes the standard score
for the normal distribution. For each statistic, the second step consists of
taking the exponent of the lower and upper limits of this constructed confi-
dence interval in these equations.
To illustrate these concepts, consider the following hypothetical 2 ×
2 table (Table 3.3.2) from a study by Massoudi and Mohit (2021), which
used the cohort design to calculate the odds and odds ratio of receiving the
2019 influenza vaccine, experiencing pulmonologist-​confirmed COVID-​
19 symptoms, and testing positive for COVID-​19, among healthcare
workers in a hospital setting in Iran. Table 3.3.2 presents summary counts
of the numbers of healthcare workers, by vaccination status (exposure)
and experience of COVID-​19 symptoms (outcome). The probabilities of
incident COVID-​19 related symptoms (outcome) among vaccinated (ex-
posed) and unvaccinated (unexposed) healthcare workers are 0.03 (i.e.,

Table 3.3.2 Association between vaccination status and COVID-​19 symptoms

Symptomatic Asymptomatic Total

Vaccinated 3 87 90
Unvaccinated 77 94 171
Total 80 181 261
180 Parameter estimation

3/​90) and 0.45 (77/​171), respectively. The corresponding odds of symp-

toms the two exposure groups are 0.03 (0.03/​[1 − 0.03]) and 0.82 (0.45/​[1 −
0.45]). The odds ratio for symptoms among vaccinated healthcare workers
relative to unvaccinated workers is 0.04 (0.03/​0.82). The 95% confidence
interval for the odds ratio is 0.01–​0.14, suggesting that the observed as-
sociation between vaccination and incidence of COVID-​19 symptoms
is statistically significant. Massoudi and Mohit (2021) also report asso-
ciations between the 2019 influenza vaccine exposure and confirmed
COVID-​19 cases.
For example, using data from Massoudi and Mohit (2021) (Table 3.3.2),
we can also calculate the relative risk of experiencing COVID-​19 symp-
toms among vaccinated and unvaccinated individuals; that relative risk is
0.07 (0.03/​0.45), with a 95% confidence interval of 0.02–​0.21, thus sug-
gesting the observed association was still statistically significant, sug-
gesting that vaccinated individuals have a significantly much lower risk
of experiencing symptoms, relative to their unvaccinated counterparts.
Notice that the relative risk is not too far apart from the odds ratio.
However, notice also that the implied treatment effect is different (larger in
this case) when measured by the relative risk, compared to the odds ratio.
This is because the comparison of risk ratios produces a stronger measure
of the potential association between the exposure and the outcome than
odds ratios do. In general, when the outcome of interest is relatively un-
common, its odds in the exposure group will be similar to its probability
in the same group. As a result, the odds ratio provides a relative measure of
the treatment effect for case–​control studies, as well as an estimate of the
risk ratio in the source population when the outcome is uncommon.
Despite the many attractive features of the relative risk ratio, it is worth-
while pointing out that the odds ratio has better statistical properties than
the relative risk ratio. For example, the odds ratio of experiencing COVID-​
19 symptoms is the inverse of the odds ratio of experiencing no symptoms.
This is not the case for the relative risk ratio: the relative risk of experien-
cing symptoms differs from the inverse of the relative risk of experiencing
no symptom. Because of these favorable statistical properties of odds ratios,
much of the data in the literature are reported as odds ratios.
Table 3.3.3 summarizes and describes some of the most commonly re-
ported statistics in the published literature, which may be relevant to the
estimation of parameters for vaccine economics models.
 3.3.2 Ways to derive model input parameters 181

Table 3.3.3 Statistics commonly reported in published scientific studies that are
relevant to vaccine economics models

Statistic Definition

Probability/​risk/​ Number of events ocurring inatime period

prevalence Number of persons at risk in the populationover that time period
Rate Number of events ocurring inatime period
Number of persons at risk in the population over that time period
Odds Probability of outcome
1 − Probability of outcome
Odds ratio Oddsof outcomeinexposed
Oddsof outcomeinunexposed
Relative risk /​risk Probability of outcome inexposed
ratio/​hazard ratio Probability of outcome inunexposeed
Risk difference Differencein probability of event among exposed and unexposed
Survival curve Number of people aliveat time (t −1) whoare aliveat time t
Mean Sumof all observations ina sample (or population)
Total number of observationsinthe sample (or population)

3.3.2.3.2 Types of input parameters needed in health

economics models
The odds ratio, relative risk (or risk ratio and hazard ratio), and risk difference
statistics in Table 3.3.3 are comparative statistics, as they compare risks or odds
of an outcome (e.g., infection) among two different population subgroups
(e.g., vaccinated versus unvaccinated). All other statistics in Table 3.3.3 are
non-​comparative statistics.
Inputs for health economic models require non-​comparative parameters.
For example, in the example of hepatitis B vaccination, we need the proba-
bility of acquiring an acute infection without vaccination. Hence, comparative
data such as odds ratios, relative risk, or risk difference must be transformed,
or combined with non-​comparative data to be informative in a vaccine eco-
nomics model. Furthermore, for the simplest types of models such as decision
tree or Markov models, we need parameters in the form of probabilities. In
more complex models (e.g., dynamic transmission models), rates might be
more desirable.
182 Parameter estimation

3.3.3 Data quality for parameter estimation

in vaccine economics applications

In practice, most parameter estimates can be judged according to two cri-

teria: quality of the underlying data (“internal validity”) and appropriate-
ness or generalizability to the analysis (“external validity”). This principle
holds whether or not the specific application relates to vaccine economics or
health economics more broadly. In evaluating the quality of underlying data,
a commonly used hierarchy may be applied from the principles of evidence-​
based medicine (Guyatt, Rennie, Meade, & Cook, 2015). These principles of
evidence-​based medicine include using all available evidence, not selectively
relying on just one source, avoiding bias when using observational sources, and
using formal evidence synthesis techniques. Certain study designs are better
equipped than others to minimize certain forms of biases. In the hierarchy of
evidence-​based medicine, often depicted as a pyramid, quality of evidence can
be assessed based on study design (listed from lowest to highest quality): case
series, case–​control studies, cohort studies, RCTs, systematic reviews, and
meta-​analyses. When considering two different RCT studies of vaccines, for
example, the analyst should first ask whether the two vaccines were studied in
a head-​to-​head trial. All things equal, priority should be given to evidence gen-
erated through head-​to-​head comparison trials. Study design is not the only
determinant of data quality, however; investigators must also assess studies—​
and their corresponding parameter estimates—​for their level of precision (e.g.,
sample size) and risk of bias.

3.3.4 Appropriateness of data to the analysis

In addition to underlying data quality, parameter values must also be

evaluated for their appropriateness to the analysis at hand. For example,
a high-​quality cluster randomized trial successfully delivered the hepa-
titis B vaccine to 97.6% of eligible neonates in Qidong, China (Qu et al.,
2014). However, applying this probability of vaccination in an economic
evaluation of a real-​world program in sub-​Saharan Africa would be in-
appropriate. Often, investigators must identify parameter estimates from
the scientific literature and consider how to adjust or modify these values
(and corresponding estimates of uncertainty) for their analysis. In making
any such modifications, it is important to be transparent about the
 3.3.5 Formatting data for analysis 183

assumptions made, so that analyses can be replicated and appropriately

judged by readers in terms of both quality and appropriateness of para­
meter estimates.

3.3.5 Data conversion into a format consistent with the

analytic structure of the model

One additional consideration is that parameter estimates from the scientific

literature are often not presented in the form required by an economic eval-
uation. For example, in the context of vaccine economics, some parameters
may be reported as incidence rates, but the analysis may involve a Markov
model with a finite time step (for example, 1 month)—​thereby requiring the
parameters to be estimated as a 1-​month probability (or “cumulative inci-
dence”). To the extent possible, investigators should attempt to structure
their analyses (e.g., age categories of hepatitis B infection risk) in a fashion
that reflects the underlying data structure. However, the natural history of
the disease, disease staging, or the clinically and economically meaningful
health states associated with the condition may not necessarily coincide with
how the data are reported in the literature. When this is the case, mathe-
matical conversions may be required. For example, a 6-​month probability
of infection may be what is reported in the literature. However, the vaccine
economics model might have a 1-​year cycle length, requiring the analyst
to extrapolate beyond the 6-​month estimate, since a 1-​year probability is
needed for the model. Furthermore, and as we have noted earlier, probabil-
ities cannot be manipulated as easily as other parameters such as rates; we
cannot simply multiply or divide them. For example, a 100% probability at
5 years does not mean a 20% probability at 1 year; likewise, a 30% probability
at 1 year does not mean a 120% probability at 4 years, and fortunately so,
because if it did, a fundamental property of probabilities—​that they cannot
exceed 1—​would be violated.
Data from the literature may also be reported as incidence rates or cumula-
tive incidence. To calculate the probability of an event from incidence rates or
cumulative incidence requires information on the duration of the follow-​up. If
the duration of the follow-​up is known, the probability of the event occurring
in each time step is simply the cumulative incidence per time step. Going from
the incidence rate to the probability involves, however, more assumptions and
steps, further described below.
184 Parameter estimation

3.3.5.1 Translating rates to probabilities

If the incidence rate (rate per unit time) is constant, the cumulative incidence
follows what we call an exponential decay. In that case, we can convert the con-
stant incidence rate into a probability, using the exponential decay formula:
Equation 3.3.8. Rate of decay:

pt = 1 − e − rt

where pt is the probability of remaining event free at time t, r is the inci-

dence rate, and t is the duration of time, with the same unit of time used in
both. Thus, for example, to convert a constant incidence rate of 0.4/​year to
a 1-​month (0.083 year) cumulative probability, one could solve for p0.083 =​
1 –​ e−(0.4)*(0.083) =​0.0327. This equals the 1-​month probability of an event that
occurs at an underlying constant incidence rate of 0.4/​year.

3.3.5.2 Translating probabilities to rates

We can also translate rates into probabilities. To do this, it suffices to do some

rearrangements of the exponential decay formula so that e − rt = 1 − pt .
Then, taking the natural logarithm of both sides of this equality, we have
−rt = ln(1 − pt ). After some additional rearrangement, we derive our desired
relationship between the rate, time, and the transition probability:
Equation 3.3.9. Probability of decay:

ln(1 − pt )
r=−
t
Thus, for example, to translate the 1-​month cumulative probability derived
ln(1 − p0.083 )
above to a constant annual rate, one could solve for r = − =​ 0.4/​year.
0.083

3.3.5.3 Changing time frames of probabilities derived from

the literature

Now that we have established the relationship between rates, probabil-

ities, and time, we can leverage these properties to manipulate probabilities,
 3.3.5 Formatting data for analysis 185

including translating them to rates, and translating back rates into prob-
abilities. This is extremely useful when one needs to change the time frame
of probability from data derived from the literature. For example, in our
hepatitis B example, we found from the literature that 30% of people have
controlled diabetes at 5 years, suggesting that the 5-​year probability of con-
trolled diabetes is 0.3. However, our model runs on a 1-​year cycle, so we
need to convert this probability into the relevant time frame. To do this, we
first convert the reported probability from the literature into a rate, using
ln(1 − 0.3)
Equation 3.3.9. So, the annual rate is − = 0.071/ year. Next, we
5
convert back the calculated rate into an annual probability, using Equation
3.3.8: the resulting annual probability is 0.069.

3.3.5.4 Translating odds to probabilities

As we have seen earlier, the odds of an event occurring is the ratio of the prob-
ability of occurrence of that event, divided by the probability of that event
p
not occurring, and is expressed mathematically as odds = , where p
1− p
denotes the probability of occurrence of the event. From this relationship, it
is straightforward to express the probability in terms of the odds: the proba-
bility becomes the odds divided by 1 plus the odds:
Equation 3.3.10. From odds to probability:

odds
p=
1 + odds

3.3.5.5 Translating relative risk or risk ratios to

probabilities

Translating relative risk or a risk ratio into a probability is reasonably straight-

forward, provided that one has the probability for one of the groups. For
example, if one knows the probability of infection in the unvaccinated pop-
ulation, we can calculate the probability of infection among vaccinated indi-
viduals by multiplying this known parameter (probability of infection in the
unvaccinated population) with the relative risk or risk ratio of infection be-
tween the two groups.
186 Parameter estimation

3.3.5.6 Translating risk difference to probabilities

Risk difference is risk in one group minus risk in the comparator group. Since
we have defined a probability to be a measure of risk, then risk difference
is simply the difference of the probabilities of events occurring. More con-
cretely, the risk difference in adverse events between the two double-​dose
mRNA vaccines approved for use against COVID-​19 under an Emergency
Use Authorization, namely BNT162b2 (developed by BioNTech, Fosun
Pharma, and Pfizer) and mRNA-​1273 (developed by Moderna and NIAID),
is the difference in the probabilities of adverse events for the two vaccines:
Equation 3.3.11. Risk difference:

Risk difference = PAE of Pfizer − PAE of Moderna

If a study reports the risk difference between two events or outcomes, it

should and will often report one of the probabilities. In modeling applica-
tions, we are often interested in estimating the treatment effect, that is, how
the intervention affects outcomes, or the risk of occurrence of certain out-
comes or events. For example, in the case of vaccination we would want to
know how the introduction of a new vaccine affects the risk of infection
among the vaccinated individuals, and in the broader population. Analysts
should therefore directly use the probability of the outcome or events of in-
terest for the treatment (intervention) group, whenever reported in the study.
If, instead, the study reports the probability of occurrence of the outcome or
events of interest for the control group, we can use that information along
with the risk difference to derive the probability of the outcome for the treat-
ment group.

3.3.5.7 Translating odds ratios to probabilities

Odds ratios are more likely to be reported in the literature than relative risk
ratios. Hence, it is important to understand how one can leverage this in-
formation to inform vaccine economics models. If the outcome is rare (i.e.,
≤10%), it may be reasonable to assume, as we have seen earlier, that the odds
ratio approximates the relative risk. In this case, the probability of the out-
come (e.g., infection) in the exposed group (e.g., vaccinated) can be calculated
as the odds ratio multiplied by the probability of the outcome in the unex-
posed group (e.g., unvaccinated). However, if the outcome is not rare, one
 3.3.5 Formatting data for analysis 187

needs to consult a statistician about how to translate odds ratio statistics into
probabilities.
Whether it is reasonable to assume that the odds ratio approximates risk
ratios depends on the probability of the outcome in the unexposed group.
This probability should be and is often reported in the published study. In
situations where the probability of the outcome in the unexposed group is
conveniently reported in the study from which the odds ratio was extracted,
this can be used, in combination with the odds ratio, to derive the probability
of the outcome in the exposed group. It suffices to notice that the probability
of the outcome in the unexposed group can be used to derive the odds for the
outcome in that group (use Equation 3.3.1). Next, we can use Equation 3.3.2
to derive the odds for the outcome in the exposed group. Finally, Equation
3.3.10 can be used to derive the probability of the outcome in the exposed
group. When the probability of the outcome in the unexposed group is not
reported in the study, the next reasonable step is to return to the literature to
find this value for a similar group of patients or population being studied.

3.3.5.8 Using survival data to estimate probabilities

In the previous sections, the probabilities are treated as constant throughout

the model. But they need not be constant. For example, the risk of death is
not, and should not be assumed to be, constant over time. In a more realistic
model, and depending on the time horizon of the model, the probability of
death will change with age. Hence, there would be multiple probabilities of
death, one for each age, or cycle of the model. But how does one estimate or
derive such data?
Probabilities of death are typically derived from survival curves, which are
constructed from survival data largely derived from vital statistics databases.
For example, in the US, these data are maintained by the Centers for Disease
Control and Prevention’s National Vital Statistics System, which tracks impor-
tant statistics, including birth, mortality, and life expectancy data. Mortality
data are reported by cause of death, and in aggregate, for any cause. These
cause-​specific mortality estimates can be age, sex, and race adjusted.
These mortality databases known as life tables or actuarial tables summa-
rize the vital experiences of the individuals in the population, from birth to
death. Life tables are widely used in the insurance industry to estimate bene-
ficiaries’ life expectancy, and to set insurance premia. A cohort life table or
follow-​up life table is a type of life table which summarizes the experiences of
188 Parameter estimation

participants over a predefined follow-​up period in a cohort study or a clinical

trial, until the outcome of interest or the end of the study, or lost-​to follow-​up,
or death, whichever comes first. For example, reported survival rates in life ta-
bles may be death rates (generally expressed per 100,000 population in a spe-
cified group). To translate survival rates to probabilities, we can leverage the
formula for the rate to probabilities conversion discussed in Equation 3.3.8:
pt =​1 –​e−rt, where pt is the probability of remaining event free (e.g., death) at
time t, r is the incidence rate of deaths, and t is the duration of time.
Condition-​or treatment-​specific survival data can also be derived from the
literature, where they may be reported as probability of death at a given time,
t, or as survival curves. A survival curve (also known as survival function or
survival distribution) represents the probability that a person “survives,” that
is, remains free of the outcome of interest, beyond a certain time point, where,
conventionally, time is shown on the x-​axis and survival (i.e., the proportion
of people at risk for the outcome) is shown on the y-​axis.
These condition-​or treatment-​specific survival data can be estimated both
parametrically and non-​parametrically. Common parametric methods in ap-
plications include the exponential, Weibull, Gompertz, and log-​normal dis-
tributions (Cox & Oakes, 1984). The most popular of them is arguably the
exponential distribution, which assumes that the likelihood of an outcome
occurring in an individual is independent of the duration of that individual
being event free. The other distributions make different assumptions about
the probability of the outcome being realized for an individual, such as the
probability increasing or decreasing over time. For more details on parametric
methods for estimating survival curves, the reader should see publications by
Hosmer and Lemeshow (1999) and Lee and Wang (2003).
Non-​parametric methods, unlike parametric methods, make no assump-
tion about how the probability of the outcome for a person changes with time.
They typically involve estimating and plotting the survival distribution (i.e.,
survival curve) from the empirical data, using Kaplan–​Meier curves which re-
port unadjusted survival data from RCTs, or Cox proportional hazard curves
which report adjusted survival data for observational studies (Cox & Oakes,
1984; Crawley, 2005; Gehan, 1965; Greenwood, 1926; Hosmer & Lemeshow,
1999; Kalbfleisch & Prentice, 2002; Kleinbaun & Klein, 2005; Lee & Wang,
2003; Mantel, 1966; Peto & Peto, 1972).
The evaluation of vaccination interventions often requires estimates of the
survival curve of the intervention. When individual patient-​level data are
available, the parametric and non-​parametric methods described above can
be used to estimate the survival rates, which can then be used to derive the
 3.3.6 Point estimate and uncertainty distribution 189

probabilities. However, as individual patient-​level data are often unavailable,

survival curves are typically calculated by fitting a nonlinear least squares
model to the Kaplan–​Meir plots available from the published literature.
Unfortunately, this approach does not account for the uncertainty associated
with the Kaplan–​Meir curve and can produce biased estimates. In addition,
Kaplan–​Meier curves may only provide survival data up to a few months or
years, when the analyst is interested in the life course of those affected by the
intervention. Extrapolation to a lifetime horizon is therefore required.

3.3.5.9 Deriving probabilities from means of

continuous distributions

Deriving probabilities from means is very challenging and should be applied

with great care. There are at least two important challenges when considering
this approach. First, one needs a validate method to generate a binary vari-
able from a continuous distribution (e.g., a threshold), in order to characterize
discrete events. For example, when using HbA1c to define controlled dia-
betes, one may want to use a threshold value of, say, 7, so that all individuals in
the population with a HbA1c value below this threshold of 7 are classified as
having controlled diabetes.
The second challenge is that we need an estimate of variation around the
mean (e.g., standard deviation or variance) or median (interquartile range).

3.3.6 Characterize parameters by both a point estimate

and an uncertainty distribution

Parameters used in models are typically not known with certainty; if they
were, decision-​making would be considerably simplified. Because of un-
certainty in parameters, it is important to characterize parameters by both
a point estimate, and an uncertainty distribution, and explore the implica-
tions of this uncertainty for the results of the analysis. Fortunately, standard
statistical methods for estimation generate a point estimate, as well as some
measure of precision such as standard errors (SEs) or 95% confidence inter-
vals. In a multivariable framework, a measure of covariance between the esti-
mated parameters is also available. Information about uncertainty should be
used in economic models to characterize parameter uncertainty and inform
uncertainty analyses. The link to the underlying evidence base should also be
190 Parameter estimation

made clear. See Chapter 4.3 for more details on handling uncertainty in eco-
nomic analysis.

3.3.6.1 Choosing distributions for parameters

Decision-​analytic model parameters are summary representations of the av-

erage population-​level experiences of potential individuals affected by an in-
tervention or a program. Because of this, the relevant uncertainty to capture
when choosing the distribution for a parameter is the second-​order uncer-
tainty of the parameter’s sampling distribution, as opposed to the first-​order
uncertainty, which relates to the variability in the values observed in the pop-
ulation (Stinnett & Paltiel, 1997). The assumption of normality, which is a pro-
perty of large samples, is commonly used in statistics, and is grounded in the
central limit theorem. But this assumption may be unrealistic for many health
economic applications. For example, normally distributed parameters can as-
sume any value between minus infinity and positive infinity. But in reality,
parameters in most applications will have logical bounds on possible values
they can assume, so the normal distribution may be inappropriate, and it is
critical to carefully chose the appropriate distribution.
At least five different types of parameters are used in vaccine economics
models. They include probabilities, resource items, unit costs, relative risks,
and effectiveness parameters. In what follows, we briefly discuss the nature
of the data that can be used to inform the estimation of these parameters, the
logical bounds on the parameters, and the ways in which one can reasonably
select their distributions.

3.3.6.1.1 Probability parameters

Probabilities for vaccine economic models can be estimated using observed
proportions of the outcome of interest (e.g., infection incidence). At the in-
dividual level, the data can be classified as success (e.g., infected) or failure
(e.g., uninfected), and can be considered to be independent Bernoulli trials,
so that the binomial distribution can be considered. So, it may be convenient
to naturally use the proportion of successful cases (i.e., proportion of indi-
viduals infected) as an estimate of the probability of infection. However, the
binomial distribution, which is a discrete distribution highly dependent on
the sample size, may be inappropriate for modeling the distribution of prob-
abilities, which are continuous measurements. To overcome these limitations
the beta distribution, which is a conjugate of the binomial distribution and
 3.3.6 Point estimate and uncertainty distribution 191

a continuous distribution on the time interval of 0 to 1, may be more appro-

priate (Briggs, 2005).

3.3.6.1.2 Resource items parameters

Full economic evaluation analyses are collectively concerned with the effi-
cient, affordable, profitable, and equitable use of scarce health resources. The
number of resource items (e.g., number of vaccine doses) needed for a par-
ticular vaccination program is a count variable and can be modeled using the
Poisson distribution with constant mean-​variance parameter λ. We can there-
fore use the Poisson estimate of the variance (λ) to calculate the SE for the
mean resource use (e.g., average number of vaccine doses in a population of
interest) to characterize its distribution. This SE is simply the squared root of
the variance. In large samples, we can even use the estimated SE and mean,
in combination with the central limit theorem, to derive a normal sampling
distribution. However, as the underlying assumption of large sample is often
violated, this approach is rarely recommended. In smaller samples, there is a
non-​zero probability that the normal distribution could take a value less than
0; this would clearly be inappropriate for resource use parameters, as they must
be 0 or strictly positive. To overcome this limitation, a Bayesian framework is
often recommended. Under this approach, we can use the gamma distribu-
tion to characterize the uncertainty distribution for the mean resource use
parameter, without concerns of generating inconsistent values. The gamma
distribution, which is constrained to be positive and is fully continuous, is a
conjugate to the Poisson distribution, meaning that the two distributions be-
long to the same parametric family of distributions (i.e., set of distributions
with the same functional form, but differing only by the value of the finite-​
dimensional parameter).

3.3.6.1.3 Unit cost parameters

To assign monetary values to resource use items, it is common to multiply the
units of resource use with the unit cost of each resource item. This can be, for
example, the cost per dose of a vaccine. If the values of the unit cost param-
eters had no uncertainty, they would not contribute to the broader uncertainty
in the cost of resources used. Indeed, in economic evaluations alongside clin-
ical trials, unit costs are often treated as fixed parameters. But in reality, unit
costs are uncertain; for example, they may vary over time, or across vendors.
The unit cost is a strictly non-​negative continuous variable, unlike data for re-
source use, which we discussed above. Therefore, a gamma distribution could
be used to represent uncertainty in unit costs (Briggs et al., 2006). If the unit
192 Parameter estimation

cost is not more variable than the resource use it will be applied to, it may be
appropriate to use a normal distribution instead.
Sometimes, costs data are presented as an aggregate value for resource use,
weighted by unit costs. In this case, a log-​normal distribution or a gamma dis-
tribution may be appropriate for approximating the uncertainty distribution
of costs.

3.3.6.1.4 Relative risk parameters

Relative risk parameters are commonly used in economic evaluation models
to derive probabilities for other groups, as we have discussed in section 3.3.2.3.
For vaccine economics models, relative risk parameters are particularly im-
portant, as they can inform us about the difference in the risk of exposures, or
outcomes between, say those who have been vaccinated and those who have
not. Relative risk is also a common primary outcome measure in clinical trial
studies. The calculation of confidence intervals for relative risk parameters in
these studies often assumes, based on the central limit theorem, that the nat-
ural logarithm of the relative risk parameter is normally distributed, so that
the confidence intervals can be calculated using standard techniques. As we
have discussed in section 3.3.2.3, the confidence interval for the relative risk
parameter is simply obtained by exponentiating the calculated confidence
intervals on the logarithmic scale. This suggests that uncertainty for relative
risk parameters may be characterized by the log-​normal distribution.

3.3.6.1.5 Effectiveness parameters

Health outcomes are represented in health economic models by effectiveness
parameters, which have inherent uncertainties. In what follows, we discuss
how these uncertainties can be characterized and incorporated in vaccine ec-
onomics models. Choosing the appropriate distribution of these parameters is
of a paramount importance to obtain a balance between accuracy of fit to the
data—​which improves model precision—​and simplicity—​which provides ease
of use.
Health outcomes are typically measured as natural units, or utilities (e.g.,
quality-​adjusted life years, disability-​adjusted life years, etc.). When the ef-
fectiveness parameter is expressed in natural units, the main source of un-
certainty is other model parameters, such as probabilities, and relative risk.
However, when the health outcomes are measured in utilities, additional
sources of uncertainties need to be taken into account.
Utilities are preference weights applied to the length of time spent in each
health state to construct the composite measure of quality-​adjusted life years.
 3.3.6 Point estimate and uncertainty distribution 193

These weights, also called health-​related quality of life (HRQoL), capture

an individual’s or population’s perceived physical and mental health over
time, and are measured using multi-​attribute utility instruments such as the
EuroQol five-​dimensional (EQ-​5D)—​the most widely used instrument—​or
other instruments such as the SF-​6D. HRQoL data are collected through sur-
veys. Hence, they have uncertainties, due to measurement errors and other
types of biases, so, it is critical to capture the impacts of these uncertainties
in models. HRQoL data are continuous integers anchored and censored at
0 (death) and 1 (full health), but with a lower bound of minus infinity (i.e.,
states worse than death). For many widely used instruments, the lower bound
is censored at −1. Hence, there is no single distribution which matches these
characteristics (Brazier, Roberts, & Deverill, 2002; Dolan, 1997; Feeny et al.,
2002). As a result, any distribution chosen to characterize HRQoLs will be
an approximation. The distributions most widely used to approximate un-
certainties in HRQoLs include the beta, gamma, log-​normal, and normal
distributions.
A related measure is the disutility, which is the disability weight associated
with a health state and is used in the calculation of disability-​adjusted life
years. Similar to the HRQoL, the disability weights also have uncertainties,
and are continuous integers on the range of 0 (full health, no disability) and 1
(disability level in a state such as death), so that their uncertainties can also be
characterized with the beta distribution.

3.3.6.2 Estimating model parameters from the

mean and standard error

As we have already noted, parameters are typically characterized with certain

measures of variation around their mean values, and the underlying distri-
butions of their uncertainty. The measure of variation is typically reported as
a SE, 95% confidence interval, or 25% and 75% quartiles. With this informa-
tion, we can fit the model’s input parameters to appropriate probability dis-
tributions by estimating the parameters of the corresponding (approximate)
probability distributions (Westwood et al., 2017).
For some model input parameters (e.g., utilities for chronic hepatitis B
states), the literature may already report the probability distributions’ param-
eters so that they can be included directly in the model. However, when they
are not reported, the analyst must estimate them, sometimes using nonlinear
fitting techniques (Westwood et al., 2017). For example, when the mean and
194 Parameter estimation

SE of a parameter is available, the parameters of the corresponding beta distri-

bution can be obtained from the mean and SE as:
Equation 3.3.12. Beta distribution:

 mean × (1 − mean) 
α = mean ×  − 1
 SE 2 

 mean × (1 − mean) 
β = (1 − mean) ×  − 1
 SE 2 

where α and β are strictly positive real numbers denoting the shape param-
eters (Westwood et al., 2017).

3.3.6.3 Estimating model parameters from a

confidence interval or an interquartile range

When a mean, and confidence interval or any other uncertainty range, is re-
ported, parameter estimation is not as straightforward as in the approaches
mentioned above. In these situations, more complex numerical procedures
may be required. For example, it may be necessary to use more advanced soft-
ware packages such as R, Stata, SPSS, Python, Julia, Matlab, Mathematica, and
so on, and more advanced nonlinear multivariate fitting techniques to esti-
mate the parameters of the probability distribution function.
Sometimes, it is necessary to resort to expert opinion to obtain information
on the most likely value of a parameter, its range of uncertainty, and poten-
tially the underlying distribution of the parameter. Multiple experts may need
to be consulted under this approach. In most applications, including vaccine
economics applications, it is common to use the beta or PERT distributions to
characterize this distribution.

3.3.7 Assessing the influence of parameter uncertainty,

assumptions, and data quality and appropriateness

Sensitivity analyses should also be conducted to assess the impacts of uncer-

tainties in model parameters, model assumptions, data quality, and data ap-
propriateness on the model’s results and conclusions. These analyses should
be guided by the same sets of principles involved with parameter estimation.
 3.3.8 Concluding remarks 195

These may include one-​way sensitivity analyses which can help identify
parameters that are most influential in the model. Multivariate and proba-
bilistic sensitivity analyses can help assess the joint impact of parameter un-
certainties on the results and can be used to construct credible confidence
intervals around the model’s predicted outcomes. See Chapter 4.3 for more
details.

3.3.8 Concluding remarks

Ultimately, for each parameter estimate in an economic evaluation, investi-

gators should strive to (1) include both a point estimate and an uncertainty
distribution; (2) incorporate empiric data (or estimates from the scientific lit-
erature) when available, minimizing the number of assumptions; (3) utilize
data of high quality; (4) assess data for appropriateness for the particular anal-
ysis; (5) convert data into a format that is consistent with the analytic structure;
and (6) evaluate the influence of these decisions on the main results and con-
clusions. Economic evaluations that carefully follow these steps in parameter
estimation are more likely to produce results that are both epidemiologically
valid and meaningful to decision makers.

References
Ades, A. E., Lu, G., & Higgins, J. P. (2005). The interpretation of random-​effects meta-​analysis in
decision models. Medical Decision Making, 25(6), 646–​654. doi:10.1177/​0272989x05282643
Ades, A. E., Sculpher, M., Sutton, A., Abrams, K., Cooper, N., Welton, N., & Lu, G. (2006).
Bayesian methods for evidence synthesis in cost-​effectiveness analysis. PharmacoEconomics,
24(1), 1–​19. doi:10.2165/​00019053-​200624010-​00001
Baker, S. G. (2006). A simple meta-​analytic approach for using a binary surrogate endpoint to
predict the effect of intervention on true endpoint. Biostatistics, 7(1), 58–​70. doi:10.1093/​
biostatistics/​kxi040
Berlin, J. A., Santanna, J., Schmid, C. H., Szczech, L. A., & Feldman, H. I. (2002). Individual
patient-​versus group-​level data meta-​regressions for the investigation of treatment ef-
fect modifiers: Ecological bias rears its ugly head. Statistics in Medicine, 21(3), 371–​387.
doi:10.1002/​sim.1023
Bilcke, J., Antillón, M., Pieters, Z., Kuylen, E., Abboud, L., Neuzil, K. M., . . . Pitzer, V. E.
(2019). Cost-​effectiveness of routine and campaign use of typhoid Vi-​conjugate vaccine
in Gavi-​eligible countries: A modelling study. Lancet Infectious Diseases, 19(7), 728–​739.
doi:10.1016/​s1473-​3099(18)30804-​1
Brazier, J., Roberts, J., & Deverill, M. (2002). The estimation of a preference-​based measure
of health from the SF-​36. Journal of Health Economics, 21(2), 271–​292. doi:10.1016/​
s0167-​6296(01)00130-​8
196 Parameter estimation

Briggs, A. (2005). Probabilistic analysis of cost-​effectiveness models: Statistical representation

of parameter uncertainty. Value in Health, 8(1), 1–​2. doi:10.1111/​j.1524-​4733.2005.08101.x
Briggs, A., Sculpher, M., & Claxton, K. (2006). Decision modelling for health economic evalua-
tion. Oxford: Oxford University Press.
Collett, D. (2015). Modelling survival data in medical research. New York, NY: Chapman and
Hall/​CRC.
Cox, D. R., & Oakes, D. (1984). Analysis of survival data. New York, NY: Chapman and
Hall/​CRC.
Crawley, M. J. (2005). Statistics: An introduction using R. New York, NY: John Wiley & Sons.
Dias, S., Ades, A. E., Welton, N., Jansen, J. P., & Sutton, A. (2018). Network meta-​analysis for
decision-​making. New York, NY: John Wiley & Sons.
Dolan, P. (1997). Modeling valuations for EuroQol health states. Medical Care, 35(11),
1095–​1108. doi:10.1097/​00005650-​199711000-​00002
Feeny, D., Furlong, W., Torrance, G. W., Goldsmith, C. H., Zhu, Z., DePauw, S., . . . Boyle, M.
(2002). Multiattribute and single-​attribute utility functions for the health utilities index
mark 3 system. Medical Care, 40(2), 113–​128. doi:10.1097/​00005650-​200202000-​00006
Gehan, E. A. (1965). A generalized Wilcoxon test for comparing arbitrarily singly-​censored
samples. Biometrika, 52, 203–​223. https://​pub​med.ncbi.nlm.nih.gov/​14341​275/​
Glick, H. A., Doshi, J. A., Sonnad, S. S., & Polsky, D. (2014). Economic evaluation in clinical
trials. New York, NY: Oxford University Press.
Gold, M. R., Siegel, J., Russell, L., & Weinstein, M. (1996). Cost-​effectiveness in health and medi-
cine. New York, NY: Oxford University Press.
Greenwood, M. (1926). A report on the natural duration of cancer: London: HMSO.
Guyatt, G., Rennie, D., Meade, M. O., & Cook, D. J. (2015). Users’ guides to the medical lit-
erature: A manual for evidence-​based clinical practice (3rd ed.). New York, NY: American
Medical Association/​McGraw-​Hill.
Hasselblad, V., & McCrory, D. C. (1995). Meta-​analytic tools for medical decision making: A
practical guide. Medical Decision Making, 15(1), 81–​96. doi:10.1177/​0272989x9501500112
Higgins, J., Thomas, J., Chandler, J., Cumpston, M., Li, T., Page, M., & Welch, V. (Eds.) (2022).
Cochrane handbook for systematic reviews of interventions, version 6.3 (updated February
2022). Cochrane. http://​www.train​ing.cochr​ane.org/​handb​ook
Higgins, J. P., Whitehead, A., Turner, R. M., Omar, R. Z., & Thompson, S. G. (2001). Meta-​
analysis of continuous outcome data from individual patients. Statistics in Medicine, 20(15),
2219–​2241. doi:10.1002/​sim.918
Hosmer, D. W., & Lemeshow, S. (1999). Applied survival analysis: Regression modeling of time to
event data. New York, NY: John Wiley & Sons.
Hunink, M., Glasziou, P., Siegel, J., & Weeks, J. (2001). Decision making in health and medi-
cine: Integrating evidence and values. New York, NY: Cambridge University Press.
Kalbfleisch, J. D., & Prentice, R. L. (2002). The statistical analysis of failure time data (2nd ed.).
New York, NY: John Wiley & Sons.
Kleinbaun, D. G., & Klein, M. (2005). Survival analysis: A self-​learning text. New York,
NY: Springer Science +​Business Median, Inc.
Lambert, P. C., Sutton, A. J., Abrams, K. R., & Jones, D. R. (2002). A comparison of summary
patient-​level covariates in meta-​regression with individual patient data meta-​analysis.
Journal of Clinical Epidemiology, 55(1), 86–​94. doi:10.1016/​s0895-​4356(01)00414-​0
Lee, E. T., & Wang, J. W. (2003). Statistical methods for survival data analysis (3rd ed.). New York,
NY: John Wiley & Sons.
Lu, G., & Ades, A. E. (2004). Combination of direct and indirect evidence in mixed treatment
comparisons. Statistics in Medicine, 23(20), 3105–​3124. doi:10.1002/​sim.1875
Lumley, T. (2002). Network meta-​analysis for indirect treatment comparisons. Statistics in
Medicine, 21(16), 2313–​2324. doi:10.1002/​sim.1201
 3.3.8 Concluding remarks 197

Mantel, N. (1966). Evaluation of survival data and two new rank order statistics arising in its
consideration. Cancer Chemotherapy Reports, 50(3), 163–​170. https://​pub​med.ncbi.nlm.
nih.gov/​5910​392/​
Massoudi, N., & Mohit, B. (2021). A case–​control study of the 2019 influenza vaccine and inci-
dence of COVID-​19 among healthcare workers. Journal of Clinical Immunology, 41(2), 324–​
334. doi:10.1007/​s10875-​020-​00925-​0
Neumann, P., Ganiats, T., Russell, L., Sanders, G., & Siegel, J. (2016). Cost-​effectiveness in health
and medicine. New York, NY: Oxford University Press.
Pannucci, C. J., & Wilkins, E. G. (2010). Identifying and avoiding bias in research. Plastic and
Reconstructive Surgery, 126(2), 619–​625. doi:10.1097/​PRS.0b013e3181de24bc
Peto, R., & Peto, J. (1972). Asymptotically efficient rank invariant test procedures. Journal of the
Royal Statistical Society, 135(2), 185–​207. http://​www.jstor.org/​sta​ble/​2344​317
Piantadosi, S., Byar, D. P., & Green, S. B. (1988). The ecological fallacy. American Journal of
Epidemiology, 127(5), 893–​904. doi:10.1093/​oxfordjournals.aje.a114892
Qu, C., Chen, T., Fan, C., Zhan, Q., Wang, Y., Lu, J., . . . Sun, Z. (2014). Efficacy of neo-
natal HBV vaccination on liver cancer and other liver diseases over 30-​year follow-​up of
the Qidong hepatitis B intervention study: A cluster randomized controlled trial. PLoS
Medicine, 11(12), e1001774–​e1001774. doi:10.1371/​journal.pmed.1001774
Rosenthal, R., & DiMatteo, M. R. (2001). Meta-​analysis: Recent developments in quantita-
tive methods for literature reviews. Annual Review of Psychology, 52, 59–​82. doi:10.1146/​
annurev.psych.52.1.59
Stinnett, A. A., & Paltiel, A. D. (1997). Estimating CE ratios under second-​order uncer-
tainty: The mean ratio versus the ratio of means. Medical Decision Making, 17(4), 483–​489.
doi:10.1177/​0272989x9701700414
Sutton, A. J., Abrams, K. R., Jones, D. R., Sheldon, T. A., & Song, F. (2000). Methods for meta-​
analysis in medical research (Vol. 348). Chichester: Wiley.
Sutton, A. J., & Higgins, J. P. (2008). Recent developments in meta-​analysis. Statistics in
Medicine, 27(5), 625–​650. doi:10.1002/​sim.2934
Sutton, A. J., Song, F., Gilbody, S. M., & Abrams, K. R. (2000). Modelling publication bias in
meta-​analysis: A review. Statistical Methods in Medical Research, 9(5), 421–​445. doi:10.1177/​
096228020000900503
Thompson, S. G., & Higgins, J. P. (2002). How should meta-​regression analyses be undertaken
and interpreted? Statistics in Medicine, 21(11), 1559–​1573. doi:10.1002/​sim.1187
Wakefield, J. (2008). Ecologic studies revisited. Annual Review of Public Health, 29, 75–​90.
doi:10.1146/​annurev.publhealth.29.020907.090821
Westwood, M., Corro Ramos, I., Lang, S., Luyendijk, M., Zaim, R., Stirk, L., . . . Kleijnen, J.
(2017). Faecal immunochemical tests to triage patients with lower abdominal symptoms
for suspected colorectal cancer referrals in primary care: A systematic review and cost-​
effectiveness analysis. Health Technology Assessment, 21(33), 1–​234. doi:10.3310/​hta21330
3.4
Measuring and valuing health
outcomes
Y. Natalia Alfonso, Stéphane Verguet, and Ankur Pandya

The impact of vaccines and immunization programs on populations can be

measured with the use of a range of outcome measures, including (1) process
and intermediate outputs (e.g., number of children fully immunized, vacci-
nation coverage); (2) natural unit outcomes specific to vaccine-​preventable
diseases (e.g., cases or deaths associated with measles or rotavirus averted);
(3) constructed summary measures of population health (e.g., incorporating
both mortality or morbidity outcomes like disability-​adjusted life years
(DALYs) or quality-​adjusted life years (QALYs)); and (4) equity metrics,
such as the distribution in health outcomes and non-​health benefits such as
financial risk protection (prevention of medical impoverishment). A com-
parison of the strengths and weaknesses of all such measures in vaccine ec-
onomics is described in detail elsewhere (Bärnighausen et al., 2014; Jit et al.,
2015; Walker, Hutubessy, & Beutels, 2010). This chapter introduces some
measures pertaining to (3) and (4) and their use in the economic evaluation
of vaccines.

3.4.1 Preference elicitation

To quantify the full impact of vaccines on the health of populations, both im-
provements in the length of life and health-​related quality of life (HRQoL)
should ideally be measured. Constructed summary measures of population
health that incorporate both components into a single unified metric are gen-
erally categorized as “health-​adjusted life years” or HALYs. To understand
how to estimate HALYs, it is important to distinguish between measuring and
valuing health outcomes:

Y. Natalia Alfonso, Stéphane Verguet, and Ankur Pandya, Measuring and valuing health outcomes In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0017
 3.4.1 Preference elicitation 199

• Measuring refers to counting the number of events. For example, the

number of deaths, infections averted, disability cases averted, and so on.
• Valuing refers to the subjective appreciation of avoiding a disease that is
not fatal. For example, how does one compare the value of living 10 ad-
ditional years without post-​polio paralysis versus 5 years with post-​polio
paralysis?

It is difficult to define a single value that can accurately represent the reduced
quality of life or enhanced disability associated with the experience of being
in one specific disease state, such as post-​polio paralysis. Preference-​based
measures provide one approach to the valuation of health states (Feeny, 2005;
Rowen et al., 2020). The basic idea is to ask people to state their preferences
for specific health states. For example, on a scale from 0 to 100, where 0 is
death and 100 is the best health state imaginable, how would you rate your or
someone else’s quality of life during a year experiencing measles?
Robust methods for eliciting an individual’s preferences (i.e., “patient pref-
erences”) for the HRQoL value of various health states require a rational
decision-​making process. This process includes understanding what is at
risk with or without a vaccine-​preventable disease, the trade-​offs at stake,
and the uncertainty of those risks and trade-​offs. In other words, this pro-
cess requires that the individual doing the quality-​of-​life assessment has
complete information and understanding of the short-​term and long-​term
risks and uncertainty with and without a vaccine-​preventable disease. See
Box 3.4.1 for an example of a preference-​based valuation scenario that incorp-
orates these aspects. This scenario presents a case in which an individual con-
siders both the health and economic attributes that influence their decisions.
Ideally, HRQoL valuation techniques would incorporate both attributes while
also assessing risk, trade-​offs, and uncertainty. However, it may be difficult
for a single valuation technique to incorporate all these characteristics. In
fact, most HRQoL techniques focus on health attributes, excluding economic
attributes, and incorporate one or two of the requirements for eliciting a ra-
tional decision-​making process. Another aspect to consider is that the best
vaccine choice is subjective based on the individual eligible to receive it and
what health or economic conditions they prefer to avoid. As such, quality-​of-​
life valuations are likely to vary by conditions such as age, sex/​gender, race,
income, and other factors.
The three most common techniques used to elicit patient preferences include
the visual analog scale (VAS), time trade-​off (TTO), and standard gamble (SG)
approaches (Feeny, 2005; Torrance, 1982; Torrance, Feeny, & Furlong, 2001).
200 Measuring and valuing health outcomes

Box 3.4.1 Example of preference-​based evaluation for the

Hemophilus influenzae type b vaccine

Rational decision-​making for patients requires trade-​offs like the following:

Short-​term common-​side effects of the vaccine:

• Pain, muscle aches.

• Fever in children.
• Costs of vaccine.

Long-​term health benefits:

• Avoid H. influenzae type b sequelae (e.g., pneumonia).

• Avoid indirect consequences like lost productivity.

The decision to proceed with the H. influenzae type b vaccine depends on the trade-​
off between short-​term treatment of vaccine and flu symptoms and longer-​term treat-
ment risks.

These techniques are used to design survey questions that facilitate a rational
decision-​making process. The focus is to assess an individual’s preferences for
both the length and quality of life based on vaccine usage. To do this, these
techniques incorporate one or more of the following types of preference meas-
ures: (1) value, (1) risk, and (3) time. See Box 3.4.2 for a description of each of
these three preference measures. Likewise, see Drummond, Sculpher, Claxton,
Stoddart, and Torrance (2015a) for details and comparisons between the VAS,
TTO, and SG techniques.
The overall objective of these measures is estimating HRQoL values for
different health states in a given population. When quality valuations reflect
preferences then the value is known as a “utility weight,” “index,” or “score.”
Utility weights are a quantifiable index of health captured on a scale ranging
from 0.0 to 1.0 to inform a HALY metric such as a QALY gained or DALY
averted (Drummond et al., 2015a). While many instruments use the VAS,
TTO, and SG to generate HALYs, there is a hierarchy in terms of each item’s
fidelity to capturing value, risk, time, and uncertainty. The VAS is easy to use
and interpret but it is not considered a true preference-​based measure given
that it does not specify aspects of risk and time to generate a utility value
 3.4.1 Preference elicitation 201

Box 3.4.2 Preference-​based measures

Value preference
How do you feel about one certain outcome relative to another certain outcome?

Risk preference
How do you feel about one certain outcome versus a gamble on another outcome?

Time preference
How do you feel about a certain outcome today versus the same outcome in the future?

(Torrance et al., 2001). The TTO and SG incorporate both a time horizon and
value preferences, but only the SG incorporates measurement for risk and
uncertainty. As such, the hierarchy of preference between these measures is
the SG, followed by TTO, followed by VAS (Drummond et al., 2015a; Gold,
Stevenson, & Fryback, 2002). See Box 3.4.3 for an example of the SG tech-
nique used to elicit patient preferences.
For more details about patient preferences and utilities see the video
on “Introduction to effectiveness, patient preferences, and utilities”
(Jacobs, 2018).
The most prominent survey instruments designed to elicit preference-​
based valuations of quality of life include the Quality of Well-​B eing Scale,
the Health Utilities Index, and the EuroQol Group’s EQ-​5D questionnaires
(Drummond, Sculpher, Claxton, Stoddart, & Torrance, 2015b; Mortimer
& Segal, 2008). The EQ-​5D questionnaires are widely used in many high-​
income countries and some low-​and middle-​income countries (LMICs)
and are translated into more than 200 languages (e.g., English, Mandarin,
Thai, Amharic) (Brooks, Boye, & Slaap, 2020; Herdman et al., 2011; Shaw,
Johnson, & Coons, 2005). The questionnaires are designed using the TTO
preference-​based technique and consist of five domains that assess an
individual’s health, including mobility, self-​care, usual activities, pain/​
discomfort, and anxiety/​depression (e.g., the EQ-​5D-​5L questionnaire)
(van Reenen & Janssen, 2015). A child-​friendly version is also available
(i.e., EQ-​5D-​Y) (Brooks et al., 2020; Devlin & Brooks, 2017; Kind, Klose,
Gusi, Olivares, & Greiner, 2015). See Box 3.4.4 for an example of a HRQoL
valuation of post-​polio syndrome (PPS). The scores from these question-
naires are converted to utility weights using country-​specific value sets
202 Measuring and valuing health outcomes

Box 3.4.3 Standard gamble examples

Imagine that you have recovered from acute measles and later begin to develop
hearing loss. You can hear well today; you are not deaf. You have occasional hearing
loss in both ears. Doctors say you will become totally deaf in the next year. But there
is a potential cure . . . Successful surgery would prevent symptoms indefinitely. But
there is an X% chance that you die from the surgery.
What X% chance of death would you be willing to risk for possibility of cure?

• 10%.
• 25%.
• 50%.
• 75%.
• 90%.

At the point of indifference, the relative value of the health state of post-​measles deaf-
ness is 1.0 − X.
Imagine that you have early-​stage symptoms of paralysis after recovering from
acute polio. You can walk today; you are not paralyzed. You have occasional weakness
in legs. Doctors say you will become completely paralyzed from the waist down in the
next year. But there is a potential cure . . . Successful surgery would prevent symptoms
indefinitely. But there is an X% chance that you die from the surgery.
What X% chance of death would you be willing to risk for possibility of cure?

• 10%.
• 25%.
• 50%.
• 75%.
• 90%.

At the point of indifference, the relative value of the health state of post-​polio paral-
ysis is 1.0 − X.

(i.e., that incorporate country-​specific preferences). For a quick summary

of this survey tool see the “Explaining the EQ-​5D in about two-​and-​a-​
half-​minutes” video (EuroQol Research Foundation, 2019). For details on
how to calculate utility weights using the EQ-​5D descriptive system, see R
package “valueEQ5D” (Manchira Krishnan, 2020).
 3.4.2 Measures of vaccine effectiveness 203

Box 3.4.4 Example of health-​related quality of life valuation

using the EQ-​5D-​5L

A study is recruiting participants to estimate HRQoL values for PPS, a vaccine-​

preventable disease. Participants are either recruited from a population of patients
with PPS or from the general population without PPS. In the latter case, participants
are provided a vignette describing what it feels like to have PPS.
Each participant is asked to rate his/​her health status with PPS regarding their
pain/​discomfort by selecting one out of the five health state options below:

• I have no problems in walking about.

• I have slight problems in walking about.
• I have moderate problems in walking about.
• I have severe problems in walking about.
• I am unable to walk about.

The individual repeats this exercise for the four remaining health domains: mo-
bility, self-​care, usual activities, and anxiety/​depression.

3.4.2 Measures of vaccine effectiveness

Underlying any measure of vaccine effectiveness is a strong causal study de-

sign that compares use of the vaccine to a relevant alternative (or set of alter-
natives). These alternatives could include strategies such as usual care or status
quo coverage when evaluating vaccine scale-​up policies, for example. Strong
causal study designs include randomized controlled trials or rigorous quasi-​
experimental designs, such as controlled interrupted time series and regres-
sion discontinuity designs (Lopez Bernal, Andrews, & Amirthalingam, 2019).
The outcomes from such studies, such as cases of disease averted from the
vaccine, are the starting point for any measure of vaccine effectiveness. These
disease-​specific outcomes might not be comparable across health conditions,
however (how would one value 50 cases of measles averted from one vaccine
to 50 cases of high-​risk human papillomavirus from another vaccine?), thus
motivating the use of HALY metrics (such as QALYs or DALYs) that com-
bined length of life and quality of life into a single number, allowing for com-
parisons across health conditions.
When used in economic evaluation, QALYs are an expectancy measure in
a similar way that life expectancy is an expectancy measure. Life expectancy
204 Measuring and valuing health outcomes

is the expected number of years that an individual will live based on relevant
life tables for that individual. Quality-​adjusted life expectancy, or expected
QALYs, is similar to life expectancy in that it captures length of life, but QALYs
also capture quality of life. Quality of life is measured on a 0.0–​1.0 scale, where
0.0 indicates death and 1.0 indicates perfect health. Although perfect health is
an abstract concept, one way it can be described is by using the EQ-​5D frame-
work where someone in perfect health must have no limitations with mobility,
self-​care, usual activities, pain/​discomfort, or anxiety/​depression. All values
between 0.0 and 1.0 represent time spent in less-​than-​perfect health, and can
be elicited using direction approaches or indirect approaches (such as the
EQ-​5D). Negative QALYs imply worse-​than-​death health states, for outcomes
such as terminal diseases. QALYs are calculated as:
Equation 3.4.1. QALY:

QALYs = ∑ years spent inhealth state i × utility value for health state i
i

DALYs are similar to QALYs in that they are HALYs that combine length of
life and quality of life into a single number, but are different in that DALYs
are a gap measure, whereas QALYs are an expectancy measure. Put simply,
more DALYs are bad; DALYs averted are good. Furthermore, more QALYs
are good. Gap measures represent how far off a given individual’s life expe-
rience is from ideal health in terms of length and quality of life. Ideal length
of life is calculated using a synthetic life table that is constructed using the
best annual survival probabilities across countries and sexes (the 2016 Global
Burden of Disease analyses used a synthetic life table that resulted in an ideal
life expectancy of 92 years, for example). Any premature death (i.e., any death
before age 92 years) contributes to years of life lost, calculated as the ideal life
expectancy minus the year of premature death. Disability weights for DALYs
are measured on a 1.0–​0.0 scale, where 1.0 indicates a state of disability that is
no more functional than death, and 0.0 implies perfect health (i.e., zero disa-
bility). Researchers have published standard DALY weights that can be used
when country-​or population-​specific estimates are lacking (Salomon et al.,
2015). The number of years lost due to disability is calculated as the product
of years lived with a certain health condition and the disability weight for that
health condition. DALYs are calculated as the sum for years of life lost and
years lost due to disability.
QALYs are typically used more in economic evaluations for high-​income
countries, while DALYs are typically used more for LMICs. The Tufts Cost-​
Effectiveness Registry catalogs all health utility values used in cost-​per-​QALY
 3.4.2 Measures of vaccine effectiveness 205

cost-​effectiveness analyses, and the Tufts Global Health Cost Effectiveness

Registry includes a DALY calculator that can be customized by country, health
condition, and other variables (Tufts Medical Center, 2021). Other studies
containing “off-​the-​shelf ” sources have been published for QALY (Sullivan &
Ghushchyan, 2006) and DALY (Salomon et al., 2015) utility values, although
analysts should pay close attention to whether or not the populations and con-
ditions that were assessed to derive those specific values match the popula-
tions of interest for their own studies.
Fig. 3.4.1 shows how the QALYs gained and DALYs averted from a vac-
cine can be estimated for a given individual. In an economic evaluation, the
QALYs gained (or DALYs averted) would be calculated for each individual
(or cohort) using a simulation model (which would project estimated QALYs
or DALYs) or (less typically) from empirical observation (from a randomized
controlled trial that collected quality-​of-​life and mortality follow-​up data, for
example).
Although the equations used to calculate QALYs and DALYs are straight-
forward and have intuitive appeal (each can be represented as areas under the
curve in Fig. 3.4.1), these metrics have important limitations that should be
understood by analysts and policymakers who use QALYs and DALYs. The
additive nature of the QALY and DALY constructions requires certain theo-
retical assumptions, such as mutual utility independence and constant pro-
portional time trade-​off, which are often violated in surveys of the general
public (Box 3.4.5) (Hammit, 2002; Pliskin, Shepard, & Weinstein, 1980).
HALY formulations that do not require these assumptions and are usable
for analyses that span interventions or populations are lacking, however,
which is one reason why QALYs and DALYs are the prevailing metrics in ec-
onomic evaluations of health programs (Carlson, Brouwer, Kim, Wright, &
McQueen, 2020).
QALYs and DALYs imply certain value judgments that can lead to prob-
lematic ethical issues (Brock, 1995, 1998, 2004). The standard QALY formula
gives no weight to distributional issues, for example, “a QALY is a QALY is a
QALY” in standard economic evaluations that use QALYs gained (or DALYs
averted) as the effectiveness measure (Weinstein, 1988), which can lead to po-
tential policy recommendations that might focus interventions on already ad-
vantaged populations (richer, literate, urban) if they are easier to reach and
more likely to take advantage of services. Another issue relates to discrimina-
tory issues with the QALY or DALY for life-​saving interventions; lives saved
for disabled individuals (or any individuals in less than perfect health) would
result in fewer QALYs gained compared to the same number of lives saved
Fig. 3.4.1 Conceptual diagrams that show QALYs gained or DALYs averted from
a hypothetical vaccine. (a) The life trajectory for a hypothetical individual in a
counterfactual scenario without Disease X. This person ended up starting life in
perfect health (utility value of 1.0), then experienced lower quality of life later in their
life before dying. (b) The life trajectory for the same individual in a counterfactual
scenario where they have Disease X, which affects the length of their life (premature
death from Disease X) but not quality of their life (which remained at a utility value
of 1.0 in early years), and the QALYs gained for this individual from a vaccine that
prevents Disease X. (c) The QALYs gained for this individual from a vaccine that
prevents Disease Y, which affects both length and quality of life. (d) The DALYs for
this individual in a counterfactual scenario without Disease Y (or Disease X). (e) The
DALYs averted from a vaccine that prevents Disease Y, which happen to be equal to
the QALYs gained from the same vaccine. In practice, it is possible that DALYs averted
can be different from QALYs gained for the same intervention (if the utility weights
used to calculate QALYs or DALYs are not equal, for example), but these differences
will usually be small.
Fig. 3.4.1 Continued
208 Measuring and valuing health outcomes

Box 3.4.5 Theoretical assumptions underlying the QALY

and DALY

The construction of QALYs and DALYs has an intuitive “area under the curve” pro-
perty that only requires the length of time spent in health states and the utility
values for those health states. This calculation requires several assumptions
(mutual utility independence, constant proportional trade-​offs, and risk neutrality
over life years), however, that are often violated by individuals in surveys of the
general public. Mutual utility independence implies that an individual’s prefer-
ences between lotteries over the length of life are independent of the quality of life.
A constant proportional trade-​off implies there is a constant exchange rate between
length of time in a health state and time in perfect health. Risk neutral over life years
means individuals are indifferent to length of life trajectories that have the same ex-
pected value of life years, even if the risks of death are different over time. Without
these assumptions, the QALYs and DALYs cannot be calculated using an “area under
the curve” approach.

for individuals in perfect (or better) health. Ignoring the quality-​of-​life di-
mension of QALYs or DALYs avoids this problem, but then gives no weight
to quality of life-​improving interventions. Newer metrics, such as the health
years in total measure (Basu, Carlson, & Veenstra, 2020), have been developed
to reconcile these competing forces, but they are not yet commonly used as
complements or substitutes for QALYs or DALYs.

3.4.3 Equity impact

The burden of vaccine-​preventable diseases (VPDs) is largely concentrated

among the poorest socioeconomic groups in LMICs. As a case in point, pneu-
monia and diarrhea deaths still constitute about 20–​30% of total under-​five
deaths in LMICs (Institute for Health Metrics & Evaluation, 2020). For in-
stance, Nigeria, India, Ethiopia, and Pakistan would face estimated diarrhea
deaths among under-​fives of 104,000, 103,000, 32,000, and 29,000, respec-
tively, in 2017, which is about 50% of total diarrhea deaths in that age group
(Institute for Health Metrics & Evaluation, 2020). Under-​five mortality is
far greater among the poorest wealth quintiles than the richest quintiles in
LMICs (Chao, You, Pedersen, Hug, & Alkema, 2018); and rotavirus-​related
 3.4.3 Equity impact 209

deaths could be several times higher among the poor than the rich in those
emerging economies (Rheingans, Atherly, & Anderson, 2012; Verguet et al.,
2013). Vaccine coverage also remains unequally distributed across the socio-
economic gradient, as documented by the Demographic and Health Surveys.
For example, in Ethiopia, inequalities in vaccine access persist and the cov-
erage of the first dose of measles vaccine was 42% in the poorest wealth quin-
tile as opposed to 83% in the richest wealth quintile in 2019 (Central Statistical
Agency (Ethiopia) & ICF, 2019; Memirie, Nigus, & Verguet, 2021).
Furthermore, with the onset of VPDs comes financial consequences and
potentially financial risks for households in LMICs. VPDs can lead to large
amounts of (1) out-​of-​pocket (OOP) direct medical costs (e.g., costs of phy-
sician consultations, drugs, hospitalizations for severe VPD cases); (2) OOP
direct non-​medical costs associated with transport (and potentially housing)
costs to seek care in health facilities; and (3) indirect costs tied to time losses
and wages lost associated with seeking care. As a case in point, inpatient visits
for severe pneumonia and diarrhea could cost up to $50–​$100 to affected
households in Ethiopia (Memirie et al., 2017). In this respect, VPDs can sub-
stantially contribute to the burden of medical impoverishment in LMICs
(World Health Organization, 2019a).
As a result, vaccines can promote equity in two ways, by (1) reducing dis-
parities in the burden of VPDs across the socioeconomic gradient and by
(2) preventing the burden of impoverishment associated with the onset of
VPDs (Verguet, 2018). More generally, beyond immunization programs, there
has been a long-​standing attention to how investing in the health sector could
participate in reducing and redressing inequalities in society. Specifically,
improving the levels and distributions of health outcomes and financial risk
protection (the prevention of medical impoverishment) are core objectives
of health systems (Roberts, Hsiao, Berman, & Reich, 2019). Therefore, ex-
tended cost-​effectiveness analysis methods were developed to evaluate health
interventions—​like immunization programs—​along four dimensions: (1) the
health benefits (e.g., deaths averted or QALYs gained); (2) the household costs
averted (e.g., OOP direct costs and indirect costs averted) by intervention and
the financial risk protection gains procured to individuals; (3) the equity im-
pact across socioeconomic status (e.g., poorest versus richest income groups);
and (4) the intervention costs (Center for Health Decision Science, 2020;
Verguet et al., 2016).
Using extended cost-​effectiveness analysis, analysts can study the likely eq-
uity impact—​both distributional impact across income quintiles and poverty
reduction impact of immunization programs (Chang, Riumallo-​Herl, Perales,
210 Measuring and valuing health outcomes

et al., 2018; Chang, Riumallo-​Herl, Salomon, et al., 2018; Riumallo-​Herl et al.,

2018). Vaccines can prevent the majority of deaths among the poorest income
quintiles of LMICs, as the poorest often have the most to gain in the first place,
due to both greater VPD burden and smaller vaccine coverage. One can also
demonstrate that vaccines can substantially reduce poverty, as the poorest
would face higher VPD-​related OOP costs compared with their disposable
income (Chang, Riumallo-​Herl, Perales, et al., 2018).

References
Bärnighausen, T., Bloom, D. E., Cafiero-​Fonseca, E. T., & O’Brien, J. C. (2014). Valuing vaccina-
tion. Proceeings of the National Academy of Sciences of the United States of America, 111(34),
12313–​12319. doi:10.1073/​pnas.1400475111
Basu, A., Carlson, J., & Veenstra, D. (2020). Health years in total: A new health objective function
for cost-​effectiveness analysis. Value in Health, 23(1), 96–​103. doi:10.1016/​j.jval.2019.10.014
Brock, D. W. (1995). Justice and the ADA: Does prioritizing and rationing health care dis-
criminate against the disabled? Social Philosophy & Policy, 12(2), 159–​185. doi:10.1017/​
s0265052500004714
Brock, D. W. (1998). Appendix E: Ethical issues in the development of summary measures
of population health status. In M. J. Field & M. R. Gold (Eds.), Summarizing population
health: Directions for the development and application of population metrics (pp. 73–​81).
Washington, DC: National Academies Press.
Brock, D. W. (2004). Ethical issues in the use of cost effectiveness analysis for the prioritization
of health resources. In G. Khushf (Ed.), Handbook of bioethics: Taking stock of the field from a
philosophical perspective (pp. 353–​380). Dordrecht: Springer Netherlands.
Brooks, R., Boye, K. S., & Slaap, B. (2020). EQ-​5D: A plea for accurate nomenclature. Journal of
Patient-​Reported Outcomes, 4(1), 52. doi:10.1186/​s41687-​020-​00222-​9
Carlson, J. J., Brouwer, E. D., Kim, E., Wright, P., & McQueen, R. B. (2020). Alternative ap-
proaches to quality-​ adjusted life-​
year estimation within standard cost-​ effectiveness
models: Literature review, feasibility assessment, and impact evaluation. Value in Health,
23(12), 1523–​1533. doi: 10.1016/​j.jval.2020.08.2092.
Center for Health Decision Science. (2020). Resource pack: Extended cost-​effectiveness analysis.
https://​rep​osit​ory.chds.hsph.harv​ard.edu/​rep​osit​ory/​col​lect​ion/​resou​rce-​pack-​exten​ded-​
cost-​effect​iven​ess-​analy​sis/​
Central Statistical Agency (Ethiopia), & ICF. (2019). Ethiopia mini demographic and health
survey 2019. https://​dhs​prog​ram.com/​publi​cati​ons/​publ​icat​ion-​FR363-​DHS-​Final-​Repo​
rts.cfm
Chang, A. Y., Riumallo-​Herl, C., Perales, N. A., Clark, S., Clark, A., Constenla, D., . . . Verguet,
S. (2018). The equity impact vaccines may have on averting deaths and medical impover-
ishment in developing countries. Health Affairs (Millwood), 37(2), 316–​324. doi:10.1377/​
hlthaff.2017.0861
Chang, A. Y., Riumallo-​Herl, C., Salomon, J. A., Resch, S. C., Brenzel, L., & Verguet, S. (2018).
Estimating the distribution of morbidity and mortality of childhood diarrhea, measles, and
pneumonia by wealth group in low-​and middle-​income countries. BMC Medicine, 16(1),
102. doi:10.1186/​s12916-​018-​1074-​y
 3.4.3 Equity impact 211

Chao, F., You, D., Pedersen, J., Hug, L., & Alkema, L. (2018). National and regional under-​5 mor-
tality rate by economic status for low-​income and middle-​income countries: A systematic as-
sessment. Lancet Global Health, 6(5), e535–​e547. doi:10.1016/​S2214-​109X(18)30059-​7
Devlin, N. J., & Brooks, R. (2017). EQ-​5D and the EuroQol Group: Past, present and future.
Applied Health Economics and Health Policy, 15(2), 127–​137. doi:10.1007/​s40258-​017-​0310-​5
Drummond, M., Sculpher, M., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015a).
Measuring and valuing effects: Health gain. In Methods for the economic evaluation of health
care programmes (pp. 133–​143). Oxford: Oxford University Press.
Drummond, M., Sculpher, M., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015b).
Methods for the economic evaluation of health care programmes (4th ed.). Oxford: Oxford
University Press.
EuroQol Research Foundation. (2019). Explaining the EQ 5D in about two-​and-​a-​half minutes.
https://​www.yout​ube.com/​watch?v=​qhZ6​goeT​SLI&feat​ure=​youtu.be
Feeny, D. (2005). Preference-​based measures: Utility and quality-​adjusted life years. In P. Fayers,
& R. Hays (Eds.), Assessing quality of life in clinical trials (pp. 405–​431). Oxford: Oxford
University Press.
Gold, M. R., Stevenson, D., & Fryback, D. G. (2002). HALYs and QALYs and DALYs, oh
my: Similarities and differences in summary measures of population health. Annual Review
of Public Health, 23(1), 115–​134. doi:10.1146/​annurev.publhealth.23.100901.140513
Hammit, J. K. (2002). QALYs versus WTP. Risk Analysis, 22(5), 985–​1001. doi:10.1111/​
1539-​6924.00265
Herdman, M., Gudex, C., Lloyd, A., Janssen, M. F., Kind, P., Parkin, D., . . . Badia, X. (2011).
Development and preliminary testing of the new five-​level version of EQ-​5D (EQ-​5D-​5L).
Quality of Life Research, 20(10), 1727–​1736. http://​www.jstor.org/​sta​ble/​41488​243
Institute for Health Metrics & Evaluation. (2020). GBD compare. https://​viz​hub.hea​lthd​ata.org/​
gbd-​comp​are/​
Jacobs, J. (2018). Introduction to effectiveness, patient preferences, and utilities. https://​www.
hsrd.resea​rch.va.gov/​for_​rese​arch​ers/​cyb​er_​s​emin​ars/​archi​ves/​video_​arch​ive.cfm?Sessio​
nID=​2400
Jit, M., Hutubessy, R., Png, M. E., Sundaram, N., Audimulam, J., Salim, S., & Yoong, J. (2015).
The broader economic impact of vaccination: Reviewing and appraising the strength of evi-
dence. BMC Medicine, 13(1), 209. doi:10.1186/​s12916-​015-​0446-​9
Kind, P., Klose, K., Gusi, N., Olivares, P., & Greiner, W. (2015). Can adult weights be used to
value child health states? Testing the influence of perspective in valuing EQ-​5D-​Y. Quality of
Life Research, 24(10), 2519–​2539. doi:10.1007/​s11136-​015-​0971-​1
Lopez Bernal, J. A., Andrews, N., & Amirthalingam, G. (2019). The use of quasi-​experimental
designs for vaccine evaluation. Clinical Infectious Diseases, 68(10), 1769–​1776. doi:10.1093/​
cid/​ciy906
Manchira Krishnan, S. (2020). valueEQ5D: Scoring the EQ-​5D descriptive system. The R Project
for Statistical Computing. https://​CRAN.R-​proj​ect.org/​pack​age=​valueE​Q5D
Memirie, S. T., Metaferia, Z. S., Norheim, O. F., Levin, C. E., Verguet, S., & Johansson, K. A.
(2017). Household expenditures on pneumonia and diarrhoea treatment in Ethiopia: A
facility-​based study. BMJ Global Health, 2(1), e000166. doi:10.1136/​bmjgh-​2016-​000166
Memirie, S. T., Nigus, M., & Verguet, S. (2021). Cost-​effectiveness and equitable access to vac-
cines in Ethiopia: An overview and evidence synthesis of the published literature. Journal of
Global Health Reports, 5, e2021008. doi:10.29392/​001c.19354
Mortimer, D., & Segal, L. (2008). Comparing the incomparable? A systematic review of com-
peting techniques for converting descriptive measures of health status into QALY-​weights.
Medical Decision Making, 28(1), 66–​89. doi:10.1177/​0272989x07309642
212 Measuring and valuing health outcomes

Pliskin, J. S., Shepard, D. S., & Weinstein, M. C. (1980). Utility functions for life years and health
status. Operations Research, 28(1), 206–​224. doi:10.1287/​opre.28.1.206
Rheingans, R., Atherly, D., & Anderson, J. (2012). Distributional impact of rotavirus vaccina-
tion in 25 GAVI countries: Estimating disparities in benefits and cost-​effectiveness. Vaccine,
30(Suppl 1), A15–​A23. doi:10.1016/​j.vaccine.2012.01.018
Riumallo-​Herl, C., Chang, A. Y., Clark, S., Constenla, D., Clark, A., Brenzel, L., & Verguet, S.
(2018). Poverty reduction and equity benefits of introducing or scaling up measles, rotavirus
and pneumococcal vaccines in low-​income and middle-​income countries: A modelling
study. BMJ Global Health, 3(2), e000613. doi:10.1136/​bmjgh-​2017-​000613
Roberts, M. J., Hsiao, W., Berman, P., & Reich, M. R. (2019). Getting health reform right, anni-
versary edition. Oxford: Oxford University Press.
Rowen, D., Rivero-​Arias, O., Devlin, N., & Ratcliffe, J. (2020). Review of valuation methods of
preference-​based measures of health for economic evaluation in child and adolescent popu-
lations: where are we now and where are we going?. Pharmacoeconomics, 38(4), 325–​340.
doi:10.1007/​s40273-​019-​00873-​7
Salomon, J. A., Haagsma, J. A., Davis, A., de Noordhout, C. M., Polinder, S., Havelaar, A.
H., . . . Vos, T. (2015). Disability weights for the Global Burden of Disease 2013 study. Lancet
Global Health, 3(11), e712–​e723. doi:10.1016/​S2214-​109X(15)00069-​8
Shaw, J. W., Johnson, J. A., & Coons, S. J. (2005). US valuation of the EQ-​5D health states:
Development and testing of the D1 valuation model. Medical Care, 43(3), 203–​220. https://​
journ​als.lww.com/​lww-​medi​calc​are/​Fullt​ext/​2005/​03000/​US_​Valuation_​of_​t​he_​E​Q_​5D​_​
Hea​lth_​Stat​es_​.3.aspx
Sullivan, P. W., & Ghushchyan, V. (2006). Preference-​based EQ-​5D index scores for chronic
conditions in the United States. Medical Decision Making, 26(4), 410–​420. doi:10.1177/​
0272989x06290495
Torrance, G. W. (1982). Preferences for health states: A review of measurement methods. Mead
Johnson Symposium on Perinatal and Developmental Medicine, 20, 37–​45. http://​europe​pmc.
org/​abstr​act/​MED/​6927​515
Torrance, G. W., Feeny, D., & Furlong, W. (2001). Visual analog scales: Do they have a role in the
measurement of preferences for health states? Thousand Oaks, CA: Sage Publications.
Tufts Medical Center. (2021). Global health cost-​effectiveness analysis registry. http://​ghcear​egis​
try.org/​ghcear​egis​try/​
van Reenen, M., & Janssen, B. (2015). EQ-​5D-​5L user guide: Basic information on how to use the
EQ-​5D-​5L instrument. Rotterdam: EuroQol Research Foundation.
Verguet, S. (2018, July 21). The equity and poverty reduction benefits of vaccines. Harvard
Health Policy Review. http://​www.hhp​ronl​ine.org/​artic​les/​2018/​7/​21/​the-​equ​ity-​and-​pove​
rty-​reduct​ion-​benef​i ts-​of-​vacci​nes
Verguet, S., Kim, J. J., & Jamison, D. T. (2016). Extended cost-​effectiveness analysis for
health policy assessment: A tutorial. PharmacoEconomics, 34(9), 913–​923. doi:10.1007/​
s40273-​016-​0414-​z
Verguet, S., Murphy, S., Anderson, B., Johansson, K. A., Glass, R., & Rheingans, R. (2013).
Public finance of rotavirus vaccination in India and Ethiopia: An extended cost-​effectiveness
analysis. Vaccine, 31(42), 4902–​4910. doi:10.1016/​j.vaccine.2013.07.014
Walker, D. G., Hutubessy, R., & Beutels, P. (2010). WHO guide for standardisation of eco-
nomic evaluations of immunization programmes. Vaccine, 28(11), 2356–​2359. doi:10.1016/​
j.vaccine.2009.06.035
Weinstein, M. C. (1988). A QALY is a QALY—​or is it? Journal of Health Economics, 7(3), 289–​290.
doi:10.1016/​0167-​6296(88)90030-​6
World Health Organization. (2019a). Global monitoring report on financial protection in health 2019.
https://​www.who.int/​hea​lthi​nfo/​univer​sal_​heal​th_​c​over​age/​rep​ort/​fp_​g​mr_​2​019.pdf?ua=​1
3.5
Reporting and interpreting results
of economic evaluation
Ijeoma Edoka, Carleigh Krubiner, Andrew Mirelman, R. Brett McQueen,
Mark Sculpher, Julia F. Slejko, and Tommy Wilkinson

3.5.1 Understanding a cost-​effectiveness

analysis plane

A cost-​effectiveness analysis (CEA) plane is a common way to visualize the re-

sults of a vaccine CEA. A CEA plane is a graph where each intervention in the
analysis can be plotted with costs on the vertical (y) axis and effects on the hori-
zontal (x) axis. While the incremental cost-​effectiveness ratio (ICER) is the key
summary metric of interest when reporting results, a CEA plane can display the
expected costs and effects of all included interventions relative to a common
comparator and is a useful way of communicating results to a nontechnical
audience.
To populate a CEA plane, each intervention is plotted according to calculated
costs and effects. The effects are the outcomes generated by the CEA and can be
natural units (e.g., infections averted) or generalizable units (e.g., the disability-​
adjusted life year (DALY) averted or quality-​adjusted life year (QALY) gained).
Using a generalizable outcome measure is the most common approach and al-
lows a CEA plane to be interpreted relative to a cost-​effectiveness threshold
(CET) and other competing health systems investments. The costs reflected for
each intervention are “net” and incorporate costs incurred from implementa-
tion of the intervention and any costs savings.
An example results table is shown (Table 3.5.1) with the corresponding
CEA plane (Fig. 3.5.1). The costs and effects in Table 3.5.1 are relative to the
status quo or existing treatment, and status quo is represented at the origin of
the CEA plane in Fig. 3.5.1.

Ijeoma Edoka, Carleigh Krubiner, Andrew Mirelman, R. Brett McQueen, Mark Sculpher, Julia F. Slejko, and
Tommy Wilkinson, Reporting and interpreting results of economic evaluation In: Handbook of Applied Health Economics in
Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0018
214 Reporting and interpreting results of economic evaluation

Table 3.5.1 Hypothetical results table for a corresponding case

study of results for cost-​effectiveness analysis of vaccines and
competing public health strategies

Alternative Costs ($) Effectiveness

(QALYs)

Vaccine A 10 3.0
Vaccine B 20 4.0
Vaccine C 50 5.0
Intervention D −40 3.0
Intervention E −20 −3.0
Intervention F 10 −5.0

The CEA plane typically consists of four quadrants, called the North West
(NW), North East (NE), South West (SW), and South East (SE). As the or-
igin of the CEA plane represents the status quo, interventions plotted in the
NE quadrant (Vaccines A, B, and C) are more expensive and more effective
than current treatment, whereas an intervention plotted in the SE quadrant

NW NE

$50 C

$25

F A
Cost ($)

‒5.0 ‒2.5 (0.0, $0) 2.5 5.0

E
ss Y)
e ne AL $‒25
iv /Q
ct ($
e ffe old
-
st esh
Co Thr D
$‒50

SW SE

Effectiveness (QALYs)

Fig. 3.5.1 Cost-​effectiveness plane of a sample vaccine case.

 3.5.1 Understanding a cost-effectiveness analysis plane 215

(Intervention D) is expected to be more effective and less costly than current

treatment. An intervention that is plotted in the SW quadrant (Intervention
E) is expected to be less costly and less effective than current treatment; and an
intervention that is plotted in the NW quadrant (Intervention F) is expected
to be more costly and less effective than current treatment.
In Table 3.5.1, Vaccine A achieves three units of health for $10 compared
to current treatment, while Vaccine B achieves an additional unit of health
for an additional $10 compared to Vaccine A, and Vaccine C achieves an ad-
ditional unit of health for an additional $30 compared to Vaccine B. Plotting
a line from the origin (status quo) to Vaccine A, to Vaccine B, and to Vaccine
C is called the cost-​effectiveness frontier (gray dotted lines). In Fig. 3.5.1, the
cost-​effectiveness frontier is becoming steeper as we move from Vaccine A to
B to C indicating that the incremental cost per unit of health gain (the ICER)
is increasing. The CET can also be plotted on a CEA plane and is shown in as a
dotted diagonal line in Fig. 3.5.1 at a hypothetical vale of $10 per unit of health
gain. Where the cost-​effectiveness frontier is shallower than the CET then an
intervention would be considered cost-​effective, and an intervention with a
CEA frontier steeper than the CET would not be considered cost-​effective.
A policy recommendation based on cost-​effectiveness would recommend an
intervention with the highest ICER that remained at or equal to the CET to
achieve optional spending across the health system.
In Fig. 3.5.1, the CEA frontier moving from the origin to Intervention
A is shallower than the CET, the CEA frontier moving from Vaccine A to
Intervention B is the same as the CET, and the CE frontier moving from Vaccine
B to Vaccine C is steeper than the CET. The interpretation is that Vaccine C is
not cost-​effective, and that Vaccines A and B represent potential policy options.
As the ICER for Vaccine B is equal to the CET, Vaccine B would be the recom-
mended policy option based on cost-​effectiveness criteria. As it is common for
a CEA to be conducted on interventions that are expected to cost more but will
be more effective than current treatment, frequently only the NE quadrant of a
CEA plane will be reported in a published CEA. However, many CEA results will
involve interventions that are plotted in the other quadrants of the CEA plane.
The interpretation of results for interventions plotted in either the SE or
NW quadrants are generally straightforward. As Intervention D is less costly
and more effective than current treatment, it can be considered cost-​effective
regardless of any CET or willingness to pay for health gain. Conversely,
Intervention F would not be considered cost-​effective even at a very high CET
as it achieves less health compared to status quo at a higher cost.
Interventions in the SW quadrant often represent a challenging area for health
policy decision makers. In Fig. 3.5.1, Intervention E is both less costly but less
216 Reporting and interpreting results of economic evaluation

effective than status quo. This means that Intervention E represents a “disinvest-
ment” option where implementing Intervention E and stopping provision of the
status quo would lead to lower direct health for the immediate patient popula-
tion. However, this could be considered a cost-​effective disinvestment decision
if the money saved (or yielded) by investing in Intervention E relative to status
quo could be invested elsewhere in the health system, thereby generating more
population health overall. The point at which a vaccine would yield more popu-
lation health than is lost as a result of disinvesting the status quo is also informed
by the CET, where interventions below the CET would yield positive overall
population health and would be considered a cost-​effective treatment option. In
Fig. 3.5.1, Intervention E is above the CET and would not be considered a cost-​
effective option as the health losses associated with implementation would not
be fully offset by the population health gains achievable from cost savings rein-
vestment in the health system. CEA can be an extremely useful tool in informing
health system disinvestments; however, the policy decision would need to take
further non-​efficiency factors into consideration including the rights of patients
to continue therapies from which they are benefiting and the feasibility, uncer-
tainty, and risks of implementing a known less effective therapy.
Fig. 3.5.1 displays deterministic results of a CEA and each intervention is
represented as a single point plotted on the graph. A CEA plane is also effective
as visualizing the results of probabilistic sensitivity analysis (see Chapter 4.3),
where each intervention is represented by a field of plotted points, each point
representing a single simulation within the probabilistic sensitivity analysis.
Viewing probabilistic sensitivity analysis results on a CEA plane shows the
range of uncertainty associated with an intervention relative to the CET and
other mutually exclusive interventions.
This section has introduced the CEA plane and described how results
plotted on the CEA plane can be interpreted for vaccine value. The CEA plane
remains the standard approach for reporting CEA results and is a powerful
mechanism for communicating results of an analysis to policymakers and
nonexpert audiences.

3.5.2 Cost-​effectiveness thresholds for

vaccine economics

3.5.2.1 Background and principles

In the process of conducting CEAs for vaccines (or other health technologies
and interventions), a CET can provide guidance on whether an investment or
 3.5.2 Cost-effectiveness thresholds for vaccine economics 217

disinvestment in that intervention is cost-​effective or not. CETs can be used

by analysts, or in decision-​making processes such as health technology assess-
ment, which are established to make investment and disinvestment decisions
for packages of health services across a health system.
In its basic form, a CET can be conceptualized in a fraction format of the
value per health benefit gained, which represents a level above or below that
something is considered cost-​effective. The value in the numerator is a com-
monly understood currency, while the value in the denominator is a summary
measure of health such as a QALY gained or a DALY averted. CET is used in-
terchangeably with another term introduced earlier, the willingness-​to-​pay
threshold (see Chapter 3.1).
When the incremental costs per QALY or DALY of a new intervention com-
pared to its next best alternative, expressed as an ICER, falls below the CET,
then the intervention is regarded as representing good value for money and is
recommended for funding. However, an ICER that falls above the threshold
implies that investing in the new intervention does not constitute good value.
CETs are therefore an important element for making resource allocation de-
cisions and have been widely used in several countries. They are generally
viewed as a means for improving efficiency in the allocation of scarce re-
sources as well as for ensuring consistency, transparency, and public support
in the decision-​making process. This is particularly important when unpop-
ular decisions must be made.

3.5.2.2 What do cost-​effectiveness thresholds represent?

The two perspectives of what CETs should represent are often referred to as
the demand-​side and supply-​side perspectives. Demand-​side thresholds re-
flect the monetary value society places on additional health gains while for-
going gains from the consumption of non-​health goods. In other words, this
perspective suggests that thresholds should represent society’s willingness to
pay for additional health gains, thus reflecting consumer demand for health-
care services. Many commonly used thresholds typically reflect this perspective
and are either based on value judgments or derived empirically by estimating
society’s willingness to pay for additional health gains (Vallejo-​Torres, García-​
Lorenzo, & Serrano-​Aguilar, 2018). One common example that had been in
favor until recently was the Commission on Macroeconomics and Health’s rule
of thumb threshold of one to three times gross domestic product (GDP) per
capita, based on value judgments on economic gains accruing from averting a
DALY (Hutubessy, Chisholm, & Edejer, 2003). In Thailand, the CET used for
218 Reporting and interpreting results of economic evaluation

reimbursement decisions is based on empirical estimates of society’s willingness

to pay for a QALY (Teerawattananon, Tritasavit, Suchonwanich, & Kingkaew,
2014; Thavorncharoensap et al., 2013).
Supply-​side thresholds on the other hand, reflect the opportunity costs
resulting from funding decisions when health budgets are constrained.
Proponents of the supply-​side perspective argue that when health budgets
are fixed, decisions to introduce a new intervention should be made based on
health gains displaced elsewhere in the system in order to free up resources for
the new intervention. When the health displaced elsewhere in the system out-
weighs health gained from funding the new intervention under evaluation,
the new intervention is not worth doing. The CET should therefore reflect the
value of health given up or the health opportunity cost of health spending.
This perspective explicitly takes into account the budget available to the health
system and/​or the efficiency with which the health system is capable of pro-
ducing health. Thresholds can be useful tools, and when they are explicitly set,
decision-​making can also be held accountable to this explicit component.
However, CETs should not be applied as the sole decision-​making criteria.
One reason could be that the CET itself could be inaccurate or not fully reflect
different consequences of a decision. For example, the GDP multiplier-​based
thresholds, have been widely criticized as not suitable for health decision-​
making. Alternatively, a threshold estimate based on the supply-​side op-
portunity cost would needs extensive data to be identified, or else it would
be difficult to reflect the health consequences of the decision being made.
Therefore, an appropriate CET is only helpful as an aid to decision-​making if it
reflects the best available evidence and is incorporated in a broader decision-​
making process that is both transparent and accountable.

3.5.2.3 Identifying cost-​effectiveness thresholds

A CET may be estimated with several techniques. One approach is to use

mathematical modeling to maximize the health objective function subject to
the budget constraint (Stinnett & Paltiel, 1996). More commonly, a causal link
between expenditure and health is identified using either cross-​country var-
iation, or within-​country variation to estimate the impact of relative amount
of health that is given up (or gained) when there is new investment (or dis-
investment). Estimating a threshold using within-​country data can be inten-
sive. This is because data are needed that establish the opportunity cost of the
system with the causal link between expenditure changes and health changes.
 3.5.2 Cost-effectiveness thresholds for vaccine economics 219

Table 3.5.2 Examples of cost-​effectiveness thresholds identified with quantitative

approaches

Example Income group Threshold Explicit threshold used Source

country estimated

Thailand Middle-​ Baht 240,000 Baht 160,000 Nimdet &

income per QALY Ngorsuraches
country (2015)
England High-​income £5,000—​£15,000 £20,000—​£30,000 Lomas et al.
country per QALY NICE recommendation (2019)
US High-​income US $100,000 US $50,000—​$150,000 Vanness et al.
country per QALY per QALY (2021)
No exact threshold
recognized

Within country approaches have been used in the UK and Spain, and there
have been attempts to use within-​country approaches in low-​and middle-​
income countries (LMICs) as well, such as South Africa (Claxton et al., 2015;
Edoka & Stacey, 2020; Vallejo-​Torres et al., 2018; Vanness, Lomas, & Ahn,
2021). However, given the need for high-​quality within-​country data for
these approaches, other approaches have been used that incorporate cross-​
country data (Ochalek, Lomas, & Claxton, 2018; Woods, Revill, Sculpher, &
Claxton, 2016).
Table 3.5.2 provides information to shows three examples of countries
with explicitly defined thresholds used in health decision-​making. With
the exception of the US, the other countries—​Thailand and England—​are
two of the few with explicitly defined thresholds (Lomas, Martin, & Claxton,
2019; Schwarzer et al., 2015). The estimated values of the thresholds that are
estimated in studies can vary from those that are used in practice.

3.5.2.4 Alternative approaches?

A review of CEA studies in LMICs identified that thresholds related to GDP

per capita are the most widely used in cost-​effectiveness studies in LMICs
(Leech, Kim, Cohen, & Neumann, 2018). However, while these types of
thresholds are widely cited in cost-​effectiveness studies, they are often mis-
used and overused. The current recommendation is to rely on context-​specific
thresholds that are not based on GDP per capita to inform decision-​making
(Bertram et al., 2016).
220 Reporting and interpreting results of economic evaluation

Given that it may either be difficult to identify a CET with quantitative ap-
proaches, or that other across-​country CETs may not be suitable for national
health decision makers, a recent paper convened by a group of experts at
Bellagio identified other approaches to making decisions when an explicitly
defined CET is not possible. These approaches include looking at decisions in
other settings, looking at evidence of affordability (budget impact analysis),
and developing league tables (Chi et al., 2020). One should note, however, that
these approaches do not necessarily substitute for using CETs, and many of the
mentioned approaches can be part of a broad-​based decision-​making process.

3.5.2.5 Cost-​effectiveness thresholds and vaccines

Vaccines, in principle, are the same as any other health intervention. They
have a cost, an associated health impact, and decisions need to be made
about whether they should be introduced or removed. Vaccines, however,
can also have important multisectoral impacts outside of the health sector
(Bärnighausen et al., 2014). This means that in addition to considering a
health-​specific CET, such as the ones referred to above, one would also
need to consider decision rules in other non-​health sectors. This is not a
straightforward task given that there may be trade-​offs between sectors that
are important to consider and that the measures of impact between sectors
may differ.
In the UK, the Joint Committee on Vaccination and Immunisation makes
decisions around vaccine introduction, and it uses the same threshold as
that used by the National Institute for Health and Care Excellence (NICE).
However, there have been recommendations to bring this value lower so it
is supported by an evidence-​based value of opportunity cost (Table 3.5.2)
(O’Mahony & Paulden, 2019).
In LMICs, there are still many studies that use GDP per capita-​based
thresholds when making conclusions about cost-​effectiveness (Fesenfeld,
Hutubessy, & Jit, 2013). Another approach is to consider many different types
of thresholds and to let decision makers make an assessment based on what
they consider to be appropriate. Loganathan et al. (2018) do this in a study
of rotavirus vaccination in Malaysia. Regardless of the threshold used, it is
important to remember that economic evaluations for vaccines (and indeed
other interventions) should encompass the principles of timeliness, inte-
gration, quality, and ownership/​institutionalization (Jauregui et al., 2015).
See Box 3.5.1.
 3.5.2 Cost-effectiveness thresholds for vaccine economics 221

Box 3.5.1 Applying CETs in an unequal world: is it ethical to

have different thresholds for different global contexts?

CEA remains a critical tool to navigate the complexities of priority setting for health
and ensure that limited resources are used effectively and efficiently to protect and
promote well-​being. Yet, the use of CEA and health economics more broadly has been
subject to various criticisms, including that it commodifies health in ways incon-
sistent with conceptions of health as a human right and that it reinforces inequities in
health (Farmer, 2015; Meyer, 2013). However, properly understood, the realization of
the right to health is unattainable without explicit priority setting subject to resource
availability (Rumbold et al., 2017). Moreover, inefficient allocations often favor the
interests of the privileged few while diverting resources away from cost-​effective serv-
ices that would most improve the health of the politically and economically disadvan-
taged. Lastly, it is worth underscoring that cost-​effectiveness alone cannot capture all
morally relevant considerations to guide decision-​making (Krubiner & Faden, 2017;
Norheim et al., 2014; World Health Organization, 2014a). In any context, there may be
compelling and principled reasons why a health intervention that fails to meet a CET
should still be covered. The use of a threshold, however, ensures that health opportu-
nity costs are considered in the priority-​setting process and that allocations resulting
in lesser aggregate health gains are reasonably justified by competing moral claims.
Regarding CETs that vary based on the resource constraints in a given country, one
can understand the perception that, on its face, this approach calculates the “value of
life” differently in rich versus poor settings. However, this interpretation fails to recog-
nize that the true aim of a CET is not to set a price on the value of individuals’ health,
but rather to help avoid morally relevant health opportunity costs in which more
people suffer from ill health and premature death as a result of a short-​sighted invest-
ment in an expensive health technology with clinical benefits for a limited proportion
of the population (Chi et al., 2020; Ochalek et al., 2018; Revill et al., 2014). Countries
must operate within a set health budget, and make tough decisions about how to
allocate domestic resources. A CET can help policymakers identify the set of health
interventions that will produce the greatest health gains for their populations for a set
of budget constraints.
Consider, for example, policy decisions about seasonal influenza immunization,
including which vaccines to procure and which populations should be included in
annual vaccination campaigns (Clements, Chancellor, Nichol, DeLong, & Thompson,
2011; Hendriks et al., 2018; Jit, Newall, & Beutels, 2013). National immunization
strategies vary widely across global contexts, with some high-​income countries like
the US electing for universal flu immunization while most LMICs opt for targeted
222 Reporting and interpreting results of economic evaluation

strategies, typically focused on populations at greatest risk of bad outcomes (e.g.,

those who are elderly, pregnant, or have underlying conditions) or on those who
contribute most to influenza transmission (e.g., children) (Principi, Camilloni, &
Esposito, 2018). Yet, cross-​country variability of influenza immunization policy is
not necessarily a sign of inequity, and contextualized CEA can help policymakers
appropriately take into account the actual burden of influenza in their countries,
the costs of influenza-​like illness for those affected, and the associated costs and
benefits of expanding vaccination program coverage (World Health Organization,
2016, 2020c). What may be a cost-​effective strategy in the US or UK in light of the
epidemiological and economic burden of influenza, the expected impact of the vac-
cination strategy, and the overall available resources for domestic health, could be
entirely inappropriate in other country settings, where significant health gains can
be realized through targeting immunization while reserving the additional funds
and resources it would take to reach wider coverage for alternative investments in
more cost-​effective health interventions (Edoka et al., 2021). There have also been
recent debates about the comparative cost-​effectiveness of different influenza vac-
cine products, namely the value for money of moving from tetravalent to quadriva-
lent influenza vaccines, in LMICs—​where the incremental cost and budget impact
may not be justified by corresponding health gains, not to mention feasibility con-
cerns given supply-​side constraints and the need for seasonal timeliness of delivery
(Teerawattananon et al., 2014). The case of seasonal influenza vaccination strategies
illustrates the need for context-​specific economic evaluation and consideration of
the epidemiological, economic, and health system realities to guide local policy.
Additionally, that richer nations can afford to spend more on health is not a
problem of thresholds—​it is a reflection of the larger backdrop of global and struc-
tural injustice, which would not be solved by setting a universal threshold for what
is considered “good value for money” across different country settings. Instead, an
appropriate threshold for poorer countries to optimize domestic health spending
could more clearly shape where global health aid from bilateral and multilateral
partners is most needed to supplement national spending. With domestic resources
reserved for the most cost-​effective interventions, global health aid could be priori-
tized for expanding coverage to less affordable interventions, both those that are high
priorities within countries and for donor objectives that may not align perfectly with
countries’ most dire needs. For example, in the context of vaccination programs, this
could translate to greater development assistance for costly vaccines crucial to global
health security efforts (e.g., safeguarding against pathogens with pandemic poten-
tial like coronavirus disease 2019 (COVID-​19) or Ebola through contributions to Gavi
and the COVID-​19 Vaccines Global Access (COVAX) facility), while enabling LMICs to
 3.5.3 ICER calculation and determining dominance 223

allocate domestic resources for more cost-​effective, affordable vaccines addressing

persistent endemic threats.
It should also be noted that CETs can play an instrumental role in facilitating afford-
able and equitable introduction of new vaccine products into lower-​income markets,
whether through bilateral price negotiations using the threshold as a reference point
(Glassman, Cañón, & Silverman, 2016; Teerawattananon & Tritasavit, 2015) or through
more innovative approaches that leverage tiered pricing models (Berkley, 2014). and
other market-​shaping mechanisms like benefits-​based advanced market commit-
ments (Chalkidou, Kettler, Ramakrishnan, Silverman, & Towse, 2020). So rather than
being contrary to the aims of global health equity, context-​appropriate CETs can pro-
mote earlier introduction of new vaccines into LMICs in ways that are more affordable
and sustainable.

3.5.3 Incremental cost-​effectiveness ratio calculation,

incremental ratios, and dominance

The ICER and related calculations such as net health benefit (NHB) and net
monetary benefit (NMB) aid decision makers in determining which vaccine
program or treatment strategy in question is the most “cost-​effective.” ICERs
and related net benefit calculations depend on monetary estimates in the
numerator and a common effectiveness or benefit measure in the denomi-
nator. Researchers often use ICERs in practice as a pairwise measure, com-
paring a new treatment to a baseline comparator regardless of the number
of strategies available to clinicians and decision makers. The ICER is a ratio
measure and interpretation requires a reference to a certain willingness-​to-​
pay threshold to determine whether the additional health purchased with the
intervention in question is of good value. ICER calculations differ slightly in
reference to whether there are two strategies being compared to each other or
more than two strategies compared to a baseline comparator (Gray, Clarke,
Wolstenholme, & Wordsworth, 2011).

3.5.3.1 Two or more vaccine strategies, treatment

programs, and so on

In the case of two strategies, the ICER calculation is as simple as it appears

in Equation 3.1.4 in Chapter 3.1, such as in a classic example of a vaccine
224 Reporting and interpreting results of economic evaluation

(A) versus a treatment (B). In the numerator, the cost of A and B are meas-
ured and incrementally compared based on direct and non-​direct medical
costs, along with indirect costs such as lost productivity in cases where a so-
cietal perspective is relevant for the decision problem. In the denominator,
the benefit of A and B are often measured and incrementally compared in
terms of the QALYs as it is a common metric that can be compared across
disease states.
However, there are other measures of benefit depending on the decision
problem. In the case of two strategies, the ICER interpretation depends
on the sign (+​or −) of both the numerator and denominator along with
a willingness-​to-​pay threshold. In the case of more than two strategies,
ICERs are often still compared to a baseline comparator which is often de-
termined from clinical guidelines and real-​world utilization, among other
factors. Interpretation for two or more comparators is informed by various
visual displays of ratio estimates such as the CEA plane, explained in detail
in section 3.5.1 and again in section 3.5.5. However, interpreting the cost-​
effectiveness of more than two strategies involves consideration of domi-
nance and extended dominance.

3.5.3.2 Dominance and extended dominance

When considering more than two strategies, certain treatments may be less
or more effective and less or more costly than other treatments. Certain treat-
ments may be “dominated” in that they have either higher costs and smaller
or equal benefit than at least one other treatment being considered; or smaller
benefits and higher or equal costs than at least one other strategy being con-
sidered (Gray et al., 2011; Paulden, 2020). For example, consider the polio vac-
cine, which prevents a condition that could cost tens of thousands of dollars
to manage and severely depletes a patient’s health utility, whereas the cost of
the polio vaccine is relatively low—​a few dollars per dose to administer—​and
provides lifelong immunity against this debilitating disease. Thus, the polio
vaccine dominates disease management since it comes at a lower cost for in-
creased clinical benefit. Similarly, “extended dominance” is an important con-
cept since it rules out any intervention that has an ICER that is greater than
that of a more effective alternative.
To illustrate dominance and extended dominance, consider the following
example with three vaccines (A, B, and C) compared to a reference standard of
care (Table 3.5.3).
 3.5.4 Defining uncertainty of value 225

Table 3.5.3 Evaluating dominance

Alternative Cost ($) Effectiveness (QALY) ICER ($/​QALY)

Vaccine A 40 3.0 20
Vaccine B 50 4.0 10
Vaccine C 5 2.0 −5
Standard of care 10 1.0 Reference

First, Vaccine C dominates standard of care since it comes at a reduced cost

compared to the reference standard of care, −$5, and improves QALYs gained
by +​1.0. For LMICs with limited financial bandwidth, Vaccine C provides an
appealing solution to the infectious disease outbreak at hand. For countries
in a better financial position, they may be willing to pay more to gain more
QALYs, making both Vaccines A and B attractive since they deliver superior
QALYs to standard of care. However, we can still use extended dominance to
rule out Vaccine A since it has an ICER of $20 per QALY, whereas Vaccine B
provides greater QALYs at a lower ICER. Since Vaccine B generates fewer op-
portunity costs per QALY, it represents an efficient option for communities
that have the budget to invest in it.
The general steps to calculate ICERs and determine the most cost-​effective
strategy involves the following:

• Step 1: rank strategies by cost or benefit.

• Step 2: rule out dominated strategies (less effective, more costly than
other treatments).
• Step 3: calculate incremental estimates.
• Step 4: rule out extendedly dominated strategies.
• Step 5: recalculate incremental values.
• Step 6: rule out final extendedly dominated strategies.

While a detailed discussion of each step in context to an example is out of the

scope of this handbook, interested readers can review literature in this area
that provides helpful examples (Paulden, 2020).

3.5.4 Defining uncertainty of value

The assessment of uncertainty in reference to a decision problem is a standard

and expected step in economic evaluation. There are a wide range of terms
226 Reporting and interpreting results of economic evaluation

used interchangeably given the interdisciplinary nature of economic evalu-

ation. In this section of this chapter, we will define uncertainty in context to
other terms used in the field and provide examples of appropriate assessment
of uncertainty.

3.5.4.1 Defining uncertainty

As many economic evaluations are designed to assist decision makers with

maximizing health gains subject to economic constraints, assessing uncer-
tainty is important for revealing potential deviations from the “best” course
of action (Briggs et al., 2012). Specifically, uncertainty assessment informs
two potential useful criteria for decision makers: (1) the confidence in
selecting a particular treatment strategy; and (2) the value of collecting ad-
ditional information, such as an additional trial or study on the treatment in
question (Briggs et al., 2012). While there are different concepts related to
uncertainty, the assessment of uncertainty should always be taken in con-
text with the decision maker’s perspective to best inform present and future
decision-​making.

3.5.4.2 Distinguishing between uncertainty, heterogeneity,

and variability

The joint International Society for Health Economics and Outcomes

Research (ISPOR)–​Society for Medical Decision Making task force defines
four different distinguishable terms related to uncertainty: stochastic un-
certainty, parameter uncertainty, heterogeneity, and structural uncertainty.
Stochastic uncertainty relates to random variability in outcomes between
the same patients which is also known as first-​order uncertainty. For ex-
ample, the response of one treatment may differ among patients with very
similar demographic characteristics. Parameter uncertainty or second-​
order uncertainty is the uncertainty in estimation of a particular input pa-
rameter itself. Parameter uncertainty often arises when data to generate an
input parameter differs in terms of sample size, study population, external
validity, and so on.
Heterogeneity is defined by the variability between patients that can be
defined through measured characteristics such as age, sex, history of co-
morbid conditions, and so forth. Finally, structural uncertainty refers to the
 3.5.4 Defining uncertainty of value 227

underlying assumptions of the decision analysis model. In other words, does

the analysis reflect the clinical plausibility and course of outcomes both with
and without the treatment in question? Structural uncertainty may require
input from clinicians and a literature review of the possible course of action
for those with and without a disease.

3.5.4.3 Uncertainty analyses

While a researcher may assess concepts such as heterogeneity and variability

in specific cases, all decision analysis modeling exercises require parameters
that must be estimated and varied in uncertainty analyses. This step involves
varying parameters across a plausible or evidence-​based range and analyzing
the impact on both cost and effectiveness outcomes. There are multiple uncer-
tainty analyses but most economic evaluations include the following: one-​way
sensitivity analyses, two-​way sensitivity analyses, and probabilistic sensitivity
analyses.
One-​way sensitivity analyses vary individual parameters while holding all
other parameters constant. This exercise is useful to identify individual param-
eters that have a large impact on model results. The presentation of one-​way
sensitivity analyses most commonly comes in the form of a tornado diagram
(Fig. 3.5.2). A tornado diagram presents each input parameter separately
starting with the parameter that has the most impact on the model results.
Lower in the tornado diagram, each input’s impact on the economic evalua-
tion results becomes smaller leading to a tornado or funnel format. It may also
be useful to include a description of how conclusions may change with large
uncertainty around certain input parameters. For example, if varying the treat-
ment effect of an intervention, does the ICER remain above or below a certain
willingness-​to-​pay threshold? It may be that small changes in an input param-
eter alter the conclusions and should be documented in the dissemination of
the uncertainty analysis.
Two-​way sensitivity analyses extend the one-​way sensitivity analysis by
varying two parameters simultaneously and assessing the impact on the
results. For example, a decision maker may want to understand how both
changes in price and effectiveness would impact the ICER. This may be
presented in a table or figure format. Finally, probabilistic sensitivity ana-
lyses vary multiple input parameters simultaneously to assess the impact
on the results. Probabilistic sensitivity analyses are presented in detail in
Chapter 4.3.
228 Reporting and interpreting results of economic evaluation

Incremental cost-effectiveness ratio

(Icer, $/QALY)
200,000 150,000 100,000 50,000 0 50,000 100,000 150,000

Treatment efficacy

Treatment cost

Social distancing cost

Rate of vaccination

Vaccine efficacy

Probability of infectivity

Vaccine cost

Number of social contacts

ICER−25% ICER+25%

Fig. 3.5.2 Example tornado diagram of one-​way sensitivity analyses for vaccine
coverage for patients at-​risk of COVID-​19.
Source: Padula, Malaviya, Reid, Tierce, and Alexander (2020).

3.5.4.4 Data sources and distributions for

parameter uncertainty

As discussed in Chapter 3.3, there are a variety of methods to incorporate un-

certainty in input parameters. However, selecting the best available evidence
in the literature can often involve subjective assessments of prior study de-
signs. Input parameters ideally will include a point estimate and uncertainty
around that estimate such as a 95% confidence interval. Input parameters ide-
ally should be from the most recent and least biased sources of information.
Many researchers employ a parameterized distribution of values as an input to
a model (Briggs et al., 2006). Choosing a distribution can be useful for setting
up a probabilistic analysis or even specifying the range of plausible values for
one-​and two-​way sensitivity analyses.

3.5.5 Interpreting the incremental

cost-​effectiveness ratio

3.5.5.1 Identifying the single most cost-​effective strategy

After calculating the ICER, we can use the sign of both the cost and benefit
estimates as well as visual plots to help inform the most cost-​effective strategy.
 3.5.5 Interpreting the incremental cost-effectiveness ratio 229

When comparing two strategies, there are four possibilities in terms of incre-
mental cost and benefit on the CEA plane (Fig. 3.5.1):

• NE (upper-​right) quadrant of the CEA plane: +​incremental cost and +​

incremental benefit.
• SW (lower-​left) quadrant of the CEA plane: − incremental cost and − in-
cremental benefit.
• SE (lower-​right) quadrant of the CEA plane: − incremental cost and +​in-
cremental benefit.
• NW (upper-​left) quadrant of the CEA plane: +​incremental cost and −in-
cremental benefit.

While interpretation of ICERs requires a reference to the CEA plane, other

measures of economic value may be useful for identifying a ranking of the
most and least cost-​effectiveness strategies when more than two strat-
egies exist.

3.5.5.2 Ranking strategies from most to least cost-​effective

While ICERs are often used to identify the single most cost-​effective
strategy, NHB or NMB may be more useful in cases with multiple treat-
ments or interventions (Paulden, 2020; Stinnett & Mullahy, 1998). Some of
the limitations of ICERs when identifying a ranking of cost-​effectiveness
strategies include leaving no “cost-​effective” strategies after ruling out dom-
inance and extended dominance as discussed earlier and the inability to
identify the most cost-​effective strategies when reviewing higher or lower
ICERs (Paulden, 2020).
However, both NHB and NMB require that a willingness-​to-​pay threshold
is specified. NHB and NMB are similar estimates of economic value but NHB
is defined in units of health whereas NMB is defined in monetary units. The
NHB has two main components: (1) the health benefit for patients who re-
ceive the treatment in question; and (2) the health loss experienced by other
patients given the new treatment in question is covered, leaving fewer resources
for health gains in other areas. To calculate NHB we first take the incremental
gains in effectiveness (e.g., QALYs) less the ratio of incremental costs to the
willingness-​to-​pay threshold (see Equation 3.1.6 in Chapter 3.1).
A positive NHB implies that overall population health would be in-
creased if the new treatment were covered while a negative NHB implies
the health benefits of the new treatment are not enough to outweigh the
health losses that would arise from not funding other treatments. NMB
230 Reporting and interpreting results of economic evaluation

uses a similar approach, however, the calculation results in a monetary

value. NMB is calculated by multiplying the incremental health benefit by
the willingness-​to-​pay threshold minus the incremental cost (see Equation
3.1.5 in Chapter 3.1).
A positive NMB indicates the treatment in question is cost-​effective or the
cost to derive the benefit is less than the maximum a decision maker is willing
to pay for the benefit. NHB and NMB overcome some of the challenges in
ICER interpretation because of the simplicity of the measure: the greater the
NHB and NMB for a particular treatment, the more cost-​effective that treat-
ment is. This intuitive interpretation assists researchers and decision makers
with identifying the most cost-​effective strategies when two or more treat-
ments are in question.

3.5.6 Limitations of economic evaluation methods

for vaccine distribution

ICERs are defined by their use of a common denominator to measure change

in effectiveness or change in utility. When the denominator measure is the
QALY, it is widely considered to reflect incremental utility gain, as a metric
comprising quality and quantity of life. This may be reported as the output of
a cost–​utility analysis, and is also accepted to be a type of CEA. This and other
utility metrics, such as DALYs, are introduced in Chapter 4.2. Despite their
wide use grounded in the economic concepts of utility and welfare economics,
there are numerous concerns related to the use of QALYs (Nord, Daniels, &
Kamlet, 2009) as discussed in Chapter 3.4.
Despite these chief concerns, the advantages of comparability among
studies in the same therapeutic area and also across therapeutic areas facili-
tates resource allocation. Economic evaluation, and cost–​utility analysis in
particular, recognized the QALY as a preferred measure for clinical effective-
ness. The US Panel on Cost-​effectiveness in Health and Medicine concluded
that QALYs are appropriate effectiveness measures by expressing health status
as a function of physical and mental components that impact survival and life
expectancy (Sanders et al., 2016). On the other hand, the Patient-​Centered
Outcomes Research Institute statutes explicitly prohibit any research “that
discounts the value of a life because of an individual’s disability,” which im-
plicates the QALY (Padula & McQueen, 2019; US Department of Health and
Human Services, 2010). Thus, the use of ICER measures over recent decades
has been supported with newly emerging measures to enhance, supplement,
 3.5.6 Limitations of economic evaluations 231

or replace these measures. There are numerous current efforts to improve on

or replace QALYs to better reflect individual preferences and address many of
the equity concerns.
There are ethical concerns around the use of QALYs and how they are used
in decision-​making(Ashcroft, 2005; Schneider, 2022). It is important to con-
sider how individuals with worse-​off baseline health can also gain health
benefit as measured through the QALY, and therefore value, as a result of in-
tervention than others with mild conditions. This raises another area of con-
cern related to equity issues with CET. This threshold estimates what a user of
healthcare services may be prepared to pay for a health benefit, assuming that
there are competing demands for those resources. In the healthcare setting,
the willingness-​to-​pay threshold is commonly used in cost-​effective analyses
to compare the resources consumed for each additional QALY gained for the
intervention, against a predetermined willingness-​to-​pay threshold. The con-
cern is that when there is spending above threshold, resources are taken away
from one group to spend on another. This becomes an equity issue as health
for one group is improved relative to another group. Equity distribution using
a threshold is often not balanced and there are arguments that more should be
spent on children, for example, or other groups who have been disadvantaged
in the past to increase their health to the level of health of others. Critics of the
specific threshold amounts may question whether thresholds represent oppor-
tunity costs. As of 2020, the UK’s NICE uses a willingness-​to-​pay threshold
of £20,000–​£30,000 (0.70 and 1.04 times GDP, respectively) per QALY gained
and it is used to make recommendations for the purchase of new technolo-
gies to the national health system based on their cost-​effectiveness. In the US,
the use of a willingness-​to-​pay threshold varies, depending on the payer. The
US-​based Institute for Clinical and Economic Review carries out CEAs and
uses a CET of $50,000–​$175,000 per QALY gained, using a higher threshold
in some cases, such as for rare disease (Institute for Clinical and Economic
Review, 2019).
Distributional CEA (Cookson et al., 2021) has been developed and ap-
plied to reflect the distribution of health outcomes across populations, and
extended CEA has sought to incorporate concerns about financial risk protec-
tion to certain groups or individuals (Verguet et al., 2016). The key objectives
of distributional CEA were described by Asaria, Griffin, and Cookson (2016)
as improving total population health and reducing unfair health inequality
by modeling and evaluating social distributions of health associated with
different interventions. There have been examples of distributional CEA in
LMICs for public health strategies such as an analysis of rotavirus vaccination
232 Reporting and interpreting results of economic evaluation

in Ethiopia by Dawkins, Mirelman, Asaria, Johansson, and Cookson (2018).

Griffin, Walker, and Sculpher (2020) analyzed the impact of air pollution re-
duction strategies from multiple perspectives using the quality-​adjusted life
expectancy framework.

3.5.7 Reporting quality of analysis with examples

Numerous tools are available to aid in the conduct or quality assessment of

CEAs. On the one hand, guidance is useful when undertaking an analysis, to
ensure that the appropriate inputs are used and that important components
for a perspective are complete. On the other hand, when a decision maker is
relying on published literature to make conclusions about value, it is useful to
have a rubric for assessing the quality of a published study to evaluate whether
the findings are reliable. When undertaking systematic reviews, it is useful
for reviewers of economic evaluations to assess quality in a consistent and
comprehensive manner (Watts & Li, 2019). For these reasons, it is useful to
become familiar with several notable tools for assessing and reporting the
quality of economic evaluation described in the following sections.

3.5.7.1 Components of a cost-​effectiveness

analysis report

The quality and reporting of CEA results is critical for decision makers to
be able to compare results across studies and generalize results from one ju-
risdiction to another. Reporting guidelines are useful because they provide
a standard set of results that CEA end-​users should expect to see in a pub-
lished CEA (Watts & Li, 2019). Additionally, the inputs and assumptions
used for the analysis should be consistently reported across studies so that
one can ascertain the methodological approach for the study. This also allows
for comparability of results stemming from studies conducted with a similar
methodological approach. Like all scientific papers, reports of CEAs should
contain introduction, methods, results, and discussion sections, with specific
components of the methods and results that are specific to CEA. For example,
methods should specifically state the study perspective, time horizon, dis-
count rate, and other key assumptions (Husereau et al., 2013). The results sec-
tion should specifically state how uncertainty was characterized alongside the
main results of incremental costs and effects.
 3.5.7 Reporting quality of analysis with examples 233

3.5.7.2 Biases, limitations, and conflicts of interest

As described in previous chapters, model inputs rely on evidence synthesis.

For each model parameter, there may be one or more published studies pro-
viding evidence for the input. These underlying studies are subject to bias
depending on their rigor and this may lead to bias in the model inputs and re-
sulting cost-​effectiveness estimates. These data source considerations should
be reported alongside the CEA methods.
Model-​based CEAs are typically subject to limitations about the availability
of reliable model inputs. For example, CEA models typically estimate cost-​
effectiveness over a longer horizon or larger population than has been studied
in clinical trials. There are potential limitations to the assumptions used to
produce the estimates, that the drug effects persist over that timeframe and
similarly in populations beyond those studied in trials.
Sources of funding for CEA studies or conflicts of interest of study authors
may exist and should be reported. Both the Consolidated Health Economic
Evaluation Reporting Standards (CHEERS) and Quality of Health Economic
Studies (QHES) checklists (described below) include items about study
funding or conflict of interest.

3.5.7.3 Reference case and impact inventory

The comparability of methods used for CEA is facilitated by use of a “refer-

ence case” to list out a consistent set of methodological choices that will guide
an analysis. One commonly used reference case example is the one proposed
by the US Public Health Service Panel on Cost-​Effectiveness in Health and
Medicine and later updated in 2016 (Gold, Siegel, Russell, & Weinstein, 1996;
Neumann, Ganiats, Russell, Sanders, & Siegel, 2016). Other reference case
sources include NICE in the UK and the International Decision Support
Initiative. A vaccine-​specific reference case is provided in Chapter 4.4.
The impact inventory is a tool for analysts to comprehensively summarize
the impacts of the intervention under consideration from both the healthcare
sector and societal perspectives. All analyses conducted from the reference
case should start with this inventory to ensure inclusion of all consequences
that could affect the ICER. The impact inventory serves to identify conse-
quences of interventions among formal and informal healthcare sectors (e.g.,
health outcomes, medical costs, and indirect medical and non-​medical costs)
as well as non-​healthcare sectors (e.g., lost productivity, future consumption
234 Reporting and interpreting results of economic evaluation

unrelated to health, and impacts on social services, criminal justice, educa-

tion, and housing systems, among others). The inventory might also highlight
impacts that are not typically considered in CEAs but important to include
elsewhere in the evaluation.

3.5.7.4 CHEERS checklist

The CHEERS statement, published in 2013, includes a 24-​item checklist and

recommendations on minimum standards for CEA reporting (Husereau
et al., 2022). At the time of publication, a panel of experts had convened to
discuss proposed updates and revisions to the original CHEERS guidelines.
The format of this checklist is particularly useful for peer reviewers, as well
as researchers aiming to identify or extract study components for systematic
reviews. The checklist recommendations are subdivided into six sections: (1)
title and abstract, (2) introduction, (3) methods, (4) results, (5) discussion,
and (6) other. There are specific recommendations for model-​based evalu-
ations versus trial-​based evaluations. These address attributes of model-​based
evaluations such as model structure, inputs, and uncertainty. The CHEERS
statement also addresses whether studies report sources of funding and po-
tential conflicts of interest.

3.5.7.5 QHES assessment instrument

The QHES instrument was developed in 2003 to support quality assessment of

cost-​effectiveness studies (Ofman et al., 2003). This 16-​question checklist as-
signs points for completion of the elements, resulting in a total score out of 100
points. The questions assign variable point values, implying increased impor-
tance and complexity of certain study attributes, such as analytic variables/​in-
puts, cost methods, model components, and conclusions/​recommendations (8
points each) versus smaller or less complex study attributes, such as subgroup
specification (1 point), funding (3 points), and perspective (4 points).

References
Asaria, M., Griffin, S., & Cookson, R. (2016). Distributional cost-​effectiveness analysis: A tuto-
rial. Medical Decision Making, 36(1), 8–​19. doi: 10.1177/​0272989X15583266
Ashcroft, R. E. (2005). Quality of life as the basis of health care resource allocation: A philosopher’s
perspective on QALYs. Virtual Mentor, 7(2).
 3.5.7 Reporting quality of analysis with examples 235

Bärnighausen, T., Bloom, D. E., Cafiero-​Fonseca, E. T., & O’Brien, J. C. (2014). Valuing vaccina-
tion. Proceedings of the National Academy of Sciences of the United States of America, 111(34),
12313–​12319. doi:10.1073/​pnas.1400475111
Berkley, S. (2014). Improving access to vaccines through tiered pricing. Lancet, 383(9936),
2265–​2267. doi:10.1016/​s0140-​6736(13)62424-​1
Bertram, M. Y., Lauer, J. A., De Joncheere, K., Edejer, T., Hutubessy, R., Kieny, M.-​P., & Hill,
S. R. (2016). Cost-​effectiveness thresholds: Pros and cons. Bulletin of the World Health
Organization, 94(12), 925–​930. doi:10.2471/​BLT.15.164418
Briggs, A., Sculpher, M., & Claxton, K. (2006). Decision modelling for health economic evalua-
tion. Oxford: Oxford University Press.
Briggs, A., Weinstein, M. C., Fenwick, E. A., Karnon, J., Sculpher, M. J., & Paltiel, A. D. (2012).
Model parameter estimation and uncertainty analysis: A report of the ISPOR-​SMDM
Modeling Good Research Practices Task Force Working Group-​6. Medical Decision Making,
32(5), 722–​732. doi:10.1177/​0272989x12458348
Chalkidou, K., Kettler, H., Ramakrishnan, G., Silverman, R., & Towse, A. (2020). Leave no one
behind: Using a benefit-​based advance market commitment to incentivise development and
global supply of COVID-​19 vaccines. Center for Global Development. https://​www.cgdev.
org/​publ​icat​ion/​leave-​no-​one-​beh​ind-​using-​bene​fit-​based-​adva​nce-​mar​ket-​com​mitm​ent-​
covid-​vacc​ine
Chi, Y. L., Blecher, M., Chalkidou, K., Culyer, A., Claxton, K., Edoka, I., . . . Winch, A. (2020).
What next after GDP-​based cost-​effectiveness thresholds? Gates Open Research, 4, 176.
doi:10.12688/​gatesopenres.13201.1
Claxton, K., Martin, S., Soares, M., Rice, N., Spackman, E., Hinde, S., . . . Sculpher, M. (2015).
Methods for the estimation of the National Institute for Health and Care Excellence cost-​
effectiveness threshold. Health Technology Assessment, 19(14), 1–​503, v–​vi. doi:10.3310/​
hta19140
Clements, K. M., Chancellor, J., Nichol, K., DeLong, K., & Thompson, D. (2011). Cost-​
effectiveness of a recommendation of universal mass vaccination for seasonal influenza in
the United States. Value in Health, 14(6), 800–​811. doi:10.1016/​j.jval.2011.03.005
Cookson, R., Griffin, S., Norheim, O. F., & Culyer, A. J. (2021). Distributional cost-​
effectiveness analysis: Quantifying health equity impacts and trade-​offs. Oxford, UK: Oxford
University Press.
Dawkins, B. R., Mirelman, A. J., Asaria, M., Johansson, K. A., & Cookson, R. A. (2018).
Distributional cost-​effectiveness analysis in low-​and middle-​income countries: illustrative
example of rotavirus vaccination in Ethiopia. Health Policy and Planning, 33(3), 456–​463.
doi:10.1093/​heapol/​czx175
Edoka, I., Kohli-​Lynch, C. N., Fraser, H., Hofman, K., Tempia, S., McMorrow, M., . . . Cohen,
C. (2021). A cost-​effectiveness analysis of South Africa’s seasonal influenza vaccination pro-
gramme. Vaccine, 39(2), 412–​422. doi:10.1016/​j.vaccine.2020.11.028
Edoka, I., & Stacey, N. K. (2020). Estimating a cost-​effectiveness threshold for health care
decision-​making in South Africa. Health Policy and Planning, 35(5), 546–​555. doi:10.1093/​
heapol/​czz152
Farmer, P. (2015, February 3). Who lives and who dies: Paul Farmer on the iniquities of health-
care funding. London Review of Books, 37(3). https://​www.lrb.co.uk/​the-​paper/​v37/​n03/​
paul-​far​mer/​who-​lives-​and-​who-​dies
Fesenfeld, M., Hutubessy, R., & Jit, M. (2013). Cost-​effectiveness of human papillomavirus vac-
cination in low and middle income countries: A systematic review. Vaccine, 31(37), 3786–​
3804. doi:10.1016/​j.vaccine.2013.06.060
Glassman, A., Cañón, O., & Silverman, R. (2016). How to get cost-​effectiveness analysis right?
The case of vaccine economics in Latin America. Value Health, 19(8), 913–​920. doi:10.1016/​
j.jval.2016.04.014
236 Reporting and interpreting results of economic evaluation

Gold, M. R., Siegel, J., Russell, L., & Weinstein, M. (1996). Cost-​effectiveness in health and medi-
cine. New York, NY: Oxford University Press.
Gray, A. M., Clarke, P. M., Wolstenholme, J., & Wordsworth, S. (2011). Applied methods of cost-​
effectiveness analysis in healthcare (Vol. 3). Oxford: Oxford University Press.
Griffin, S., Walker, S., & Sculpher, M. (2020). Distributional cost effectiveness analysis of
West Yorkshire low emission zone policies. Health Economics, 29(5), 567–​579. doi:doi.org/​
10.1002/​hec.4003
Hendriks, J., Hutubessy, R. C. W., Grohmann, G., Torelli, G., Friede, M., & Kieny, M. P. (2018).
Quadrivalent influenza vaccines in low and middle income countries: Cost-​effectiveness,
affordability and availability. Vaccine, 36(28), 3993–​3997. doi:10.1016/​j.vaccine.2018.05.099
Husereau, D., Drummond, M., Augustovski, F., de Bekker-​Grob, E., Briggs, A. H., Carswell,
C., . . . Staniszewska, S. (2022). Consolidated Health Economic Evaluation Reporting
Standards (CHEERS) 2022 Explanation and Elaboration: A Report of the ISPOR CHEERS
II Good Practices Task Force. Value Health, 25(1), 10–​31. doi:10.1016/​j.jval.2021.10.008.
Erratum in: Value Health. 2022 Jun;25(6):1060. PMID: 35031088.
Husereau, D., Drummond, M., Petrou, S., Carswell, C., Moher, D., Greenberg, D., . . . Loder, E.
(2013). Consolidated Health Economic Evaluation Reporting Standards (CHEERS) state-
ment. Value in Health, 16(2), e1–​e5. doi:10.1016/​j.jval.2013.02.010
Hutubessy, R., Chisholm, D., & Edejer, T. T. (2003). Generalized cost-​effectiveness analysis for
national-​level priority-​setting in the health sector. Cost Effectiveness and Resource Allocation,
1(1), 8. doi:10.1186/​1478-​7547-​1-​8
Institute for Clinical and Economic Review. (2019). Value assessment framework 2020. https://​
icer.org/​our-​appro​ach/​meth​ods-​proc​ess/​value-​ass​essm​ent-​framew​ork/​
Jauregui, B., Janusz, C. B., Clark, A. D., Sinha, A., Garcia, A. G., Resch, S., . . . Andrus, J. K.
(2015). ProVac Global Initiative: A vision shaped by ten years of supporting evidence-​based
policy decisions. Vaccine, 33(1), A21–​A27. doi:10.1016/​j.vaccine.2014.12.080
Jit, M., Newall, A. T., & Beutels, P. (2013). Key issues for estimating the impact and
cost-​ effectiveness of seasonal influenza vaccination strategies. Human Vaccines &
Immunotherapeutics, 9(4), 834–​840. doi:10.4161/​hv.23637
Krubiner, C., & Faden, R. (2017). A matter of morality: Embedding ethics and equity in the
health benefits policy. In A. Glassman, U. Giedon, & P. C. Smith (Eds.), What’s in, what’s out?
Designing benefits for universal health coverage (pp. 290–​326). Washington, DC: Center for
Global Development.
Leech, A. A., Kim, D. D., Cohen, J. T., & Neumann, P. J. (2018). Use and misuse of cost-​
effectiveness analysis thresholds in low-​and middle-​income countries: Trends in cost-​per-​
DALY studies. Value in Health, 21(7), 759–​761. doi:10.1016/​j.jval.2017.12.016
Loganathan, T., Ng, C. W., Lee, W. S., Hutubessy, R., Verguet, S., & Jit, M. (2018). Thresholds for
decision-​making: Informing the cost-​effectiveness and affordability of rotavirus vaccines in
Malaysia. Health Policy and Planning, 33(2), 204–​214. doi:10.1093/​heapol/​czx166
Lomas, J., Martin, S., & Claxton, K. (2019). Estimating the marginal productivity of the English
National Health Service from 2003 to 2012. Value in Health, 22(9), 995–​1002. doi:10.1016/​
j.jval.2019.04.1926
Meyer, R. (2013, November 6). “The Biggest Fraud Ever Perpetrated on the World?” How
a Twitter diatribe turned into an article in one of the world’s most prestigious journals. The
Atlantic. https://​www.thea​tlan​tic.com/​tec​hnol​ogy/​arch​ive/​2013/​11/​is-​econom​ics-​the-​bigg​
est-​fraud-​ever-​perp​etra​ted-​on-​the-​world/​281​211/​
Neumann, P., Ganiats, T., Russell, L., Sanders, G., & Siegel, J. (2016). Cost-​effectiveness in Health
and Medicine. New York, NY: Oxford University Press.
Nimdet, K., & Ngorsuraches, S. (2015). Willingness to pay per quality-​adjusted life year for life-​
saving treatments in Thailand. BMJ Open, 5(10), e008123. doi:10.1136/​bmjopen-​2015-​008123
 3.5.7 Reporting quality of analysis with examples 237

Nord, E., Daniels, N., & Kamlet, M. (2009). QALYs: some challenges. Value in Health, 12(1),
S10–​S15. doi:10.1111/​j.1524-​4733.2009.00516.x
Norheim, O. F., Baltussen, R., Johri, M., Chisholm, D., Nord, E., Brock, D., . . . Wikler, D. (2014).
Guidance on priority setting in health care (GPS-​Health): The inclusion of equity criteria
not captured by cost-​effectiveness analysis. Cost Effectiveness and Resource Allocation, 12, 18.
doi:10.1186/​1478-​7547-​12-​18
O’Mahony, J. F., & Paulden, M. (2019). The Joint Committee on Vaccination and Immunisation’s
advice on extending human papillomavirus vaccination to boys: Were cost-​effectiveness
analysis guidelines bent to achieve a politically acceptable decision? Value in Health, 22(11),
1227–​1230. doi:10.1016/​j.jval.2019.07.010
Ochalek, J., Lomas, J., & Claxton, K. (2018). Estimating health opportunity costs in low-​income
and middle-​income countries: A novel approach and evidence from cross-​country data.
BMJ Global Health, 3(6), e000964. doi:10.1136/​bmjgh-​2018-​000964
Ofman, J. J., Sullivan, S. D., Neumann, P. J., Chiou, C.-​F., Henning, J. M., Wade, S. W., & Hay,
J. W. (2003). Examining the value and quality of health economic analyses: Implications
of Utilizing the QHES. Journal of Managed Care Pharmacy, 9(1), 53–​61. doi:10.18553/​
jmcp.2003.9.1.53
Padula, W. V., Malaviya, S., Reid, N., Tierce, J., & Alexander, C. (2020). Economic value of treat-
ment and vaccine to address the COVID-​19 pandemic: A U.S. cost-​effectiveness and budget
impact analysis. Journal of Medical Economics, 24(1), 1060–​1069. https://​doi.org/​10.1080/​
13696​998.2021.1965​732
Padula, W. V., & McQueen, R. B. (2019). Expanding the role of the Patient-​Centered Outcomes
Research Institute: Reauthorization and facilitating value assessments. Applied Health
Economics and Health Policy, 17(6), 757–​759. doi:10.1007/​s40258-​019-​00525-​z
Paulden, M. (2020). Calculating and interpreting ICERs and net benefit. PharmacoEconomics,
38(8), 785–​807. doi:10.1007/​s40273-​020-​00914-​6
Principi, N., Camilloni, B., & Esposito, S. (2018). Influenza immunization pol-
icies: Which could be the main reasons for differences among countries? Human Vaccines &
Immunotherapeutics, 14(3), 684–​692. doi:10.1080/​21645515.2017.1405188
Revill, P., Walker, S. M., Madan, J., Ciaranello, A., Mwase, T., Gibb, D. M., . . . Sculpher, M. J.
(2014). Using cost-​effectiveness thresholds to determine value for money in low-​and middle-​
income country healthcare systems: Are current international norms fit for purpose? Centre for
Health Economics, University of York. http://​epri​nts.whiter​ose.ac.uk/​136​186/​
Rumbold, B., Baker, R., Ferraz, O., Hawkes, S., Krubiner, C., Littlejohns, P., . . . Hunt, P.
(2017). Universal health coverage, priority setting, and the human right to health. Lancet,
390(10095), 712–​714. doi:10.1016/​s0140-​6736(17)30931-​5
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M., . . . Ganiats, T.
G. (2016). Recommendations for conduct, methodological practices, and reporting of cost-​
effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and Medicine. JAMA,
316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
Schneider, P. (2022). The QALY is ableist: On the unethical implications of health states worse
than dead. Quality of Life Research, 31(5), 1545–​1552. doi:10.1007/​s11136-​021-​03052-​4.
Epub 2021 Dec 9. PMID: 34882282; PMCID: PMC9023412.
Schwarzer, R., Rochau, U., Saverno, K., Jahn, B., Bornschein, B., Muehlberger, N., . . . Siebert,
U. (2015). Systematic overview of cost-​effectiveness thresholds in ten countries across
four continents. Journal of Comparative Effectiveness Research, 4(5), 485–​504. doi:10.2217/​
cer.15.38
Stinnett, A. A., & Mullahy, J. (1998). Net health benefits: A new framework for the analysis of
uncertainty in cost-​effectiveness analysis. Medical Decision Making, 18(2 Suppl), S68–​S80.
doi:10.1177/​0272989x98018002s09
238 Reporting and interpreting results of economic evaluation

Stinnett, A. A., & Paltiel, A. D. (1996). Mathematical programming for the efficient alloca-
tion of health care resources. Journal of Health Economics, 15(5), 641–​653. doi:10.1016/​
s0167-​6296(96)00493-​6
Teerawattananon, Y., & Tritasavit, N. (2015). A learning experience from price negotiations for
vaccines. Vaccine, 33(1), A11–​A12. doi:10.1016/​j.vaccine.2014.12.050
Teerawattananon, Y., Tritasavit, N., Suchonwanich, N., & Kingkaew, P. (2014). The use of eco-
nomic evaluation for guiding the pharmaceutical reimbursement list in Thailand. Zeitschrift
fur Evidenz, Fortbildung und Qualitat im Gesundheitswesen, 108(7), 397–​404. doi:10.1016/​
j.zefq.2014.06.017
Thavorncharoensap, M., Teerawattananon, Y., Natanant, S., Kulpeng, W., Yothasamut, J.,
& Werayingyong, P. (2013). Estimating the willingness to pay for a quality-​adjusted life
year in Thailand: Does the context of health gain matter? ClinicoEconomics and Outcomes
Research: CEOR, 5, 29–​36. doi:10.2147/​CEOR.S38062
US Department of Health and Human Services. (2010). Patient Protection and Affordable Care
Act. U.S. Government Publishing Office. https://​www.congr​ess.gov/​111/​plaws/​publ​148/​
PLAW-​111​publ​148.pdf
Vallejo-​Torres, L., García-​ Lorenzo, B., & Serrano-​ Aguilar, P. (2018). Estimating a cost-​
effectiveness threshold for the Spanish NHS. Health Economics, 27(4), 746–​761. doi:10.1002/​
hec.3633
Vanness, D. J., Lomas, J., & Ahn, H. (2021). A health opportunity cost threshold for cost-​
effectiveness analysis in the United States. Annals of Internal Medicine, 174(1), 25–​32.
doi:10.7326/​m20-​1392
Verguet, S., Kim, J. J., & Jamison, D. T. (2016). Extended cost-​effectiveness analysis for
health policy assessment: A tutorial. PharmacoEconomics, 34(9), 913–​923. doi:10.1007/​
s40273-​016-​0414-​z
Watts, R. D., & Li, I. W. (2019). Use of checklists in reviews of health economic evaluations,
2010 to 2018. Value in Health, 22(3), 377–​382. doi:10.1016/​j.jval.2018.10.006
Woods, B., Revill, P., Sculpher, M., & Claxton, K. (2016). Country-​level cost-​effectiveness
thresholds: Initial estimates and the need for further research. Value in Health, 19(8), 929–​
935. doi:10.1016/​j.jval.2016.02.017
World Health Organization. (2014a). Making fair choices on the path to universal health cov-
erage. https://​www.who.int/​publi​cati​ons/​i/​item/​978924​1507​158
World Health Organization. (2016). Guidance on the economic evaluation of influenza vaccina-
tion. https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​250​086/​WHO-​IVB-​16.05-​eng.pdf
World Health Organization. (2020c). Seasonal influenza vaccines: An overview for decision
makers. https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​336​951/​978924​0010​154-​eng.pdf
3.6
Budget impact analysis and return
on investment
Elizabeth Watts

3.6.1 Budget impact analysis

Economic evaluations for investment cases may be enhanced by per-

forming a budget impact analysis (BIA), which demonstrates affordability
independent of a vaccine’s cost-​effectiveness. While a vaccine may be cost-​
effective, the scale of the program may have an outsize impact on the country’s
(or insurer’s) budget relative to other health spending (Bilinski et al., 2017).
A well-​ conducted BIA can inform budgeting and planning for vaccine
decision-​making.
BIA defines who pays for the difference in cost for the vaccine program
compared to the status quo and frames the cost relative to how much is spent
on other health interventions. BIA is not the total cost of the program, which
is influenced by the size of the target population and vaccine uptake (or cov-
erage), but the cost per person paying for the vaccine. Thus, BIA is typically
performed from the payer’s perspective (Sullivan et al., 2014). For publicly
funded programs, the number of payers is the number of taxpayers. For pri-
vate insurance schemes, the number of payers is the number of insurance plan
members.
The budget impact is calculated from the difference in cost for the vaccine
program compared to the status quo or alternative intervention. Additionally,
future cost savings that may accrue from preventing the disease may be sub-
tracted from the cost if those savings would accrue to the payer (Sullivan
et al., 2014).
BIA strictly refers to costs of a program and can be extrapolated using
the numerator of the incremental cost-​effectiveness ratio from a cost-​
effectiveness analysis (CEA). Building off of a CEA that estimates the

Elizabeth Watts, Budget impact analysis and return on investment In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0019
240 Budget impact analysis and return on investment

difference in cost per vaccinated individual, the total difference in cost can be
estimated by multiplying the per-​person cost difference by the target popu-
lation for the vaccine. Budget impact is then estimated by dividing the total
cost by the number of individuals who pay either taxes or premiums into
the health system or insurance scheme that will cover the cost of the vac-
cine. If the government or insurance scheme will only pay for a portion of
the vaccine cost, this should be reflected in the BIA.
Since BIA is calculated from the payer’s perspective, the time horizon
should reflect the budgeting period (1–​3 years), rather than the lifetime of the
vaccine effect, as typically presented in a CEA (Sullivan et al., 2014). Analyses
covering a long time horizon have high uncertainty, which may be subject
to changing technology or changes in the target population (Sullivan et al.,
2014). Changes in technology may include the development of more effec-
tive vaccines, or changes in formulation. Likewise, the target population may
change due to demographic factors or changes in recommendations for the
vaccine delivery. Lastly, costs in BIA should not be discounted because they
will reflect the cost in the respective year of payment (Sullivan et al., 2014).
Consider a case of hepatitis B vaccine where adding the vaccine would
cost an additional US $34,685.21 per patient with an at-​risk population of
10,000 over 50 years (the estimated life expectancy of individuals without
infection) over the current standard of treatment. If the vaccine program
will be fully funded through a private insurance scheme with 100,000 mem-
bers, the budget impact would be calculated by dividing the total cost by
the number of insurance program members. The resulting budget impact
would be that members would pay an additional $3,468.52 over the impact
period of 50 years. In annual terms, the cost would be $69.37 per member
per year, or $5.78 per member per month. This case is considered further in
Chapter 3.7.
From the numerator of the incremental cost-​effectiveness ratio, the esti-
mated per-​patient program cost is $34,685.21:

$34, 685.21
10, 000 at − risk individuals

Given there are 100,000 members, divide the total program cost to compute
the per-​member cost:

$346, 852, 080

= $3, 469.52 per member
100, 000 members
 3.6.2 Return on investment analysis 241

Divide the lifetime cost per payer by the number of years:

$3, 469.52
= $69.37 per member per year
50 years

This cost can be further contextualized by comparing the budget impact of

the vaccine program to the total annual payments per member. For example,
if the plan members contribute $1,800 per year to the insurance scheme, the
additional cost for the vaccine program would represent a 3.9% increase in the
cost of the insurance program to add the hepatitis B vaccine. Alternatively,
the cost could be compared to the expenditure for existing vaccines covered
by the insurance plan.
Budget impact shows the additional cost for decision makers by presenting
the relative additional cost of the vaccine compared to what is already being
spent on health or other vaccines. It allows decision makers to assess whether
introducing a cost-​effective vaccine will be affordable.

3.6.2 Return on investment analysis

Return on investment (ROI) analysis is a useful presentation format for deci-

sion makers and donors considering investment in immunization programs.
In contrast to CEA which compares the health impact of vaccine programs to
the costs, ROI compares benefits to costs in monetary terms. ROI is the ratio
of net benefits (benefits minus costs) to costs. It is sometimes presented as the
number of years required for the health system to break even and recoup costs.
Costs of vaccine programs include costs of vaccines (vaccine doses, syr-
inges, and safety boxes), the cost of delivering vaccines (cold chain and health-
care labor costs), as well as social mobilization efforts. Vaccine costs may be
adjusted by applying a buffer stock rate and wastage rate. When conducting
ROI as an extension of CEA, the costs included for the ROI analysis should
align with the assumptions used in the CEA.
Economic benefits of immunization should be clearly defined and presented
in monetary terms, not quality-​adjusted life years or disability-​adjusted life
years. Benefits may be calculated using several different methods. A conser-
vative approach is the cost-​of-​illness approach, which examines averted costs
of vaccine-​preventable diseases, such as costs of treatment, household out-​
of-​pocket expenditure (transportation costs), and productivity loss (caregiver
missed work or productivity loss due to disability or death). If calculating ROI
242 Budget impact analysis and return on investment

from the payer perspective, economic benefits should only include cost sav-
ings for the government or healthcare sector. A broader approach may include
additional cost savings to households, such as transportation costs and lost
wages due to caregiver absenteeism.
Other approaches for estimating economic benefits of vaccines include
the value of statistical life approach or value of statistical life-​year approach
(Robinson et al., 2019), which reflect the average willingness to pay to reduce
risk of death for individuals in a population. Given the number of ways to es-
timate economic benefits of vaccines and the impact these different perspec-
tives have on results (Ozawa et al., 2016; Sim, Watts, Constenla, Brenzel, &
Patenaude, 2020), it is prudent to present results using multiple approaches
and to clearly describe the methodology (Sim, Jit, Constenla, Peters, &
Hutubessy, 2019).
An important consideration when comparing economic benefits and costs
of vaccine programs is that benefits may accrue years after the initial invest-
ment and implementation of vaccine programs. When estimating the ROI,
benefits should be discounted to the year of initial investment or the year of
vaccination. Typically, economists apply a 3% discount rate, but alternative
discount rates (between 0% and 5%) should be explored in case they have a
large impact on the result.
ROI is computed by dividing net benefits by costs.
Equation 3.6.1. ROI:

Benefits − Costs
ROI = × 100%
Costs

As shown in Equation 3.6.1, if the ROI is positive, the benefits exceed costs.
An ROI of 50% means that for every dollar invested, there is a 50% return or
$1.5 in benefits ($0.5 in net benefits). An ROI of 2000% means that for every
dollar invested, $21 in benefits ($20 in net benefits) are generated. An alter-
native calculation is the benefit–​cost ratio (BCR), which divides benefits by
costs. If the BCR is greater than 1, benefits exceed costs.
ROI may also be presented as the number of years needed to recoup costs
after the initial investment. However, it is important to remember that ROI is
the average net benefit per dollar invested and is not adjusted for changes in
benefits or costs over time (Padula, Lee, & Pronovost, 2017). To calculate the
time to recoup costs, the time horizon considered is divided by the ROI. For
example, if assessing a vaccine program with an ROI of 2.0 over 10 years, costs
would be recouped after only 5 years (Padula et al., 2017).
 3.6.2 Return on investment analysis 243

The above example of hepatitis B vaccine can also be applied to an ROI

analysis. We saw earlier that the vaccine program cost for 10,000 at-​risk in-
dividuals would be $346,852,080. Let’s assume we estimated that the vaccine
would generate a present value of $450,000,000 in economic benefits in this
population. The ROI would be calculated as follows:
First subtract the costs from the benefits to calculate net benefits:

Net benefits = $450, 000, 000 − $346, 852, 080

Then divide net benefits by costs and multiply by 100%:

$103,147, 920
ROI = × 100%
$346, 852, 080
ROI = 29.7%

Implementing the hepatitis B vaccine in this population would yield a return

of 30 cents per dollar invested, on average. Because the ROI is greater than 0,
the benefits of implementing the vaccine exceed the costs. Likewise, the BCR
(calculated by dividing benefits by costs) of 1.3 shows that benefits of this vac-
cine exceed costs because they are greater than 1.
Researchers should be mindful of which parameters are most influential
over the results and which parameters have the greatest uncertainty and per-
form sensitivity analysis to reflect the uncertainty. One-​way sensitivity anal-
ysis may also be used to demonstrate which variables have the largest impact
on the results. Multivariate sensitivity will show overall uncertainty for all
parameters included in sensitivity analysis.
When presenting ROI to decision makers, net benefits should be pre-
sented alongside the ROI, as net benefits demonstrate scale of benefits which
is not inferable from the ROI or BCR. Decision-​makers may be inclined to
compare the ROI for the vaccine to other health or non-​health investments.
In this case, it is best to clearly state the methods used to estimate the ROI.
Depending on the approach used to estimate the economic benefits, ROI may
be comparable to benefits and costs of alternative investments, not only in
the health sector, but also in education, infrastructure, and other competing
priorities. Recent studies on the ROI of vaccines showed that vaccines against
ten pathogens would return $20 in benefits for each dollar invested using a
cost-​of-​illness approach and $52 in benefits using a value of statistical life
approach in 94 low-​and middle-​income countries (Sim et al., 2020). Other
health interventions, including nutrition supplementation and tuberculosis
244 Budget impact analysis and return on investment

Table 3.6.1 Select return on investment estimates of public health interventions

Intervention Country ROI Approach for estimating

or region economic benefits

HIV testing (Hutchinson et al., 2012) US 1.95 Cost of illness averted

Supervised injection facilities Canada 3.6 Cost of illness averted
(Andresen & Boyd, 2010)
Syringe exchange (Nguyen, Weir, Des US 7.6 Cost of illness averted
Jarlais, Pinkerton, & Holtgrave, 2014)
Vaccines against ten pathogens (Sim LMICs 19.8 Cost of illness averted
et al., 2020)
Tuberculosis prevention (Stop TB LMICs 44 Value of statistical life
Partnership, 2019)
Childhood nutrition (Horton & LMICs 44 Value of statistical life
Hoddinott, 2014)
Vaccines against ten pathogens (Sim LMICs 52.2 Value of statistical life
et al., 2020)

LMICs, low-​and middle-​income countries.

prevention, have also been evaluated using a value of statistical life approach
(Table 3.6.1) and reflect high returns. ROIs that estimate economic benefits
by estimating averted costs typically have lower ROIs than if a broader ap-
proach was taken, which may be a more appropriate approach depending on
the target audience for the analysis. A systematic review of published ROI
analyses focused on public health interventions estimated a median ROI
of 14.3 across investments in public health interventions (Masters, Anwar,
Collins, Cookson, & Capewell, 2017).

References
Andresen, M., & Boyd, N. (2010). A cost-​benefit and cost-​effectiveness analysis of Vancouver’s
supervised injection facility. International Journal of Drug Policy, 21(1), 70–​76. doi:10.1016/​
j.drugpo.2009.03.004
Bilinski, A., Neumann, P., Cohen, J., Thorat, T., McDaniel, K., & Salomon, J. A. (2017). When
cost-​effective interventions are unaffordable: Integrating cost-​effectiveness and budget
impact in priority setting for global health programs. PLOS Medicine, 14(10), e1002397.
doi:10.1371/​journal.pmed.1002397
Horton, S., & Hoddinott, J. (2014). Benefits and costs of the food and nutrition targets for the post-​
2015 development agenda. Copenhagen Consensus Center. https://​www.ign.org/​docum​ent.
cfm?page​_i​ d=​142003​378
Hutchinson, A. B., Farnham, P. G., Duffy, N., Woliski, R., Sansom, S., Dooley, S., . . . Mermin, J.
(2012). Return on public health investment: CDC’s expanded HIV testing initiative. Journal of
Acquired Immunodeficiency Syndrome, 59(3), 281–​286. doi:10.1097/​QAI.0b013e31823e5bee
 3.6.2 Return on investment analysis 245

Masters, R., Anwar, E., Collins, B., Cookson, R., & Capewell, S. (2017). Return on investment
of public health interventions: A systematic review. Journal of Epidemiology and Community
Health, 71(8), 827–​834. doi:10.1136/​jech-​2016-​208141
Nguyen, T., Weir, B. W., Des Jarlais, D. C., Pinkerton, S. D., & Holtgrave, D. R. (2014). Syringe
exchange in the United States: A national level economic evaluation of hypothetical in-
creases in investment. AIDS and Behavior, 18, 2144–​2155. doi:10.1007/​s10461-​014-​0789-​9
Ozawa, S., Clark, S., Portnoy, A., Grewal, S., Brenzel, L., & Walker, D. G. (2016). Return on
investment from childhood immunization in low-​and middle-​income countries, 2011–​20.
Health Affairs, 35(2), 199–​207. doi:10.1377/​hlthaff.2015.1086
Padula, W. V., Lee, K. K. H., & Pronovost, P. J. (2017). Using economic evaluation to illustrate
value of care for improving patient safety and quality: Choosing the right method. Journal of
Patient Safety, 17(6), e568–​e574. doi:10.1097/​PTS.0000000000000410
Robinson, L. A., Hammitt, J. K., Cecchini, M., Chalkidou, K., Claxton, K., Cropper, M.,
. . . Wong, B. (2019). Reference case guidelines for benefit-​cost analysis in global health and
development. Bill & Melinda Gates Foundation. https://​cdn1.sph.harv​ard.edu/​wp-​cont​ent/​
uplo​ads/​sites/​2447/​2019/​05/​BCA-​Gui​deli​nes-​May-​2019.pdf
Sim, S. Y., Jit, M., Constenla, D., Peters, D. H., & Hutubessy, R. (2019). A scoping review of in-
vestment cases for vaccines and immunization programs. Value in Health, 22(8), 942–​952.
doi:10.1016/​j.jval.2019.04.002
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353. doi:10.1377/​hlthaff.2020.00103
Stop TB Partnership. (2019). The paradigm shift 2018–​2022. Geneva: Stop TB Partnership.
Sullivan, S. D., Mauskopf, J. A., Augustovski, F., Jaime Caro, J., Lee, K. M., Minchin, M., . . . Shau,
W.-​Y. (2014). Budget impact analysis—​principles of good practice: Report of the ISPOR
2012 Budget Impact Analysis Good Practice II Task Force. Value in Health, 17(1), 5–​14.
doi:10.1016/​j.jval.2013.08.2291
3.7
Introduction to decision tree
modeling
William V. Padula

Decision tree modeling is a difficult task that requires attention to the clin-
ical pathway in addition to the key outcomes. There is no one way to make a
decision tree. It may be best to know that all decision trees have three major
components:

• A decision node—​where comparators are pitched against one another.

• Chance nodes—​points where clinical outcomes are weighted by the prob-
abilities of their potential.
• Terminal nodes—​clinical endpoints where costs and effectiveness meas-
ures accumulate.

Costs and effectiveness measures are weighted by the probability of clinical

outcomes that occur at each chance node.
For vaccine outcomes, it may be best to start with a standardized model
structure such as the classic Susceptible–​Infected–​Recovered (SIR) model
since such a practice has clinical validity and allows you to forgo compli-
cated model development and focus on parameter estimation.
To understand decision tree modeling, a case study can help introduce the
principles. Therefore, we provide the following case study of the hepatitis B
vaccine to walk through the exercise. More examples of vaccine economic
models that produce both health and economic benefits of vaccines can be
found through the Vaccine Impact Modeling Consortium (VIMC; http://​
www.vaccin​eimp​act.org).

William V. Padula, Introduction to decision tree modeling In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0020
 3.7.1 Case study of hepatitis B vaccine 247

3.7.1 Case study of hepatitis B vaccine

Hepatitis B is a concerning and debilitating infectious disease worldwide that

can be transmitted through blood, sexual contact, or from mother to child at
birth. If infected, a patient can have short-​term (acute) or long-​term (chronic)
consequences related to the disease, including costs for continuous health-
care. Fortunately, there is a vaccine on the market for hepatitis B virus to pro-
tect people in high-​risk areas from infection.
The vaccine is considered highly valuable to people who are high risk based
on behavioral activity or local prevalence of disease. However, there is situa-
tionally equipoise about the value of hepatitis B vaccination for primary pre-
vention in the general population. Questions about value have much to do
with the impact that herd immunity can have on entire populations. Value
also varies by whether people are low risk either because they are located in
areas where prevalence of hepatitis B virus is low, or because their own beha-
vior would minimize exposure to the infection.
The case of hepatitis B vaccine provides a useful exercise for conducting
cost-​effectiveness analysis and budget impact analysis from multiple perspec-
tives. As it turns out, depending on the perspective (e.g., healthcare sector,
payer, patient), the value may vary substantially. This exercise uses a straight-
forward decision tree in order to calculate the cost-​effectiveness and budget
impact from each of these perspectives.

3.7.1.1 Conceptual framework

To begin this exercise, open the Microsoft Excel worksheet entitled “Hep B Part
A” and start on the worksheet tab [DECISION MODEL]. Additional content
on (Hep B Part A) is available online, 10.1093/oso/ 9780192896087.012.0001.
You will note the structure of the model (Fig. 3.7.1). Some important informa-
tion about the model design should be noted here:

• Study design. The model uses the classic epidemiological framework re-
ferred to as SIR in order to perform cost-​effectiveness analysis and budget
impact analysis.
• Model structure. This model is exclusively a decision tree; Markov model
approaches are covered in Chapter 4.2.
• Perspective. The model uses information in order to make judgments about
value from three perspectives—​healthcare sector, payer, and patient.
248 Introduction to decision tree modeling

HepB_CEA-Tree_PartA

Fig. 3.7.1 Example hepatitis B vaccine decision tree for conducting economic evaluation.

• Comparators. The model compares instances for either “Vaccine

Available” or “No Vaccine.”
• Time horizon. This model provides data on outcomes for individuals
through 50 years. Patients who become infected with hepatitis B virus
may not have the same degree of survival or life expectancy on average.
• Endpoints. The SIR model provides classic endpoints including No
Infection, Recovered, and Death. These endpoints impact a number of
key measures important to economic evaluation, including costs and
health utility.
• Main outcome measures. This model is built with the goal of being able
to conduct economic evaluation in order to derive information about the
incremental cost-​effectiveness ratio (ICER) and budget impact in terms of
per member per month.

3.7.1.2 Decision model

The SIR framework assumes that not all patients are infected at baseline, and
that some individuals who become infected can recover after an acute phase,
which is consistent with the timeline of hepatitis B virus. In addition, a vac-
cine can protect many individuals from contracting hepatitis B virus, so that
they remain in a long-​term susceptible state. The following paragraphs de-
scribe the model in greater detail.
During their life course, patients start in the Susceptible (S) state. For
communities where a vaccine is not available, individuals’ experiences are
simulated in the upper arm of the model which refers to the “No Vaccine”
 3.7.1 Case study of hepatitis B vaccine 249

comparator. At any given node, individual outcomes are weighted by the

probability of these outcomes as noted by population prevalence measures.
Many individuals who do not receive a vaccine will go directly to the terminal
node entitled “No Infection.” At this endpoint, there are no costs incurred for
a vaccine or healthcare costs related to hepatitis B virus. Alternatively, a pa-
tient can advance to “HepB Infection” if they contract the virus, which is the
infected (I) state of the SIR model. This rate of infection as reflected in this
decision tree is small but non-​zero. Patients who are infected with hepatitis
B virus then advance to the two terminal nodes of this branch, “Recovered”
(R) or “Death.”
The alternative comparator arm of the model “Vaccine Available” reflects
scenarios where a vaccine exists but may not be accessible or necessary for all
individuals in the population. Whatever the reason may be, a proportion of
individuals are “Vaccinated.” Once vaccinated, we can safely assume that most
individuals in the model are protected from infection, and the same sub-​tree
presents as in the “No Vaccine” arm. These options including “No Infection”
or “HepB Infection,” which proceeds onto “Recovered” and “Death.” The other
component of the “Vaccine Available” comparator includes an option for “Not
Vaccinated” due to issues based on access or lack of need for the vaccine, per-
haps due to herd immunity or other measures. After “Vaccine Available,” the
same nodes for infection present as in the other endpoints of the decision tree.
When decision trees have similar sub-​trees past major decision nodes, this is
referred to as model symmetry.

3.7.1.3 Parameter estimation

Under the [PARAMETERS] worksheet of the exercise, you are presented

with information about critical data points needed to perform economic
evaluation in this decision model. Key parameters include information about
probabilities of outcomes, effect modifiers with the presence of a vaccine, and
cost-​sharing proportions between key stakeholder in the model. There are
also a number of measures of costs for vaccine and provider administration,
as well as hepatitis B treatment. Finally, there are data about health utility in
units of disability-​adjusted life years (DALYs) and survival. The values for
these data are currently left out of the model, so the empty cells in column B
are highlighted red. You will have to extract these data from the forthcoming
passages to put into the model to perform calculations and interpret results.
As you put these data into the cells in column B, the highlighted color of the
cells will change to yellow.
250 Introduction to decision tree modeling

3.7.2 Part A: data extraction

Open up the file “Hep B Part A” if you have not already, and navigate to the
[PARAMETERS] tab. The following information is critical to the economic
evaluation of the hepatitis B vaccine. Most of these data are translated from
the case study of hepatitis B vaccination in adults according to Hoerger et al.
(2013). As data are presented, you should input this information into the
appropriate cell.

I. Probabilities:
A. Hepatitis B occurs in the population at an incidence rate of 1%.
B. The probability of recovery from hepatitis B is 5% if infected.
C. The probability of receiving a vaccine for hepatitis B, if available,
is 85%.
D. The probability of risk reduction for hepatitis B infection with a vac-
cine is 95%, suggesting that this vaccine is highly efficacious.
E. The vaccine provides herd immunity to those in the population who
remain unvaccinated, which reduces risk of infection by a proba-
bility of 75%.
F. A payer (i.e., health insurance) will cover 90% of the cost of the
vaccine’s price.
G. A payer will cover 80% of the cost of vaccine administration, by
which a provider (e.g., physician, nurse, pharmacist, or public health
worker) injects the vaccine into an individual.
H. If a patient is infected with hepatitis B virus, they are responsible
for 25% of all diagnostic and treatment costs for the duration
of care.
II. Costs:
A. The cost of the vaccine is US $28.00 per patient.
B. The cost of vaccine administration is US $14.42 per patient.
C. The cost of a provider to diagnose hepatitis B virus in an individual is
US $250.00.
D. The annual cost to manage and treat hepatitis B virus is US $1,824.00.
III. Health utility and survival:
A. Patients diagnosed with hepatitis B who die with the virus lose 0.050
DALYs, and have an average life expectancy of 5 years.
B. Patients diagnosed with hepatitis B who recover from the virus lost
0.025 DALYs, and have an average life expectancy of 20 years.
C. Patients who are not infected have a population average disutility
0.010 DALYs, and have an average life expectancy of 50 years.
 3.7.3 Part B: exploring cost-effectiveness analysis 251

As you transcribe these data into the appropriate cell, you will notice that the
cells in column B turn from red to yellow. You’ll also notice that data are trans-
lated into probabilities for [DECISION MODEL]. It is important to note here
that the probabilities in the model are comprehensive of all population health
outcomes at each node. Given that the decision nodes are dichotomous (i.e., two
outcomes per node), you only need one probability per node to calculate the
outcomes. All nodes should add up to 100%, so knowing one of the two nodes
allows you to calculate the difference for the alternative node. For example, click
on cell F7 under No Vaccine ⇨ No Infection. This cell does not specify its own
probability, but is calculated by subtracting the probability of hepatitis B infec-
tion (i.e., hepatitis B incidence) from 100%. The probabilities of Death and Not
Vaccinated in the model are also remainders from known parameters.
In addition, not all information about probabilities in the model are trans-
lated directly translated into a probability. The probabilities of risk reduction
for hepatitis B infection due to the presence of a vaccine or herd immunity
are not actually probability of decision nodes, but effect modifiers. For ex-
ample, click on cell H12 and you will see that the rate of hepatitis B infection
is multiplied by the difference of the vaccine risk-​reduction from 100%. The
same type of calculation for herd immunity is presented in cell H18 since the
presence of herd immunity protects unvaccinated people from hepatitis B in-
fection. However, vaccination does not impact the rates of death or recovery if
a patient does become infected.

3.7.3 Part B: exploring cost-​effectiveness analysis

If you feel that you have correctly input all parameters into the model in Part
A, then you can continue with the current working model. Otherwise, we rec-
ommend opening the second iteration of the model “Hep B Part B” to con-
tinue following along this exercise. Additional content on (Hep B Part B)
is available online, 10.1093/oso/9780192896087.012.0001.
Health state probabilities are calculated in real-​ time in [DECISION
MODEL] as data are populated under [PARAMETERS]. These probabilities
provide critical data that act as weights on costs, health utilities and survival
for economic evaluation. At this point, we recommend that you click through
some of the probabilities below health states on [DECISION MODEL] in
order to understand where the probabilities come from in [PARAMETERS]
and how they are calculated.
The probability weights from the model endpoints are presented in the
[C-​E ANALYSIS] tab under column B. These probabilities are calculated as
252 Introduction to decision tree modeling

the product of node probabilities for a specific pathway. For instance, the
endpoint probability of “No Vaccine ⇨ HepB Infection ⇨ Recovered” is de-
termined by the probabilities of hepatitis B infection multiplied by recovery
(0.01 × 0.05 =​0.0005).
Once probabilities and outcomes are calculated, there should be a value in
column B associated with each pathway. As a rule, these probabilities should
add up to 100% for all health states in a decision tree arm. You can build a
check into the model to confirm this by summing up all probabilities in each
arm and ensure that they add up to 1.0. This is illustrated in column C for
each arm of the model.
For this part of the exercise, try building in the check yourself using the
“SUM()” function in excel. Do this in the following steps:

1. Sum probabilities for No Vaccine:

a. Place the cursor over cell C9 which should be a red, blank cell.
b. Enter the following formula to calculate the total probabilities of the
No Vaccine Arm: =​SUM(B4:B8)
c. Then click; the color of the cell should change to yellow.
2. Sum probabilities for Vaccine Available:
a. Place the cursor over cell C21 which should be a red, blank cell.
b. Enter the following formula to calculate the total probabilities of the
No Vaccine Arm: =​SUM(B10:B20)
c. Then click; the color of the cell should change to yellow.

If the probabilities do not total 1.0 in cells C9 and C21, then you have may an
error in the way that you put probabilities into the model.
Columns D through K have to do with cost outcomes. These are broken
down into several categories of fixed and variable costs that offer some time-​
dependency to the model’s function. These costs are also broken down by
perspective, since some costs are pertinent only to the payer or patient per-
spective. We can safely assume that the summation of payer and patient costs
represent healthcare sector costs as well.
In columns D and E are fixed costs to payers and patients. These fixed costs
are the costs that only would come up once related to vaccination or hepatitis
B management during the course of a patient’s lifetime. For patients with hep-
atitis B infection, we assume that the cost of diagnosis is a one-​time fixed cost
upfront to establish that the patient is infected. Thus, all individuals who are
infected have this cost applied to them, regardless of whether they are in the
Vaccine Available or No Vaccine arms of the model. The other fixed costs are
 3.7.3 Part B: exploring cost-effectiveness analysis 253

that of the Vaccine and Vaccine Administration. These costs are only applied
to patients in the Vaccine Available arm of the model who receive the vaccine.
The costs are broken down into payer and patient proportions based on the in-
formation you would have put into the [PARAMETERS] to ensure that costs
are distributed between both perspectives. While payers pay for the majority of
these costs, patients (or their guardians) are responsible for an “out-​of-​pocket”
portion of costs to ensure responsible utilization of these services by individ-
uals in the community, and could be interpreted as a typical co-​pay.
In columns F and G are annual, or variable, costs incurred by patients who
become infected with hepatitis B, either with or without a vaccination. We as-
sume that patients who are infected incur an average annual cost of hepatitis B
treatment and management no matter what, which again is broken down into
payer and patient components.
The total costs to each perspective appear in columns H and I, which are a
summation of the one-​time fixed cost, and the annual costs from each per-
spective multiplied by the assumed duration of survival for a model sub-
group. For instance, in the pathway No Vaccine ⇨ No Infection, there are
US $0.00 since there is neither a cost for vaccination nor treatment from any
perspective. In another instance, a patient in No Vaccine ⇨ HepB Infection
⇨ Recovered, the total cost is calculated by multiplying the annual cost of
hepatitis B treatment by the average life expectancy with hepatitis B in-
fection (recovered), added to the one-​time cost of hepatitis B diagnosis.
Whereas the alternate pathway, No Vaccine ⇨ HepB Infection ⇨ Death, is
calculated in the same fashion, but the cost of treatment is multiplied by an
specified life expectancy for hepatitis B infection until death. In a final in-
stance, we should point out that costs in the Vaccine Available arm include
the fixed cost of the Vaccine and its administration distributed to payers and
patients if the model specifies that the pathway included that the individual
was Vaccinated.
Columns J and K provide important information about costs in the process
of developing an economic model, which is to weight costs to each perspective
by the probability of outcome (column B). While costs for hepatitis B treat-
ment or vaccination can seem high, their weight in the model only matters to
the extent of the probability of an outcome. Thus, weighted costs in columns
J and K are calculated simply by multiplying the total costs to patients and
payers by the corresponding probability of outcomes in the same row. Note
that a single probability of outcome is applied to all perspectives. It is unusual
that you would have different probabilities of outcomes for different perspec-
tives when building a simple model such as this decision tree.
254 Introduction to decision tree modeling

Next, we provide important information about health utility and sur-

vival. This model analyzes information about patient health utility in units of
DALYs. However, DALYs, as interpreted for an economic model, do not pro-
vide direct data for results. DALYs provide information about the disability
accumulated with a disease, but we want to understand the utility, or gain of
health status and function as a result of avoiding infectious disease with a vac-
cine. Therefore, it is important here to convert DALYs to “DALYs Averted”
so that we can express utility on a scale of 0.0 (total disability) to 1.0 (no disa-
bility). Thus, column L calculated DALYs averted by subtracting the reported
DALYs for a particular endpoint health state from 1.0.
Column M provides the corresponding life expectancy with each health
state. In column N, these weighted life expectancies are calculated by multi-
plying life expectancy (column M) and probability of outcome (column B). In
column O, Total DALYs Averted can be illustrated from multiplying DALYs
Averted (column L) by Life Expectancy (column M). In the final column P,
critical information about weighted health utility in units of DALYs averted
is expressed by multiplying Total DALYs Averted (column O) by Probability
of Outcome (column B). This information is used in the forthcoming cost-​
effectiveness analysis calculation.

3.7.4 Part C: determining the incremental

cost-​effectiveness ratio

If you feel that you have correctly input all parameters into the model in
Part B, then you can continue with the current working model. Otherwise,
we recommend opening the third iteration of the model “Hep B Part C” to
continue following along this exercise. Additional content on (Hep B
Part C) is available online, 10.1093/oso/9780192896087.012.0001.
You now have a working model where all information is available in order to
make determinations about the cost-​effectiveness of hepatitis B vaccine from
different perspectives. Go to the [C-​E RESULTS] tab to perform these calcu-
lations. You will note that many of the cells on this tab are currently blank and
red. You will need to reference data from other tabs in the model to complete
the calculations.
Select cell C4, which is the payer cost of No Vaccine. This is the sum of all
weighted payer costs on the [C-​E ANALYSIS] tab. Using the “SUM()” com-
mand, enter a formula in C4 that references the three weighted payer costs
in column J of the [C-​E ANALYSIS] tab and adds them up in total. (Hint: the
formula should appear in the same syntax as Part B, Step 1b.)
 3.7.4 Part C: determining the incremental cost-effectiveness ratio 255

Continue these steps of adding up costs from the payer and patient per-
spectives on the [C-​E RESULTS] tab by filling in cells C5, D4, and D5. You will
need to use the same “SUM()” command to reference values in columns J and
K on the [C-​E ANALYSIS] tab. Note that for the vaccine comparator (values
in row 5), there are 6 cells that you must reference in columns J and K on the
[C-​E ANALYSIS] tab.
Once you have costs from each perspective, you can calculate the total costs
from the healthcare sector perspective for the two alternatives. Place formulas
in cells B4 and B5 that add up the costs from payer and patient perspectives in
each arm of the model for corresponding rows (i.e., rows 4 and 5).
Now you are ready to total up the DALYs averted for each arm of the study
in rows 8 and 9. Because DALYs averted is a universal measure of health
utility, we assume here that it does not vary by perspective (that is, the same
amount of DALYs averted is observed to the patient, payer, and the health-
care sector).
Select cell D8, which is the patient DALYs averted of No Vaccine. This is
the sum of all weighted DALYs averted on the [C-​E Analysis] tab. Using the
“SUM()” command, enter a formula in D8 that references the three weighted
DALYs averted in column P of the [C-​E Analysis] tab and adds them up in
total. (Hint: the formula should appear in the same syntax as Part B, Step 1b.)
Continue these steps of adding up DALYs averted from the payer, patient,
and healthcare sector perspectives on the [C-​E RESULTS] tab by filling in
cells D9, B8, B9, C8, and C9. You will need to use the same “SUM()” command
to reference values in column P on the [C-​E ANALYSIS] tab. Note that for the
vaccine comparator (values in row 9), there are 6 cells that you must reference
in column P on the [C-​E ANALYSIS] tab.
You’ll notice that as you input the total costs and DALYs averted from each
perspective, the model will automatically calculate differences in costs and ef-
fectiveness between perspectives. This information is what you need to calcu-
late the ICERs for each perspective in row 12.
For the ICER calculation, start with the total ICER calculation first from
the healthcare sector perspective. Create a formula in cell B12 that divides the
value in cell B6 by the value in cell B10. (Hint: make sure to use the “=​” sign at
the beginning of your formula.)
For the formulas in cells C12 and D12, you can copy/​paste the formula from
cell B12, and excel will automatically update the column reference points
for you.
You should now be able to see the ICER values and observe the cost-​
effectiveness determination (cells B15:D17) at different willingness-​to-​pay
thresholds. How do the ICERs vary by perspective? Which option is cost
256 Introduction to decision tree modeling

effective from which perspective at US $50,000/​DALY averted versus US

$150,000/​DALY averted?

3.7.5 Part D: exploring budget impact analysis

If you feel that you have correctly input all parameters into the model in Part
C, then you can continue with the current working model. Otherwise, we rec-
ommend opening the fourth iteration of the model “Hep B Part D” to con-
tinue following along this exercise. Additional content on (Hep B Part D)
is available online, 10.1093/oso/9780192896087.012.0001.
You now have a working model where all information is available in order
to make determinations about cost-​effectiveness. The next step is to calcu-
late budget impact. To start, go to the [BUDGET IMPACT] tab. There are two
blank cells, B3 and B4, which you will need the following data to populate:

• Assume that you are working with a health insurance payer that covers
health plans for 100,000 people. Use this information in B3.
• Assume that 10% of your population would benefit from immediate in-
tervention with the hepatitis B vaccine. Input the 10% value in cell B4.

The remaining values in column B for the budget impact analysis should automat-
ically populate. What you will now see is how the payer cost difference for vaccine
in [C-​E RESULTS] cell C6 plays into the budget impact analysis. The total cost
of the vaccine program does not present a burden to the 10,000 recipients of the
vaccine, but rather the 100,000 individuals who participate in a risk pool who can
help pay for it. Since the benefits of the vaccine are modeled over a total time ho-
rizon of up to 50 years for the cohort specified, this time plays out in dividing up
total budget impact by year, and then by month. The final budget impact value,
represented as per-​member per-​year or per-​member per-​month (cell B10) esti-
mates are important measures to many payers (both state and commercial payers)
since they operate on revenues from monthly premiums or annual tax collections.

3.7.6 Part E: interpreting budget impact

If you feel that you have correctly input all parameters into the model in Part D,
then you can continue with the current working model. Otherwise, we recom-
mend opening the final iteration of the model “Hep B Part E” to continue fol-
lowing along this exercise. Additional content on (Hep B Part E) is available
online, 10.1093/oso/9780192896087.012.0001.
 3.7.6 Part E: interpreting budget impact 257

Does the per-​member per-​month budget impact seem like a reasonable in-
vestment of a payer’s risk pool? Does this budget impact compare to other
classic examples of vaccine interventions in terms of affordability for a payer
to commit to? How does the budget impact compare to the cost-​effectiveness
of the hepatitis B vaccine at certain willingness-​to-​pay thresholds?

Reference
Hoerger, T. J., Schillie, S., Wittenborn, J. S., Bradley, C. L., Zhou, F., Byrd, K., & Murphy, T.
V. (2013). Cost-​effectiveness of hepatitis B vaccination in adults with diagnosed diabetes.
Diabetes Care, 36(1), 63–​69. doi:10.2337/​dc12-​0759
 4
ADVANCED METHODS IN
ECONOMIC EVALUATION

Edited by William V. Padula, Emmanuel F. Drabo, and Ijeoma Edoka

4.0
Section introduction: advanced
methods in economic evaluation
Emmanuel F. Drabo and William V. Padula

Economic evaluation has emerged in health as the science of assessing value, to

inform the efficient use of scarce healthcare resources. To meet the objective,
it is critical that the information generated by economic evaluation studies
be as accurate as possible. This implies that both the structure of the model
and data quality meet higher standards than one might find in a simple model
with limited parameter inputs. Both data quality and model structure will be
explored in greater detail in this section. The World Health Organization also
provides a valuable resource for additional tips on modeling complex systems
of vaccine delivery (World Health Organization: Immunization Vaccines and
Biologicals, 2019).

4.0.1 Rationale for advanced models

Health economic evaluation models are computer-​based simulation models

which focus on the population-​level impacts of diseases and disease mitiga-
tion strategies can help overcome these challenges to the evaluation of policies
by combining the best available data (e.g., epidemic, clinical, economic) from
multiple sources, to permit comparisons of the current and future epidemio-
logical, clinical, and economic impacts of multiple alternative disease-​related
interventions or strategies, and for different population subgroups.
Ideally, decisions would be guided by evidence from randomized con-
trolled trials. However, randomized controlled trials are not always possible
to conduct, and may be impractical for providing economic-​specific evidence
(e.g., costs and probabilities of some long-​term outcomes). For example, as
there may be multiple strategies to assess, and given that strategies may vary
in “intensity” of implementation or may combine multiple interventions, it
is impractical and costly to compare them in a single randomized controlled

Emmanuel F. Drabo and William V. Padula, Section introduction: advanced methods in economic evaluation In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0021
262 Advanced methods in economic evaluation

trial. In these cases, modeling techniques such as decision trees can be useful.
However, to account for the long-​term consequences of various decisions,
more complex models are needed. These models can be used to extrapolate
results from the observed clinical evidence over a short time frame to a longer
time frame, randomized controlled trials are often conducted over a relatively
short period of time (Sonnenberg & Beck, 1993).

4.0.2 Model structure

Health economic evaluation models can estimate the distribution of the pop-
ulation or a cohort of individuals targeted or affected by a given vaccination
strategy in various health states (e.g., healthy, sick, or dead), at any given point
in time. In addition, they quantify how health itself is affected by alternative
strategies. By moving from simple decision trees to a more complex and all-​
encompassing model approach, such as Markov models, economic analysis
can begin to better reflect the actual sequelae of infectious diseases. Markov
models can also more accurately capture the impact on health that a vaccine
could potentially have over time, given that Markov models encapsulate many
elements of time dependency.

4.0.3 Material covered

This section draws on a wide range of methods and applications developed

specifically for advanced economic modeling in vaccine economics to offer a
synopsis of current and emerging methods of modeling approaches for health
economic evaluation. The section is organized into four chapters which walk
the reader through these advanced vaccine economic modeling techniques
and their applications with concrete examples and exercises, starting with
simpler Markov models, and moving progressively towards more complex
modeling approaches such as transmission models.
Chapter 4.1 introduces the reader to Markov modeling using an applied ex-
ercise with a hepatitis B vaccination decision problem.
Chapter 4.2 introduces the reader to static and dynamic modeling such as
disease transmission models, with examples from both high-​and low-​income
settings. Readers will become familiar with when to use each type of model.
Chapter 4.3 focuses on probabilistic sensitivity analysis and value of
information.
 4.0.3 Material covered 263

Chapter 4.4 provides a reference case for vaccine economic evalua-

tion using a novel coronavirus disease 2019 (COVID-​19) vaccine program
Markov model.

References
Sonnenberg, F. A., & Beck, J. R. (1993). Markov models in medical decision making: A practical
guide. Medical Decision Making, 13(4), 322–​338. doi:10.1177/​0272989X9301300409
World Health Organization: Immunization Vaccines and Biologicals. (2019). WHO guide for
standardization of economic evaluations of immunization programmes. https://​www.who.int/​
immun​izat​ion/​docume​nts/​who​_​ivb​_​19.10/​en/​
4.1
Introduction to Markov modeling
Emmanuel F. Drabo and William V. Padula

To assess vaccination strategies more accurately in terms of their cost-​

effectiveness and to help guide public health decisions-​making, it is critical
to account for both the short-​and long-​term consequences of each available
strategy. As randomized controlled trials may be impractical to assess long-​term
outcomes, models are needed to extrapolate results from the observed clinical
evidence over both short-​to long-​term time frames (Sonnenberg & Beck, 1993).
Economic models examining time-​dependent outcomes are constructed to es-
timate the distribution of the population or a cohort of individuals targeted or af-
fected by a vaccination strategy in various health states (e.g., healthy, sick, or dead).
In addition, models quantify how health itself is affected by alternative strategies—​
both in terms of intermediate changes in health status, as well as health endpoints.
The economic and health consequences associated with a chosen strategy are
measured as costs (e.g., direct medical care or drug costs) and health outcomes
(e.g., life years, quality of life), respectively, and assigned to each health status, to
estimate the expected cost and effectiveness of that strategy.
Chapter 3.7 introduced the decision tree model for vaccine economics, a
simplified type of modeling approach for extrapolating the long-​term out-
comes of a vaccination strategy. While decision trees are useful for modeling
instantaneous or one-​time probabilistic events, they are limited in capturing
the effects of the duration of the time spent in each health state. This chapter
introduces the reader to Markov decision-​analytic models, which are a widely
used modeling approach in economic evaluation (Drummond, Sculpher,
Claxton, Stoddart, & Torrance, 2015b).
In 1906, Andrey Markov introduced the Markov assumption, that all fol-
lowing states are independent of all past states. Markov models depict transi-
tioning fluidly between health states based on prior probabilities of assumed
outcomes. However, Markov models with dynamic features still allow a mod-
eler to partition events so that patients with certain features are limited in
terms of what can happen next. For instance, a person cannot develop pneu-
monia without first contracting influenza, but a person with influenza could

Emmanuel F. Drabo and William V. Padula, Introduction to Markov modeling In: Handbook of Applied Health Economics in
Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0022
 4.1.1 Definition of a Markov model 265

have a probability of transitioning to a state of pneumonia based on our un-

derstanding of prior probabilities. More about this will be explained later
throughout this chapter and additional considerations and data are presented
in Appendix 4 and “SVIRD”. Additional content on (SVIRD) is available
online, 10.1093/oso/9780192896087.012.0001.

4.1.1 Definition of a Markov model

Markov models can represent disease processes that evolve over time. These
models can be designed to keep track of the costs and health-​related quality
of life (HRQoL) changes of spending time in a particular health status, by rep-
resenting health status as a series of finite, discrete health states. Changes in
health status over time are captured through transition probabilities between
health states. This process of transitioning between states can then be modelled
as finite-​state Markov chains with the corresponding state transition prob-
abilities. These models are particularly useful in the context of decision prob-
lems involving risk that continues and potentially changes over time, when the
timing of events is important to outcomes, and when events occur more than
once (Sonnenberg & Beck, 1993).
Markov models are well suited for the value assessment of vaccination
strategies, as they can capture disease transmission patterns and disease re-
currence, and estimate long-​term costs and health outcomes (e.g., life years
gained, quality-​adjusted life years (QALYs)) associated with alternative vac-
cination strategies. It would be straightforward to sum up the total time spent
in each health state and multiply this by the appropriate HRQoL factors and
costs per unit time for each health state to derive the total intervention health
effects and costs for transitioning between states.
Markov modeling provides a more convenient way of modeling the ec-
onomic and health outcomes associated with vaccination decisions. The
Markov modeling approach achieves this by modeling the process of transi-
tioning between states as finite-​state Markov chains and the corresponding
state transition probabilities.
The foundational elements of a Markov model are health states. A health
state, s, can correspond to an illness status (e.g., healthy versus sick), a par-
ticular disease stage, the presence of treatment, the receipt of a vaccine, some
other disease-​related status (e.g., circumcision status, screening status, etc.),
or death. These health states must be mutually exclusive and exhaustive. Two
states are said to be mutually exclusive if the same individual cannot be in both
simultaneously, at any given point in time. The state set is said to be exhaustive
266 Introduction to Markov modeling

if it includes all possible states relevant to the disease process. The other nec-
essary element of a Markov model is the transition of individuals in a cohort
between health states, according to their risk of contracting a disease or de-
veloping a condition, the natural progression of the disease or its underlying
biological or behavioral factors, or due to the presence of a health intervention
such as vaccination or treatment. For this reason, Markov models are some-
times referred to as transition models.
Each transition between any two health states (s, s′) has a transition proba-
bility, T (s, s ′). Although Markov models often have far more than two states,
the basic concepts can be presented by considering just two states. The tran-
sition probabilities between any two states can be described by a square two-​
dimensional transition probability matrix, T:

T=

S S′
S T(s,s) T(s,s′)
S′ T(s′,s) T(s′,s′)

The rows of the transition matrix represent the current state (s) of a given in-
dividual in the cohort or population, and the columns represent the transi-
tion state (s). Each cell (s, s′) of the transition matrix represents the transition
probability T (s, s ′) between a current state s and a transition state s′. The value
of each transition probability ranges between 0.0 and 1.0. Each row of the state
transition matrix must also sum to 1, as the states are mutually exclusive, and
one must be in a state. Transitions occur over a period, called a cycle. At any
cycle t, the probability of being in (i.e., transitioning to) a given states′ is the
product of the probability of being in states at the start of the model, and the
probability of transitioning to states ’. This can be more generally denoted by
the following formula:

S ×Tt

where S is a vector of the probabilities of being in each states at the start of

the model, and T t is the product of multiplying t state transition matrices. It
should be noted here that it is possible for the person to remain in the current
state, in which case s =​ s′. Finally, each health state can have a cost and benefit
associated to it. These costs and benefits are called rewards or payoffs, R(s).
 4.1.2 Steps in developing a Markov model 267

In Markov chains, it is possible to have a state, s′, from which individuals

cannot move out once they enter. If such a state exists, it is called an absorbing
state, and the Markov chain is referred to as an absorbing Markov chain. Any
non-​absorbing state in an absorbing Markov chain is called transient. When
S × T t = S , we say that the system is at a stationary (or steady) state. For a
more in-​depth treatment of the use of Markov models in health economic
evaluation, the reader should refer to prior work by Beck and Pauker (1983),
Sonnenberg and Beck (1993), and Briggs and Sculpher (1998).

4.1.2 Steps in developing a Markov model

The development of transparent and policy-​relevant Markov models for ec-

onomic evaluation of vaccines involves the following critical steps. First,
one must clearly define the states and allowable transitions that govern the
disease and decision process, as well as how a vaccine may or may not alter
the pathway of disease progression for some, if not all, individuals in the
population being modeled.
Second, the analyst must select an appropriate cycle length. When thinking
about the patient’s journey, we must consider the duration and frequency of
stay in each state, and the associated costs and outcomes of each state. The
cycle length, that is, the minimum time people spend in a state, is a convenient
way to measure duration of stay. The timescale of a Markov model depends
on the biology and economics of the events. If infectious diseases are pro-
gressing in important ways on a daily basis, the time should be days; if they are
progressing over months, the time cycle should be months, and so on. Cycle
lengths could be as brief as on the scale of hours or days, or as long as months
or years, if not more depending on the situation.
Related to cycle length is the time horizon, which was discussed with re-
spect to decision trees in Chapter 3.7. This is the amount of time that an in-
fectious disease or vaccine-​related immunity may be relevant. Thus, time
horizons can also be as brief as days, months, or years, but rarely exceed the
duration of immunity offered by a vaccine, or a patient’s life expectancy. For a
study of an influenza vaccine, a 12-​month time horizon is relevant since most
flu shots are only indicated for that time period. However, other vaccines such
as yellow fever and polio are considered to confer immunity ranging closer to
a patient’s full life expectancy.
Third, a set of transition probabilities between states must be specified to
construct the state transition matrix.
268 Introduction to Markov modeling

Fourth, a cost and health outcome (e.g., natural unit, utility) must be as-
signed to each health state.
Fifth, the initial set-​up of the population across the various health states
must be determined.
Sixth, methods of evaluation of the alternative strategies, such as cost-​
effectiveness analysis, must be predefined.
Finally, a perspective should be well defined at the time that the model is
being conceived so that the model accurately reflects health outcomes with
respect to costs and clinical benefits being accrued that matter to the per-
spective. As discussed in the previous chapter, perspectives can range from
patient, to provider, to health system, to payer, to government, to society, or
some combination of these examples.

4.1.3 Example of a Markov model: hepatitis B

vaccination transitions

To illustrate these concepts, consider the hepatitis B vaccination decision

problem in the example in Chapter 3.7, adapted from Hoerger et al. (2013).
The original study by Hoerger et al. (2013) found that with a 10% uptake rate,
hepatitis B vaccination among all US adults with diagnosed diabetes would
lead to 528,047 people vaccinated which could prevent 4,271 acute and 256
chronic hepatitis B infections, at an incremental healthcare cost of $91.4 mil-
lion. The intervention was estimated to generate 1,218 QALYs, thus trans-
lating into a cost-​effectiveness ratio of $75.094 per QALY gained.
Our adaption of Hoerger et al.’s Markov model of the hepatitis B vaccina-
tion decision problem includes five health states: Susceptible (S, i.e., unin-
fected, but at risk of infection), Vaccinated (V ), Infected (I ), Recovered (R
), and Dead (D). Models of this type are referred to in the epidemiological
modeling literature as SVIRD models. This model can be depicted using an
influence or state transition diagram, to show the transitions of individuals
between health states (Fig. 4.1.1).
Looking at Fig. 4.1.1, we can notice that the five health states, represented by
the ovals, are mutually exclusive, that is, individuals can only be Susceptible,
Vaccinated, Infected, Recovered, or Dead within the model. For example, if
people are in the Susceptible state, they can become infected and transition to
the Infected state, become vaccinated and transition to the Vaccinated state,
die of natural cause and transition to the Death state, or remain Susceptible.
Similarly, Vaccinated individuals can receive protection from the vaccine and
remain in the Vaccinated state, lose protection from the vaccine (if the vaccine
 4.1.4 Parameterization of Markov models 269

Susceptible Infection Recovered

(S) (I) (R)

Vaccinated Death
(V) (D)

Fig. 4.1.1 State transition diagram of the hepatitis B vaccine Markov model. The states
of the model are represented by the circles; the transitions between health states are
represented by the arrows.

is imperfect, or when the immunity from the vaccine wanes) and become in-
fected, thus transitioning into the Infected state, or die of other causes and
transition to the Dead state.
Once in the Infected state, individuals can recover from the infection and
move to the Recovered state, die of other causes or from hepatitis B complica-
tion and transition to the Dead state, or remain Infected. Those who recover
are assumed to gain lifetime immunity and remain in the Recovered state or
die of other causes and transition to the Dead state. Once people are in the
Dead state, they cannot move from that state to any other state; hence, the
Dead state is an absorbing state, and all other states are non-​absorbing tran-
sient states. The directed arrows that connect the ovals (health states) rep-
resent the directions of the transitions between states. Each of these arrows
is associated with a state transition probability, T (s, s ′). It is also worthwhile
noticing that in this model, individuals cannot return to their initial state
once they leave it. Equally noteworthy is the observation that it is possible to
remain in the same health state after a cycle.

4.1.4 Parameterization of Markov models

To best inform decision-​making, it is critical that the Markov models incorpo-

rate the best available evidence on the process that is being modeled, and the
270 Introduction to Markov modeling

interventions that are being evaluated (Briggs & Sculpher, 1998). To achieve
this, the input parameters for the model typically incorporate a wide range
of information from various sources, such as estimated vaccine efficacy from
clinical trials, disease evolution from epidemiological cohorts, quality-​of-​life
values or disability weights from population-​level studies, transition prob-
abilities from life tables, and so on. The implementation of the Markov model
should also be sufficiently flexible to accommodate these various sources
of data. As done with decision models, transition probabilities for Markov
models are derived from clinical trials, observational data, meta-​analyses,
expert panels, surveys, and other relevant sources of evidence. Chapter 3.3
provides more details on how to identify, appraise, synthesize, and transform
data from disparate sources to inform decision-​analytic models.
We draw from the reported clinical, epidemiological, and economic input
parameters in the Hoerger et al. (2013) study to parameterize the model. To
illustrate, we have summarized in Table 4.1.1 the transition probabilities (ma-
trix) associated with this Markov model for the no-​vaccination (status quo)
policy, for a cohort of diabetics aged 20 years.
These transition matrices should be read from left to right. Each cell rep-
resents the probability of transitioning from a state s to another state s′. For
example, the probability of transitioning from the Susceptible state (S) to the
Infected state (I) is 0.70. For the actual input data used to calculate the entries
of the transition matrix depicted in Table 4.1.1, the reader should refer to
Table 4.1.2 and the “Parameters” tab of the Excel file for the model.
As observed earlier, the probabilities in each row of the transition matrices
add up to 1. When developing a Markov model, one should always check that
this fundamental property holds. The constraint that the sum of the transition
probabilities from one state to all other states must sum up to 1 captures the
notion that we must always account for all transitions in each cycle, and the
idea that probabilities are exhaustive and mutually exclusive. As noted earlier,

Table 4.1.1 Illustrative state transition probabilities for the vaccination scenario

S V I R D Check

Susceptible (S) 0.198 0.100 0.700 0.000 0.002 1.000

Vaccinated (V) 0.000 0.998 0.000045 0.000 0.002 1.000
Infected (I) 0.000 0.000 0.129 0.865 0.006 1.000
Recovered (R) 0.000 0.000 0.000 0.998 0.002 1.000
Deaths (D) 0.000 0.000 0.000 0.000 1.000 1.000
 4.1.6 Simulating the transitions of a cohort 271

it is worthwhile noticing that the Dead state is a final (absorbing) state, so that
the probability of remaining in the Dead state is 1. Entries of the transition
matrix with values of 0 reflect the fact that such transitions are not permitted.
For example, given that individuals cannot transition from the Infected state
to the Susceptible state, the transition probability from I to S is 0. The transi-
tion matrix therefore must agree with the transition diagram.

4.1.5 Adding cost and benefits

We will first conduct a cost analysis using the Markov model we have intro-
duced. In their study, Hoerger et al. (2013) reported the costs associated with
healthcare utilization, as well as the health outcomes in different health states.
We use these cost estimates to conduct the cost analysis of the vaccination and
no-​vaccination interventions in our illustrative example. Our cost data in-
clude vaccination costs (e.g., cost-​share of the vaccine covered by a payer, vac-
cine price, etc.), the costs of hepatitis B diagnosis, the costs of treating acute
hepatitis B infection, as well as the costs of treating chronic hepatitis B. These
costs are summarized in Table 4.1.2. The costs summarized in the table are
costs per cycle and per person. It is worthwhile noticing that while costs in-
crease with worse states, costs in the worst state (Dead state) are 0. We also
draw from these estimates to calculate the health outcomes associated with
each health state.

4.1.6 Simulating the transitions of a cohort and

conducting a cost-​effectiveness analysis

With these data, we can calculate the expected costs and expected health out-
comes of each vaccination strategy.
In our example, we simulated the transitions of a representative cohort of
1,000 diabetic patients aged 20 years, in each type of vaccination intervention
(vaccination versus no-​vaccination policies) over 59 years, with each cycle
corresponding to 1 year. The 59-​year time horizon is used because the average
life expectancy of uninfected individuals (Susceptible and Vaccinated) in the
cohort is approximately 59 years. Hence, the analysis is effectively a lifetime
horizon.
At each cycle of the simulation, individuals age, and some die (e.g., 1,868
out of 1 million individuals in the first cycle under the first strategy). We can
272 Introduction to Markov modeling

Table 4.1.2 Model input parameter and values

Parameter Value Source

Probabilities
Acute hepatitis B transmission probabilities
Rate of asymptomatic infections (acute) 0.700 Hoerger et al. (2013)
Hospitalization rate 0.380 Hoerger et al. (2013)
Share of hospitalizations that are fulminant cases 0.040 Hoerger et al. (2013)
Probability of progression from acute to chronic 0.053 Calculation
hepatitis, weighted
Probability of death during an acute infection 0.003 Calculation
Probability of recovery from an acute infection 0.942 Calculation
Chronic hepatitis B transmission probabilities
Prevalence of chronic hepatitis B 0.082 Schillie, Xing, Murphy, &
Hu (2012)
Average annual mortality rate among persons 0.036 Calculated from the chronic
with chronic hepatitis B hepatitis B Markov model
of Hoerger et al. (2013)
Vaccination probabilities
Probability of accepting the vaccine 0.100 Assumption (Hoerger et al.,
2013)
Awareness of hepatitis B infection
Proportion of chronic hepatitis B aware of 0.339 Kim, Billah, Lieu, &
infection Weinstein (2006)
Proportion of acute hepatitis B aware of infection 0.345
Costs US $
Acute hepatitis B costs
Annual cost to treat a hepatitis B infection 1,744.00 Calculation
Test for antibody to hepatitis B-​core antigen 15.00 Kim et al. (2006)
(anti-​HBc)
Cost of chronic hepatitis B diagnosis (serologic 20.00 Assumption
testing for hepatitis B surface antigen (HBsAg))
Out-​of-​pocket cost-​share for hepatitis B 0.25 Assumption
diagnosis to patient
Out-​of-​pocket cost-​share for hepatitis B 0.25 Assumption
infection treatment to patient
Vaccination costs
Cost of vaccine 28.00 Centers for Disease Control
and Prevention (2011);
Hoerger et al. (2013)
Cost of vaccine administration 14.42 Hoerger et al. (2013); Miriti
et al. (2008)
Cost-​share of vaccine covered by payer 0.90 Assumption
Cost-​share of vaccine administration by payer 0.80 Assumption
 4.1.6 Simulating the transitions of a cohort 273

Table 4.1.2 Continued

Parameter Value Source

Utilities: quality-​adjusted life years QALYs

EQ-​5D health utility score
Diabetes (EQ-​5D) 0.751 Hoerger et al. (2013);
Sullivan & Ghushchyan
(2006)
HRQoL utility weights
Susceptible 1.000 Assumption
Vaccinated 1.000 Assumption
Infected 0.992 Calculated from the input
parameters and Markov
model of Hoerger et al.
(2013)
Recovered 1.000 Assumption

Discount rates
Discount rate 0.030 Assumption

therefore use this information to calculate the life years in each cycle. In the
first cycle, under the No Vaccination strategy, 1,868 people died in the cohort,
but 998,132 were alive, and accumulated 998,132 life years. Similarly, in the
second cycle, 996,263 people were alive, and accumulated 996,263 years of
life. We can repeat these calculations and derive the accumulated years of life
associated with each cycle. At the end of the 59 cycles, we sum all the years of
life over the 59 cycles to obtain the total life years. We can also use this infor-
mation to produce the survival curve. The area under the survival curve repre-
sents the total years of life accumulated under the strategy.
Now, we need to calculate the total costs associated with each strategy. To
do this, we need estimates of the annual costs of spending a year in each health
state, as well as estimates of the expected total number of individuals in each
state at each cycle. We must also calculate and add both the fixed and variable
costs associated with each strategy, such as hepatitis B diagnosis costs, and
vaccination costs.
In our simplified example, we only track the direct vaccination program
and medical costs. We assume that the there are no additional medical costs
associated with spending time in the Susceptible, Vaccinated, Recovered, and
Death states. Hence, only the Infected state produces direct medical costs.
However, at each cycle, newly vaccinated individuals will incur program costs
in the form of vaccination. Similarly, a fraction of infected individuals will
274 Introduction to Markov modeling

incur hepatitis B diagnosis costs. Of course, from a societal perspective, more

cost components should be included; for example, we should be accounting
for productivity loss from mortality, or from morbidity. Nevertheless, these
complexities are omitted here for ease of exposition of the model.
In our model, the Infected state consists of both individuals with an acute
infection, and those with a chronic infection. There are even more nuances,
as shown in the Hoerger et al. (2013) study. Again, for simplicity, we aggre-
gated these substates into acute and chronic infections. The annual expected
cost of spending a year in the Infected state was estimated to be $1,744 (see
Appendix 5 for derivations). The calculation of the total costs simply requires
multiplying the annual expected costs of spending time in each state with
the expected prevalence of individuals in each state during each period. We
must also add the total expected program costs during each period. For ex-
ample, for this cohort, 100,000 new individuals were vaccinated during the
first cycle, accumulating $4,118,447 in program costs.
Now that we have the total life years and total costs over the time horizon
for any given strategy (vaccination versus no vaccination), we can accumulate
these total annual life years and costs over all periods to derive the cumulative
life years and costs. For this cohort, 48,848,055 total undiscounted life years
are accumulated under a no-​vaccination strategy, compared to 48,848,808
total undiscounted life years accumulated under the vaccination strategy.
Hence, the total number of life years gained under a vaccination strategy, rel-
ative to the no-​vaccination strategy, is 753 life years. Similarly, the total costs
accumulated under the no-​vaccination strategy was $63,272,588 compared
with $66,470,281 under the vaccination strategy. Hence, the incremental costs
of the vaccination strategy, relative to no vaccination, is $3,197,693. With
the calculated incremental costs and life year gains, we can calculate the in-
cremental cost-​effectiveness ratio (ICER) associated with the vaccination
strategy, relative to no vaccination; this ICER estimate is $4,247 per life year
gained. If a decision maker was willing to pay this price per life year gained,
vaccination would be good value for money. Notice that neither the costs nor
life years are discounted yet.
Adjusting for quality (i.e., preferences) is relatively straightforward in Markov
models. Since each health state has a preference weight, that is, HRQoL weight,
associated with it, we can multiply these weights by the utility score for the di-
abetic population, and the average time spent in each state during each cycle.
The relative effectiveness of vaccination compared to no vaccination would then
be the difference in the accumulated QALYs produced by the two strategies. In
our example, the HRQoL weights associated with the Susceptible, Vaccinated,
 4.1.6 Simulating the transitions of a cohort 275

and Recovered states are 1.00 for each state. The corresponding weights for the
Infected and Death states are 0.992 (see Appendix 5 for weighting and deriv-
ations), and 0.00, respectively. Notice that the HRQoL weights for the Infected
state are high here; this is because the vast majority of infection cases are acute
infections which resolve. The EuroQol five-​dimensional (EQ-​5D) utility score
for diabetes patients is 0.751 (Hoerger et al., 2013; Sullivan & Ghushchyan,
2006). Deriving the QALYs associated with each intervention only required cal-
culating the QALYs at each cycle and accumulating them over the observation
period. The QALY associated with any given health state at any given cycle is
calculated by multiplying the utility score for the population of interest (diabetes
patients) with the total number of individuals in that health state, the duration
of time spent in the state, and the HRQoL utility weight associated with the state.
The total QALYs for the cycle is simply calculated by aggregating the state-​level
QALYs across all health states. For example, under the Vaccination strategy,
the total QALYs associated with the uninfected (Susceptible, Vaccinated, and
Recovered) states during the first cycle is 748,914 QALYs (0.751 × 1 year ×
[1.00 × 897,222 Susceptible +​1.00 × 100,000 Vaccinated +​1.00 × 0 Recovered).
For the Infected state, a total of 678 QALYs (0.751 × 1 year × [0.992 × 910
Infected]) are produced during the first cycle. The number of QALYs produced
by the Death state is 0. The total number of QALYs produced during the cycle,
across all states, is therefore 749,591 QALYs. We can then do these calculations
for each cycle of the model and trace the total QALYs associated with each.
A total of 36,685,366 QALYs are accumulated over the observation period and
across all health states, under the Vaccination strategy, compared to 36,684,672
QALYs under the No Vaccination strategy, thus resulting in 694 incremental
QALYs gained under the Vaccination strategy. This suggests an ICER of $4,605
per QALY gained.
We can also add life years and calculate the disability-​adjusted life years, as we
have done with the decision tree model presented in Chapter 3.7. We will not
expand on these calculations here, for brevity.
Discounting is also easy to do in Markov models; one simply needs to cal-
culate the present values of the cycle-​level costs and benefits associated with
each strategy, and aggregate over the total number of cycles. In our example,
we assume an annual discount rate of 3% for both costs and utilities. The dis-
counted total costs, life years, and QALYs after the first cycle are $1,544,799,
969,060 life years, and 727,759 QALYs, respectively, under the Vaccination
strategy. The cumulative discounted costs, life years, and QALYs over the en-
tire simulation period are $49,254,863, 24,674,647 life years, and 18,530,600
QALYs, respectively, for the Vaccination strategy. For the No Vaccination
276 Introduction to Markov modeling

strategy, the corresponding estimates are $38,448,013, 24,674,257 life years,

and 18,530,235 QALYs, respectively. Together, these estimates suggest that on
a discounted basis, the ICER for the Vaccination strategy, when compared to
the No Vaccination strategy is $27,733 per life year gained and $29,635 per
QALY gained.
It is worthwhile noting that the usual exercise of adjusting inputs (e.g.,
translating all costs into a single year’s value) must be conducted, although we
have not discussed it here.

4.1.7 Half-​cycle correction

It should be noted that in the approach described above to calculate the life
years accumulated in each cycle, we have implicitly assumed that those who die
died at the start of the cycle. However, they could have died at any point during
the cycle, thus we should take this into account to avoid underestimating costs
and benefits. Specifically, since people die during each cycle and have spent
some time alive during the cycle in which they died, it seems reasonable that
we should credit this time alive (as well as its associated costs) to the strategy.
This assumption is less of an issue if it is applied to all strategies being evalu-
ated, and when the cycle length is short.
A reasonable and less biased approach would be to assume that people die
in the middle of the cycle. This is a result of assuming that individuals die,
uniformly, during the cycle. In other words, we would assume that over each
cycle, death follows a uniform distribution, that is, the probability of death is
the same every time during the cycle. This assumption is known as the half-​
cycle correction. It consists of adding half the time to those who died during
the cycle. We have illustrated it in our Excel example.

4.1.8 Markov models as “trees”

In Chapter 3.7, we introduced the decision tree modeling approach for exam-
ining this problem. For recurrent events, however, the decision tree approach
is not suitable. While it is possible to construct a recursive decision tree with
one tree per cycle, it would be too complicated to do so as one would rapidly
run into the “bushy” tree problem. By this, we mean that a decision tree with
many independent nodes representing state transitions and time-​dependent
outcomes over long time horizons may be overly complex to construct.
 4.1.9 Memoryless property of Markov cohort models 277

Fortunately, the same problem can be represented as a Markovian chain pro-

cess to help address much more complex questions such as vaccination against
influenza.
Influenza provides a helpful example for this purpose because on an annual
basis, a flu shot offers the patient immunity from the infection for 12 months.
However, after 12 months, the immunity wanes and the vaccination process
should be started all over again. If looking at the value of an influenza immu-
nization program beyond 12 months, then a Markov model certainly provides
simplicity in its Susceptible, Infected, Recovered, and/​or Vaccinated structure as
compared with trees branching out for recurrent years of vaccination.

4.1.9 Memoryless property of Markov

cohort models

Markov models of the type described above assume that transitions to a

state are independent of the past or the duration an individual has been in a
state. In other words, there is no “memory” of the past. This is known as the
memoryless property or Markov assumption.
The memoryless property is a limitation of Markov models as in many
situations the duration of stay in a health state influences the chances of out-
comes. For example, most individuals with hepatitis C virus are known to
have delayed onset of symptoms. One could mitigate this limitation by adding
additional transition states (e.g., asymptomatic, acute symptoms, chronic
symptoms, liver transplantation, etc.), and making transition probabilities
conditional on past events.
As we have assumed in our hepatitis B vaccination example, we allowed for
different probabilities of death by order of liver transplantation.

References
Beck, J. R., & Pauker, S. G. (1983). The Markov process in medical prognosis. Medical Decision
Making, 3(4), 419–​458. doi:10.1177/​0272989X8300300403
Briggs, A., & Sculpher, M. (1998). An introduction to Markov modelling for economic eval-
uation. PharmacoEconomics, 13(4), 397–​409. doi:10.2165/​00019053-​199813040-​00003
Centers for Disease Control and Prevention. (2011). CDC vaccine price list: Adult vaccine price list.
http://​www.cdc.gov/​vacci​nes/​progr​ams/​vfc/​cdc-​vac-​price-​list.htm
Drummond, M., Sculpher, M., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015).
Methods for the economic evaluation of health care programmes (4th ed.). Oxford: Oxford
University Press.
278 Introduction to Markov modeling

Hoerger, T. J., Schillie, S., Wittenborn, J. S., Bradley, C. L., Zhou, F., Byrd, K., & Murphy, T.
V. (2013). Cost-​effectiveness of hepatitis B vaccination in adults with diagnosed diabetes.
Diabetes Care, 36(1), 63–​69. doi:10.2337/​dc12-​0759
Kim, S. Y., Billah, K., Lieu, T. A., & Weinstein, M. C. (2006). Cost effectiveness of hepatitis B
vaccination at HIV counseling and testing sites. American Journal of Preventive Medicine,
30(6), 498–​506. doi:10.1016/​j.amepre.2006.01.017
Miriti, M. K. K., Billah, K., Weinbaum, C., Subiadur, J., Zimmerman, R., Murray, P.,... Buffington,
J. (2008). Economic benefits of hepatitis B vaccination at sexually transmitted disease clinics
in the U.S. Public Health Reports (Washington, D.C.: 1974), 123(4), 504–​513. doi:10.1177/​
003335490812300412
Schillie, S. F., Xing, J., Murphy, T. V., & Hu, D. J. (2012). Prevalence of hepatitis B virus infection
among persons with diagnosed diabetes mellitus in the United States, 1999–​2010. Journal of
Viral Hepatitis, 19(9), 674–​676. doi:10.1111/​j.1365-​2893.2012.01616.x
Sonnenberg, F. A., & Beck, J. R. (1993). Markov models in medical decision making: A practical
guide. Medical Decision Making, 13(4), 322–​338. doi:10.1177/​0272989X9301300409
Sullivan, P. W., & Ghushchyan, V. (2006). Preference-​based EQ-​5D index scores for chronic
conditions in the United States. Medical Decision Making, 26(4), 410–​420. doi:10.1177/​
0272989x06290495
4.2
Static and dynamic modeling
Ann Levin and Colleen Burgess

In order to calculate the costs and health utilities associated with the im-
plementation of various vaccine programs, we must first have a grasp of the
epidemiological impacts of these strategies. To do this, we must be able to
predict the morbidity and mortality associated with the given infectious
disease, as well as any variations in these outcomes resulting from vacci-
nation and other interventions. This involves the utilization of epidemio-
logical modeling techniques. This chapter presents a discussion of static and
dynamic modeling of infectious diseases and the advantages and disadvan-
tages of using the two approaches. Researchers employ this type of mod-
eling for estimating the impact of vaccination on morbidity and mortality
for economic evaluations.

4.2.1 When is a static model appropriate? When is a

dynamic model needed?

There are two approaches to modeling the impact of infectious

diseases on morbidity and mortality: static and dynamic. A static model as-
sumes that the force of infection is constant or changes only as a function of
age and other individual characteristics. Static models estimate the direct ef-
fect of vaccination on vaccinated individuals—​that is, the reduction in mor-
bidity and mortality—​and are affected by the incidence and prevalence of
the disease and efficacy of the vaccine. Static models can generate outcomes
quickly and fairly simply and can often be constructed using easily accessible
software programs such as Microsoft Excel or Google Sheets. The mathe-
matics involved in static models tend to be relatively straightforward, such as
applying an infection rate to a population to calculate the expected number of
disease cases over a short period of time. These types of models can obscure
significant epidemiological interactions, however, such as the geographic

Ann Levin and Colleen Burgess, Static and dynamic modeling In: Handbook of Applied Health Economics in Vaccines.
Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0023
280 Static and dynamic modeling

heterogeneity of vaccinated populations, or the threshold for herd immunity

required to drive an outbreak to extinction.
A dynamic model, on the other hand, captures both direct and indirect
effects of vaccination on vaccinated and unvaccinated individuals, in-
cluding herd immunity.
In contrast with static models, a dynamic disease transmission model as-
sumes that the force of infection can vary throughout the course of time and
as a function of population interactions, often in nonlinear ways. To simulate
transmission, dynamic models take into account data on the contact patterns
of individuals, the distribution of the infection in the population, and the
transmissibility of the infection (Kim & Goldie, 2008). If the protective effect
of vaccination is sufficiently high and enough individuals are vaccinated, the
herd immunity threshold can be reached. This threshold is “the proportion of
a population that needs to be immune in order to halt the spread of a commu-
nicable disease” (Nymark, Sharma, Miller, Enemark, & Griffiths, 2017). That
is, if enough individuals are vaccinated, others are less likely to be infected
since fewer persons around them are susceptible to the disease. Some diseases
are highly transmissible, such as measles, and most of the population need to
be immune to the disease to achieve herd immunity. Other diseases such as
polio and Haemophilus influenza type b are not as transmissible and have a
lower threshold (Nymark et al., 2017).
Dynamic models are preferable for modeling the impact of a vaccine when
considerations of herd immunity are likely to be important. Best practice
guidelines on economic evaluation of vaccines recommend dynamic models
when “the rate at which susceptible individual acquire infection is reduced
due to vaccination or when it is not possible to obtain a conservative estimate
with a static model” (Nymark et al., 2017). Static models can be used as a
conservative estimate when indirect effects are not expected to be influential,
the disease is not infectious (e.g., cancer), and there is no human-​to-​human
transmission (e.g., rabies, tetanus, Japanese encephalitis) (World Health
Organization: Immunization Vaccines and Biologicals, 2019). See Fig. 4.2.1
for a flowchart for deciding whether to use static or dynamic models.

4.2.2 Disease transmission modeling

In this section, different types of static and dynamic models are described
(Table 4.2.1).
 4.2.2 Disease transmission modeling 281

Vaccination against human

diseases

Noninfectious
Infectious disease
disease (e.g. cancer)

Human-to-human transmission Human-to-human transmission

Static model
nonexistent (e.g. rabies, common, including via a vector
acceptable
tetanus, Japanese encephalitis) (e.g. varicella, malaria)

None of the At least one of

intervention the
options impact intervention
One of the The eligible
substantially options
eligible target target groups
on impacts
groups is or are not or do
epidemiologic substantially
includes an not include an
ally influential on
epidemiologic epidemiologic
nonhuman epidemiologic
ally influential ally influential
groups or an ally influential
subgroup (e.g. subgroup (e.g.
environmental nonhuman
children for the elderly for
reservoir (e.g. groups or an
airborne pneumococcus
animal environmental
infections) or influenza)
vaccination) reservoir that
that transmits transmits to
to humans humans

Static model Dynamic model always preferred Static model

acceptable and sometimes required acceptable

Fig. 4.2.1 Flow chart on when to use static and dynamic models.
Source: Reproduced from WHO guide for standardization of economic evaluations of immunization
programmes, WHO, Copyright (2019).

4.2.2.1 Deterministic models

The transition from one disease state to the next is governed by rates or prob-
abilities, depending on the type of model. A deterministic model assumes
that there is no randomness to the way the disease behaves in the population,
and that the system can be fully described by a set of mathematical equations
and parameters. A deterministic model for the simple Susceptible–​Infected–​
Recovered–​Vaccinated (SIRV) system can be represented by a system of ordi-
nary differential equations:
282 Static and dynamic modeling

Table 4.2.1 Classification of mathematical model types used in health economic

evaluation

Static Dynamic

Deterministic
Aggregate level 1. Deterministic aggregate-​level 2. Deterministic aggregate-​level
(compartmental/​ static model (compartmental) dynamic
cohort) 1.1 Decision trees model
1.2 State-​transition models 2.1 Discrete difference
(e.g., Markov model) equations model
1.3 Hybrid models (e.g., a (discrete time)
decision tree embedded 2.2 Ordinary differential
with Markov models) equation (ODE) model
(continuous time)
2.3 Partial differential
equation (PDE) model
(continuous time)
2.4 Other types of models that
allow for
interaction
Individual level Uncommon Uncommon
Stochastic (probabilistic)
Aggregate level 3. Stochastic aggregate-​level 4. Stochastic aggregate-​level
(compartmental/​ static model, e.g., Monte dynamic model, e.g., individual
cohort) Carlo simulation (sampling of sampling of compartmental
outcomes) of a decision tree dynamic model
or a state-​transition model
Individual level 5. Static microsimulation 6. Dynamic microsimulation
model (e.g., Monte Carlo model
microsimulation of a decision 6.1 Monte Carlo simulation
tree or a state-​transition of a Markov model with
model) interaction
6.2 Discrete-​event
simulation model
6.3 Agent-​based model

Equation 4.2.1. Deterministic model:

dS
= −βSI − ρS
dt
dI
= −βSI − γ I
dt
dR
= γI
dt
dV
= ρS
dt
 4.2.2 Disease transmission modeling 283

The transition from susceptible to infected is governed by the disease trans-

mission rate β and the number or proportion of infected people in the popu-
lation, I. Individuals can also leave the susceptible population via vaccination,
at a rate ρ which is governed by vaccination coverage and vaccine effective-
ness. Once infected, an individual recovers at the rate γ, which is generally
driven by the duration of the infectious period for the disease. In a determin-
istic model, the rates β, ρ and γ are constants or functional forms which can be
defined clearly in mathematical terms.
A deterministic model assumes the system is fully describable by a defined set
of equations and parameters, and thus every time the deterministic model is run
it will always generate the same result. In contrast, a stochastic model assumes
the presence of randomness which could affect a single parameter or multiple
components within the system. The presence of uncertainty surrounding tran-
sitions between disease states can be represented by defining rate parameters as
random values or probability distributions around a mean, as opposed to a con-
stant value or smooth mathematical function. Such stochasticity can represent
random events such as an influx of immigrants, the arrival of an infected tourist,
the unexpected loss of vaccine efficacy due to a breakdown in the cold chain, or
even a mutation in the virus. Because of this randomness, a stochastic model
will generate a different result each time the model is run. Therefore, when util-
ized, stochastic models must be run for many iterations—​in the order of hun-
dreds or thousands—​in order to explore the range of possible results. This will
result in the clustering of model outcomes, and central estimates and ranges for
epidemiological outputs such as cases and deaths can then be fed into economic
calculations to determine the resulting spread of costs and health utilities.
While some deterministic models are solvable, in many cases, the math-
ematics behind transmission models are complex enough that it is impos-
sible to solve the equations for an explicit solution.

4.2.2.2 SIRV model

From a disease perspective, a population can be divided into subgroups based

on disease states—​those who are susceptible to disease (S); those who are cur-
rently infected and able to transmit the disease to others (I); and those who
have already been infected, have since recovered, and presumably now pos-
sess some level of immunity to re-​infection (R). Taken together, this defines
what is called an “SIR” model—​Susceptible–​Infected–​Recovered (Fig. 4.2.2).
The structure of the SIR model parallels the progression of a disease outbreak
as the population flows from one group into the next.
284 Static and dynamic modeling

S I R

Fig. 4.2.2 Susceptible–​Infected–​Recovered (SIR) model.

Susceptibles become infected via disease transmission, and infected individ-

uals recover from illness. The basic SIR model can be modified in a myriad of
ways—​the inclusion of an “incubating” phase, adding both acute and chronic
stages of infection, or incorporating disease-​related mortality, along with the
inclusion of age structure, geographic variation, and so on.
Including vaccination in the model introduces a new subgroup: those who
have been vaccinated against the disease and possess partial or full protection
against infection, which may or may not wane over time (V). Now we have the
“SIRV” model (Fig. 4.2.3).
In the SIRV model, individuals leave the susceptible population either
by infection or by vaccination, and once in the vaccinated category, they no
longer participate in disease transmission within the population (depending
on vaccine efficacy, the potential for breakthrough infections, etc.). The in-
clusion of the vaccinated subgroup (in the simplest implementation) has the
direct effect of reducing the size of the susceptible subgroup—​which subse-
quently reduces the future size of the infected class.
The deterministic SIRV differential equation model is continuous—​that is,
transitions occur smoothly over time.

4.2.2.3 Contact matrices

For all of the transmission model structures described above, the functional
form of the disease transmission rate β is highly dependent upon the disease
being modeled. However, for all direct transmission diseases (i.e., diseases
which do not require a vector or fomite for transmission) β will depend on

S I R

Fig. 4.2.3 Susceptible–​Infected–​Recovered–​Vaccinated (SIRV) model.

 4.2.2 Disease transmission modeling 285

effective interactions between susceptible and infected individuals. The defi-

nition of an effective interaction varies by disease—​for bloodborne pathogens
this implies direct contact with the blood or blood products of an infected
person; for sexually transmitted diseases this requires sexual contact between
susceptible and infected partners; and for respiratory viruses this includes
close proximity contacts that result in the inhalation of respiratory secretions
from an infectious individual. The effectiveness of the interaction can depend
on the attack rate of the disease, the duration of exposure, the infectious dose,
and many other factors including the implementation of disease interventions.
Contacts within populations are often driven by social or demographic fac-
tors, including age structure, family structure, work status or job function,
socioeconomic status, and so on. Quantification of these interactions can
be incorporated into disease transmission models to add realism to trans-
mission functions and reflect non-​vaccination contact-​related interventions
such as personal protective equipment, school closures, social distancing, and
quarantine.
Mathematically, this quantification often takes the form of a contact ma-
trix. A contact matrix is a pairwise quantitative description of the number of
contacts per unit time between individuals in any two subpopulations. One
of the most utilized age-​structured contact matrix systems in epidemiolog-
ical modeling is the set of POLYMOD matrices which provide estimates for
the number of unique contacts per day between different age groups in eight
different European countries, based on daily contact diaries (Mossong et al.,
2008). This was later expanded to 152 countries by mathematical methods
(Prem, Cook, & Jit, 2017). Contact matrices based upon other population
subdivisions can be based on any number of other structures, such as the
number of patients a healthcare worker interacts with daily, or the average
ridership of public transportation systems.

4.2.2.4 Stochastic models

A stochastic model simulates the world by assuming that events occur ran-
domly rather than assuming that events occur in a prespecified manner as in
the case of deterministic models (Kim & Goldie, 2008). While adding com-
plexity, a stochastic model provides a more comprehensive evaluation of the
impact of variability and uncertainty. These types of models can be used when
modeling an infectious disease outbreak in a small population because
these are affected by chance and infectious agents are transmitted with dif-
ferent transmission probabilities (Kim & Goldie, 2008).
286 Static and dynamic modeling

Stochastic models can be simulated numerically on a computer using

programming languages such as R or software applications such as Matlab
or Mathematica, for example, to generate morbidity and mortality results
which can be used to calculate costs, quality-​adjusted life years, disability-​
adjusted life years, incremental cost-​effectiveness ratios, and other health
utility outcomes. When disease transmission models are simulated on a
computer, however, it is often more straightforward to work with variables
in discrete time, working with incremental time units of a single day, week,
month, and so on, based on the dynamics of the disease. This changes the
interpretation of the parameters in the SIRV model, which now takes the
following form:
Equation 4.2.2. Stochastic model:

St +1 = −βSt I t − ρSt
I t +1 = βSt I t − γ I t
Rt +1 = γ I t
Vt +1 = ρSt

4.2.2.5 Dynamic stochastic microsimulation

A microsimulation is an individual-​based model “in which individual instanti-

ations of a system—​such as a patient’s lifetime or the course of an epidemic—​are
generated using a random process to ‘draw’ from probability distributions a large
number of times, in order to examine the central tendency and possibly the dis-
tribution of outcomes” (Weinstein, 2006). Some examples of dynamic stochastic
microsimulations are:

• Monte Carlo simulations of a Markov model.

• Discrete-​event simulation models.
• Agent-​based models.

4.2.2.6 Discrete event simulation

Discrete event simulations assume that events occur at a discrete time in-
terval, known as time steps, rather than at any time on a continuum as in a
 4.2.3 Data requirements for validation 287

continuous model. While a continuous model may provide more accurate re-
sults, the computational burden is large and such models are sometimes not
solvable. Thus, modelers often prefer to approximate continuous solutions
with discrete models. The time steps used often affect the model’s results and
thus it is important to choose the interval carefully.

4.2.2.7 Agent-​based models

The agent-​based model is a type of dynamic microsimulation model and is

a class of model for simulating the actions and interactions of autonomous
agents with their own behavioral rules.

4.2.3 Data requirements for validation

It has been said that “all models are wrong, but some are useful” (Box, 1976).
However, if we are trying to provide scientific policy guidance such as an in-
vestment case for the introduction of a new vaccine, or a detailed compar-
ison between vaccination strategies, it is critical to minimize the degree to
which our model is “wrong,” and simultaneously maximize its “usefulness.”
We want our infectious disease transmission model to be as close to reality as
possible—​that is, the model must be validated against real-​world data. Model
validation can employ a variety of different statistical techniques to compare
model-​generated outputs to real data sets. The output generated by the models
described above includes cases (and possibly deaths) associated with the trans-
mission of a given infectious disease within a specific setting. Thus, we need to
acquire location-​specific case-​count, disease incidence, or seroprevalence data
sets to use as our comparator in the model validation process. This can be his-
torical reported disease cases, if the disease is reportable, and accounting for
potential underreporting, specific to that population. This level of data spec-
ificity can be difficult to find though. In the case in which population-​specific
real-​world data cannot be found, it may be appropriate to use another popula-
tion or location, for which quality data is available, as a proxy—​based on simi-
larity between the two populations in terms of demographic, social, economic,
and political characteristics, along with comparable structure of the respective
health systems.
The quality of data sets used for model validation is very important, as is
understanding the original source of the data and any manipulations that
288 Static and dynamic modeling

have been made to it. If, for example, mortality was counted differently for
populations under 80 years old than for those above 80 in the data set, this
could prohibit any one-​to-​one comparison with model outputs unless similar
age-​specific transformation is applied to model-​generated outcomes as well.

4.2.4 Examples of static and dynamic modeling

for vaccination models

An example of static modeling is Atherly, Lewis, Tate, Parashar, and Rheingans’

(2012) study that used a decision-​analytic model to project the health outcomes
and direct costs of introducing the rotavirus vaccination in GAVI-​eligible
countries during 2011–​2030 (Atherly, Lewis, Tate, Parashar, & Rheingans,
2012). To estimate health outcomes, they used national estimates of rotavirus
medical visits and deaths. They applied regional-​specific vaccine efficacy rates
to low-​and lower-​middle income countries as well as pooled random effects
for countries with high child mortality rates (under-​five mortality rates
>30/​1,000) to project cases averted. Scenario analyses were also conducted
for variables with uncertainty and indirect effects (assuming herd immunity).
In another static modeling study, Sander et al. (2010) conducted a cost–​
utility analysis of the universal influenza immunization program in Canada.
They used multivariate regression models to project influenza infections,
controlling for age, sex, province, influenza surveillance, and temporal
trends. They then estimated the reduction in infection rates with vaccination.
They used this information to calculate quality-​adjusted life years lost from
influenza.
Two studies that use dynamic modeling are Khan et al.’s (2018) study on the
introduction of an oral cholera vaccine in Dhaka, Bangladesh, and Burger, Sy,
Nygård, Kristiansen, and Kim’s (2015) study on the cost-​effectiveness of a de-
layed catch-​up program using the 4-​valent human papillomavirus vaccine
in Norway. Khan et al. estimated the impact of different vaccination targeting
strategies in Dhaka, Bangladesh, with an SIR model of cholera transmission that
simulates how a person can be infected by another individual or from the en-
vironment (Khan et al., 2018). Burger et al. used a model that simulates sexual
mixing and HPV transmission among females and males to project which fe-
males will acquire an HPV-​16 or -​18 infection, the two types that cause 70% of
cervical cancers (Burger et al., 2015). They also used a microsimulation model of
cervical carcinogenesis to capture cervical cancer outcomes associated with all
HPV types and screening strategies.
 4.2.4 Examples of static and dynamic modeling 289

References
Atherly, D. E., Lewis, K. D., Tate, J., Parashar, U. D., & Rheingans, R. D. (2012). Projected
health and economic impact of rotavirus vaccination in GAVI-​ eligible countries:
2011–​2030. Vaccine, 30(Suppl 1), A7–​A14. doi:10.1016/​j.vaccine.2011.12.096
Box, G. E. P. (1976). Science and statistics. Journal of the American Statistical Association,
71(356), 791–​799. doi:10.1080/​01621459.1976.10480949
Burger, E. A., Sy, S., Nygård, M., Kristiansen, I. S., & Kim, J. J. (2015). Too late to vaccinate? The
incremental benefits and cost-​effectiveness of a delayed catch-​up program using the 4-​valent
human papillomavirus vaccine in Norway. Journal of Infectious Diseases, 211(2), 206–​215.
doi:10.1093/​infdis/​jiu413
Khan, A. I., Levin, A., Chao, D. L., DeRoeck, D., Dimitrov, D. T., Khan, J. A. M.,... Qadri, F.
(2018). The impact and cost-​effectiveness of controlling cholera through the use of oral
cholera vaccines in urban Bangladesh: A disease modeling and economic analysis. PLOS
Neglected Tropical Diseases, 12(10), e0006652. doi:10.1371/​journal.pntd.0006652
Kim, S. Y., & Goldie, S. J. (2008). Cost-​effectiveness analyses of vaccination programmes: A fo-
cused review of modelling approaches. PharmacoEconomics, 26(3), 191–​215. doi:10.2165/​
00019053-​200826030-​00004
Mossong, J., Hens, N., Jit, M., Beutels, P., Auranen, K., Mikolajczyk, R.,... Edmunds, W. J. (2008).
Social contacts and mixing patterns relevant to the spread of infectious diseases. PLOS
Medicine, 5(3), e74. doi:10.1371/​journal.pmed.0050074
Nymark, L. S., Sharma, T., Miller, A., Enemark, U., & Griffiths, U. K. (2017). Inclusion of the
value of herd immunity in economic evaluations of vaccines. A systematic review of methods
used. Vaccine, 35(49 Pt B), 6828–​6841. doi:10.1016/​j.vaccine.2017.10.024
Prem, K., Cook, A. R., & Jit, M. (2017). Projecting social contact matrices in 152 countries
using contact surveys and demographic data. PLOS Computational Biology, 13(9), e1005697.
doi:10.1371/​journal.pcbi.1005697
Sander, B., Kwong, J. C., Bauch, C. T., Maetzel, A., McGeer, A., Raboud, J. M., & Krahn, M.
(2010). Economic appraisal of Ontario’s Universal Influenza Immunization Program: A
cost-​utility analysis. PLoS Medicine, 7(4), e1000256. doi:10.1371/​journal.pmed.1000256
Weinstein, M. C. (2006). Recent developments in decision-​analytic modelling for economic eval-
uation. PharmacoEconomics, 24(11), 1043–​1053. doi:10.2165/​00019053-​200624110-​00002
World Health Organization: Immunization Vaccines and Biologicals. (2019). WHO guide for
standardization of economic evaluations of immunization programmes. https://​apps.who.int/​
iris/​bitstr​eam/​han​dle/​10665/​329​389/​WHO-​IVB-​19.10-​eng.pdf
4.3
Probabilistic sensitivity analysis and
value of information analysis
Ciaran N. Kohli-​Lynch

All research findings are subject to uncertainty. We never know the true
efficacy of a vaccine, but rather a range of likely values based on observa-
tions from clinical trial and population health data. Uncertainty arises in
clinical studies because data measurement and sampling are never perfect,
leading to potential biases in study results. A variety of statistical methods
have been developed to define reasonable “confidence intervals” for study
results—​or the potential variability around an expected value (e.g., the
mean treatment effect of a vaccine). Like other areas of research, uncertainty
exists in decision-​analytic modelling. Three types of uncertainty are impor-
tant to consider:

• Methodological uncertainty.
• Structural uncertainty.
• Parametric uncertainty.

This chapter reviews the impact of uncertainty on economic modeling. It fo-

cuses mainly on parametric uncertainty, given the impact that such uncer-
tainty can have on estimates of vaccine value.
Parametric uncertainty refers to uncertainty around the inputs for a de-
cision model. Decision-​analytic modelling often involves synthesizing data
from multiple studies. Inevitably, uncertainty in these input data propagates
into modelling results. Probabilistic sensitivity analysis (PSA) quantifies the
impact of parametric uncertainty on model outputs.
As discussed in Chapter 3.5, we can explore the impact of parametric un-
certainty through deterministic sensitivity analysis. The Second Panel on
Cost-​Effectiveness in Health and Medicine notes that PSA is a more thorough
approach for quantifying the uncertainty inherent in decision modelling

Ciaran N. Kohli-​Lynch, Probabilistic sensitivity analysis and value of information analysis In: Handbook of Applied Health
Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford
University Press 2023. DOI: 10.1093/oso/9780192896087.003.0024
 4.3.1 Uncertainty in economic evaluation 291

(Sanders et al., 2016). It involves modelling parameter values as distributions

rather than point estimates and running a model multiple times with input
values stochastically sampled from these distributions. Value of information
analysis produces actionable decisions based on the results of PSA. It quan-
tifies the costs and benefits of investing in further research to reduce uncer-
tainty in a decision.

4.3.1 Uncertainty in economic evaluation

Uncertainty in decision-​analytic modelling can broadly be separated into

three categories: methodological, structural, and parametric uncertainty
(Bojke, Claxton, Sculpher, & Palmer, 2009; Briggs, 2000). These types
of uncertainty arise through different processes and must be addressed
differentially.

4.3.1.1 Methodological uncertainty

Methodological uncertainty refers to the fact that different analytic ap-

proaches may be adopted in a modelling study. Uncertainty in some model
parameters does not arise due to imprecise data, but rather because of disa-
greement in the optimal approach needed to address a research question. This
handbook has explored key analytic considerations when conducting an eco-
nomic evaluation of a vaccination program. These include study perspective,
inclusion of unrelated disease (or “background”) costs, and the metric used
to value health benefit (e.g., quality-​adjusted life years (QALYs), disability-​
adjusted life years (DALYs) averted).
The aim of a healthcare decision maker is to choose which treatments to
implement, given multiple options and a limited budget. Hence, decision
modelling studies are most informative when they adhere to a common
methodological approach that enables comparisons between multiple dif-
ferent interventions. Consistency in analytic approaches can be achieved
by adopting a “reference case” of core analytic methodology (International
Decision Support Initiative, 2017; National Institute for Health and Clinical
Excellence, 2013; Sanders et al., 2016). Optimally, each decision maker
would define their own reference case. Researchers can explore the impact
and appropriateness of reference case criteria in sensitivity analyses.
292 Probabilistic sensitivity analysis and value of information analysis

4.3.1.2 Structural uncertainty

Structural uncertainty pertains to decisions made by researchers when con-

structing a decision model. These include the type of model employed (e.g.,
decision tree, Markov model, dynamic transmission model), the way transi-
tion probabilities are defined (e.g., fixed versus time-​dependent probabilities),
and the health states included in a model (e.g., health outcomes and clinical
endpoints).
Cost-​effectiveness analysis is often a complex process laden with assump-
tions. There is a constant trade-​off between research time, computing effi-
ciency, and model accuracy. Two researchers asked to analyze the same
question with the same input data will likely produce different models.
Neither model is likely to comprehensively track all patient outcomes, but
they may capture the outcomes of greatest interest, economic impact, or epi-
demiologic burden.
The solutions for dealing with structural uncertainty are demanding. The
Second Panel on Cost-​Effectiveness in Health and Medicine proposes that
structural uncertainty can be reduced by tasking multiple research teams with
model development and averaging outcomes across different models (Sanders
et al., 2016). This is clearly a resource-​intensive process and determining
weights to apply to different models is difficult. It is sometimes possible to
“parameterize” sources of structural uncertainty. Imagine there is uncertainty
regarding whether a health state should be included in a Markov model. This
health state could be included conditional on a probability which aligns with
the researchers’ prior beliefs that it is an important part of the disease pathway.

4.3.1.3 Parametric uncertainty

Parametric uncertainty relates to uncertainty in the input parameters for a

model. Estimates of parameters will inevitably be imprecise. Costs, health
state valuations, and probabilities are all sources of parametric uncertainty.
Any value which is estimated through observational or experimental means,
such as a cohort study or clinical trial, can be described as a random variable.
Because these variables are estimated through partly random processes, like
the process of recruiting a generalizable cohort, there is inherent uncertainty
in their true value.
Model inputs are often derived from published literature (see
Chapter 3.3). Greenhalgh (1997) provides a hierarchy for the reliability, validity,
 4.3.1 Uncertainty in economic evaluation 293

and generalizability of published data. In descending order, Greenhalgh’s hi-

erarchy includes:

• Systematic reviews and meta-​analyses of randomized controlled trials.

• Randomized controlled trials with definitive results.
• Randomized controlled trials with non-​definitive results.
• Cohort studies.
• Case–​control studies.
• Cross-​sectional surveys.
• Case reports.

In research areas with little quantitative data, expert opinion may be included
at the end of this list (Leal, Wordsworth, Legood, & Blair, 2007). Deriving
model inputs from any of these methods will introduce parametric uncer-
tainty into the decision modelling process.
Published studies often present a point estimate, or expected value for the
value of a parameter, alongside a confidence interval (CI). The point esti-
mate of random variable X can be referred to as its expected value, E[ X ]. In
a sample of random variables, the arithmetic mean, often represented by x, is
an unbiased estimator of E[ X ]. The CI represents a plausible range for the true
value of the parameter. It describes uncertainty in the point estimate and is a
function of the variance of a random variable. The sample variance, s 2, is an
unbiased estimator of the population variance, σ2 .
Clinical research cannot establish the true value of a parameter due to
“sampling” and “non-​sampling” errors. Sampling errors exist because most
clinical research elicits information from a subset of the total population of
interest. A randomized controlled trial for a new vaccine will only recruit a
small proportion of the potential patient population. It is therefore impossible
to establish what the exact treatment effect would be in the overall population.
Non-​sampling errors arise due to imperfectly conducted research. Survey an-
swers may be misunderstood, data collection may be incomplete, anthropo-
morphic variables may be incorrectly measured, and data may be incorrectly
processed.
The CI indicates the likely range for the value of a parameter. For example,
a meta-​analysis by Cochrane synthesized results from randomized controlled
trials of vaccination against human papillomaviruses in young women to pre-
vent cervical precancer (Arbyn, Xu, Simoens, & Martin-​Hirsch, 2018). They
found that the relative risk of precancer for young women receiving the human
papillomavirus vaccine versus placebo was 0.05 with a 95% CI of 0.03–​0.10.
294 Probabilistic sensitivity analysis and value of information analysis

With 95% confidence, they state that the true relative risk in the studied pop-
ulation lies between 0.03 and 0.10. Using the point estimate as a parameter in
a decision modelling study would disregard this acknowledged uncertainty.
Parametric uncertainty should not be confused with subgroup-​level vari-
ability. The former refers to uncertainty in the true value of a parameter for
a population. The latter refers to real differences in parameters that exist be-
tween patient subgroups (Kohli-​Lynch & Briggs, 2019). When there is reason
to believe that parameters vary based on identifiable patient characteristics
(e.g., age, sex, presence of biomarker), decision analysis should be conducted
separately for each subgroup (Sculpher, 2008).

4.3.2 Probabilistic sensitivity analysis

It is important to reflect parametric uncertainty in decision modelling studies.

Failing to address this uncertainty will bias cost-​effectiveness estimates.
Additionally, quantifying uncertainty allows researchers to assess the benefits
of obtaining more information regarding model inputs. A standard approach
for dealing with uncertainty in decision modelling studies is to conduct PSA.

4.3.2.1 Basic concept

PSA involves running a decision model multiple times. Each run of the model
can be referred to as an iteration. Key parameters are not included in the model
as unitary values when conducting a PSA. Instead, parameters are included as
draws from probability distributions. Probability distributions are functions
that assign probabilities to the range of possible values that the variable can
assume, based on the likelihood that they are the true value. For example, the
probability distribution for a fair coin toss would be a function that assigns
a probability of 0.5 to outcome “heads” and a probability of 0.5 to outcome
“tails.” The standard normal distribution, shown in Fig. 4.3.1, is another prob-
ability distribution. It has a mean value of 0 and a standard deviation of 1, with
dispersion around the mean described by a mathematical formula. Similar
distributions, often with a different mean and standard deviation, are com-
monly used to describe biomedical phenomena.
Using a probability distribution rather than a point estimate to repre-
sent a model parameter acknowledges the fact that there is uncertainty sur-
rounding the true value of the parameter. Every time the model is run in a
 4.3.2 Probabilistic sensitivity analysis 295

40%

35%

30%
Probability density

25%

20%

15%

10%

5%

0%
−3.0 −2.0 −1.0 0.0 1.0 2.0 3.0

Fig. 4.3.1 The standard normal distribution.

PSA, a different set of parametric inputs are probabilistically sampled from

their respective distributions. The process of running the model repeatedly
with probabilistically sampled inputs is called “Monte Carlo simulation.” At
the end of each run, model outcomes (e.g., QALYs, DALYs averted, costs) are
recorded. Mean cost-​effectiveness is then calculated by averaging across all
recorded outcomes. The multiple estimates produced during a PSA enable
estimation of confidence intervals for model outputs.
Consider the model shown in Fig. 4.3.2, adapted from a cost-​effectiveness
analysis of the seasonal influenza vaccine in South Africa (Edoka et al., 2021).
This decision tree model estimates the cost-​effectiveness of vaccinating eld-
erly individuals. There is considerable uncertainty in many of the param-
eters that determine the cost-​ effectiveness of the vaccination program.
Unvaccinated elderly individuals have a 3.2% (95% CI 1.4–​5.6%) probability
of developing seasonal influenza, a 4.4% (95% CI 1.8–​7.8%) probability of
dying if they contract the disease, and incur a loss of 0.031 QALYs (95% CI
0.025–​0.037 QALYs) if they develop severe influenza. A vaccine is available
with 58% (95% CI 34–​73%) effectiveness.
Deterministic analysis refers to running a decision model with point esti-
mate inputs. When we evaluate the seasonal influenza model deterministi-
cally, we estimate that the vaccination program would produce 294 QALYs
at a cost of around $615,000 (USD in 2018). This leads to an incremental
cost-​effectiveness ratio (ICER) of around $2,090/​QALY and is displayed on
296 Probabilistic sensitivity analysis and value of information analysis

GP visit, non-fatal

Non-medically attended, non-fatal

Influenza GP visit and hospitalisation, non-fatal

TIV vaccination Hospitalisation, non-fatal

programme
Influenza-related mortality

No Influenza

No vaccination
programme

Fig. 4.3.2 Decision tree model, seasonal influenza vaccination for elderly adults in
South Africa. TIV, trivalent inactivated influenza vaccine.
Source: Reprinted from Vaccine, 39(2), Edoka, I., Kohli-​Lynch, C. N., Fraser, H., Hofman, K.,
Tempia, S., McMorrow, M.,... Cohen, C., A cost-​effectiveness analysis of South Africa’s seasonal
influenza vaccination programme, 412–​422, (2021), with permission from Elsevier.

the cost-​effectiveness plane with a black cross (Fig. 4.3). South Africa’s cost-​
effectiveness threshold should be $3,040/​QALY according to recent research
(Edoka & Stacey, 2020). This threshold is shown on the cost-​effectiveness
plane. As the deterministic estimate of cost-​effectiveness lies below the
threshold, it should be implemented based on this result.
We could also run the seasonal influenza model probabilistically. The
cloud of red spots on Fig. 4.3.3 shows the results of 1,000 PSA iterations.
In each iteration, input parameters for the model were pulled from re-
spective probability distributions and led to varying estimates for the
cost-​effectiveness of the vaccine. Using PSA, uncertainty in the model
inputs has been propagated into the model outputs and we gain a greater
understanding of the uncertainty inherent in the decision-​making pro-
cess. Additionally, the average cost-​effectiveness across the 1,000 PSA it-
erations will vary, albeit marginally, from the deterministic outcome of
the model.
While there is no objective means of determining how many iterations
are required to produce accurate estimates with a PSA, it is common to run
a model several thousand times. Ultimately, the number of iterations chosen
should reduce any uncertainty that arises simply due to the Monte Carlo sim-
ulation (Hatswell, Bullement, Briggs, Paulden, & Stevenson, 2018). There
should be “model convergence” whereby PSAs with the same number of iter-
ations produce similar mean outputs.
 4.3.2 Probabilistic sensitivity analysis 297

1,500,000 PSA iteration

Deterministic estimate
Cost-effectiveness threshold
1,250,000

1,000,000
Incremental costs ($)

750,000

500,000

250,000

0
0 100 200 300 400 500 600 700 800
Incremental QALYs

Fig. 4.3.3 Cost-​effectiveness scatter plot with deterministic and probabilistic results.

4.3.2.2 Purpose of probabilistic sensitivity analysis

Most clinical research relies on statistical tests to validate findings. While

these tests vary, researchers typically strive to prove that their findings have a
type-​1 error rate of 5% or lower (i.e., p-​value <0.05), meaning they have 95%
confidence that the null hypothesis of a study can be rejected. In most clinical
studies, the null hypothesis states that some new treatment is equally as effec-
tive as a comparator. If an effect is observed with a p-​value less than 0.05, there
is strong evidence of a difference in effect between the treatment and compar-
ator. Confidence intervals and p-​values can be derived for cost-​effectiveness
estimates in decision modelling studies through PSA.
One use of a PSA is to determine whether we can reject the null hypothesis
that two treatment options are equally cost-​effective. Regulatory bodies often
require that manufacturers of novel healthcare products prove that a treat-
ment is efficacious before approving it for widespread use. Healthcare payers
may be similarly loath to fund an intervention if they are not confident that it
is more cost-​effective than established comparators.
Claxton (1999) argues that arbitrary hurdles like the statistical tests de-
scribed above should be avoided when deciding whether to implement a
298 Probabilistic sensitivity analysis and value of information analysis

healthcare intervention. If a healthcare decision maker must choose between

a new intervention and standard care, they should implement the interven-
tion which is most cost-​effective according to a central cost-​effectiveness
estimate, regardless of uncertainty. It is pure historical happenstance that
determines which option from a set of competing interventions is currently
considered standard care. While some decisions will inevitably be incorrect,
the net effect of this approach will be that net benefit in the population is
maximized.
The Arrow–​Lind principle states that, under certain circumstances, the cost
of investing in social programs with some uncertainty in outcome tends to 0
as the population becomes very large (Arrow & Lind, 1970). This supports
the idea that healthcare decision makers should base decisions on central es-
timates when evaluating investment opportunities when the risk of an incor-
rect decision is spread over a large population.
If a central estimate of cost-​effectiveness is sufficient to inform health-
care decision-​making, regardless of its confidence interval, it is worth con-
sidering the necessity and relevance of PSA. Expected cost-​effectiveness
results derived from deterministic analysis are often biased. Conducting
a PSA helps correct this bias and produce an accurate estimate of cost-​
effectiveness. Beyond improving model accuracy, PSA also provides in-
formation regarding uncertainty in the estimates produced by a decision
model. This information can be used to elicit the probability of making an
incorrect decision based on model results and estimate the value of reducing
uncertainty in the decision.
Consider the results shown in Fig. 4.3.3. The central estimate for the ICER
is below the cost-​effectiveness threshold. Based on this result, the treatment
should be implemented. However, 21% of PSA simulations are above the
threshold. Accordingly, there is a 21% chance that a decision based on the cen-
tral estimate will be incorrect and the vaccine will not be cost-​effective for eld-
erly individuals in South Africa. Further research to reduce this uncertainty
may be desirable.

4.3.2.3 Expected cost-​effectiveness

When we evaluate a decision model, we are interested in its expected value. All
decision models are essentially functions; that is, they are mathematical op-
erators that accept a range of inputs which they combine to estimate outputs
(e.g., health and cost outcomes). To calculate the expected value of a function,
 4.3.2 Probabilistic sensitivity analysis 299

g (x ), we must evaluate the function at each possible value of input variable, x.

Next, we weight each outcome by the likelihood that it is the true parameter
value and sum across all potential inputs.
The process of calculating the expected value of a function with continuous
inputs can be presented mathematically as follows: E[ g ( X )] = ∫ g (x ) p(x )dx.
The function p(x ) is a probability distribution, outlining all the possible values
that the random variable x may assume and the likelihood that they are the
true parameter value. A random variable x is described as discrete when it has
countable number of possible values. The expected value of a function of a
discrete random variable can be estimated as E[ g (x )] = ∑ X p(x ) g (x ).
More generally, the expected value of function g ( X ) of any random variable x
can be expressed as Ex [ g ( X )].
For most decision models, evaluating Ex [ g ( X )] is very difficult. Even simple
models represent long, complicated functions with multiple interacting param-
eters. We can estimate the expected outcome of a model by running a PSA. This
imitates the process described in the calculations above. We assess the output
of the model at a range of input values which are sampled from probability dis-
tributions. The distributions are chosen to ensure that likely values are sampled
more frequently than unlikely values. As we increase the number of PSA runs
and average over observed outcomes, our cost-​effectiveness estimate tends to-
wards the expected value of the model.
Depending on the complexity of a decision model, we may be able to use
non-​probabilistic methods to obtain an unbiased estimate for the expected
value of the model. The scale and complexity of decision models vary greatly,
from simple decision trees to complex microsimulation and discrete event
simulation models and beyond. The least complicated decision models, in-
cluding some worksheet-​based models and decision trees, are “linear in
their input variables.” This means that the model can be expressed as a se-
ries of terms added together. An example of a simple linear function is
g ( X ) = mX + c , where X is a random variable with expected value E[ X ] and m
and c are constant. A common proof in mathematics shows that the expected
value of a linear function is equal to the value of the function evaluated at
the expected value of its parameters. Mathematically, Ex [ g ( X )] = g (E[x]). For
our purposes, this means that some simple decision models can be run de-
terministically without biasing results. Running a PSA adds no additional
information regarding the expected cost-​effectiveness of an intervention in
this scenario.
Most decision-​ analytic models are nonlinear in their input vari-
ables. They cannot be expressed as a simple linear function. Rather, they
300 Probabilistic sensitivity analysis and value of information analysis

represent complicated, multivariate functions where parameters com-

bine multiplicatively and in power relations. Even the simplest Markov
models quickly become nonlinear as transition probabilities must be
cross-​multiplied and combined as powers to estimate outcomes. Unlike
linear models, the expected value of a linear function is not equal to the
value of the function evaluated at the expected value of its parameters.
Mathematically, Ex [ g ( X )] ≠ g (E[x]). Hence, plugging point estimates for
parameters into a nonlinear decision model produces biased estimates. It
is important to recognize that running a nonlinear model deterministically
is unlikely to cause large biases in model results; Ex [ g ( X )] rarely differs
greatly from g (E[ x]). Nonetheless, using probabilistic methods is mathe-
matically more accurate.

4.3.2.4 Making models probabilistic

Conducting a PSA involves running a model several times with different

input parameters. Values for parameters are chosen by sampling from prob-
ability distributions. Suitable distributions for parameters should be defined
with consideration for the point estimate of the parameter, its variance, and
logical constraints on the values that it can take.
Many types of probability distributions exist. Key among these is the normal
distribution, shown in Fig. 4.3.1. Values for this distribution are distributed
symmetrically around the population mean and decrease in frequency as they
move away from the mean. While other distributions share these characteris-
tics, the normal distribution is important as it describes a wide range of nat-
ural phenomena. Indeed, according to the central limit theorem, the normal
distribution describes the sampling distribution of the mean of any set of
random variables with a sufficient sample size. In practice, few parameters are
derived from large enough samples to justify using the normal distribution to
describe their distribution in a PSA.
Often logical data constraints make it preferable to employ alternatives to
the normal distribution in PSA. Fig. 4.3.4 displays an illustrative histogram
of influenza vaccine costs observed in a fictional South African costing
study with 1,000 observations. Similar data were used to inform this param-
eter in the seasonal influenza vaccine model. Most vaccines cost less than
$5.00, but some payers spent significantly more per dose. Such price varia-
tion is common in multi-​payer markets where large insurers can negotiate
discounts from listed prices when buying products in bulk. The observed
 4.3.2 Probabilistic sensitivity analysis 301

250
Normal distribution
225 Gamma distribution
Observed data
200

175

150
Frequency

125

100

75

50

25

0
–5 –4 –3 –2 –1 0 0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Vaccine cost ($)

Fig. 4.3.4 Influenza vaccine cost, normal distribution, and gamma distribution.

vaccine cost data are skewed, meaning they are not symmetrically distrib-
uted around the mean. If we use descriptive statistics from this dataset to
produce a normal distribution, it will not provide an accurate model for the
skewed data.
A normal distribution is depicted by the red curve on Fig. 4.3.4. It was con-
structed using the mean ($3.00) and standard error ($2.34) from the vaccine
cost data and is scaled to estimate the frequency of different costs expected in
a cohort of 1,000 observations. This distribution assigns non-​negative like-
lihood to negative costs. Negative unit vaccine costs are impossible; no be-
nevolent manufacturer is paying insurers to provide the seasonal influenza
vaccine in South Africa. We must find an alternative distribution to describe
this parameter. The alternative distribution would preferably maintain the
observed mean and variance from the data but must not allow sampling of
negative unit costs.
While many distributions can be employed to describe uncertainty in
a random variable, the choice of distribution for a parameter in a deci-
sion model is often straightforward. The core parameters in most decision
models relate to costs, utilities, probabilities, and treatment effect sizes.
Each of these parameter types is well described by a limited number of
common distributions. Appendix 6 includes a detailed description of some
candidate distributions for different types of parameters. In addition, it
302 Probabilistic sensitivity analysis and value of information analysis

describes how distributions can be operationalized when conducting PSA

in Microsoft Excel.

4.3.2.5 Interpreting probabilistic sensitivity

analysis results

Once a PSA has been conducted, there are several ways to report its results.
These include reporting expected values derived from simulations, reporting
uncertainty intervals for estimates, and visually presenting results in scatter
plots and cost-​effectiveness acceptability curves.

4.3.2.5.1 Expected outcomes and confidence intervals

Results from a PSA can be used to compute the expected value of a model.
Effectively, this involves averaging over all model outcomes recorded in the
PSA. As discussed in Chapter 2.3, computing the expected value of the model
through PSA produces an unbiased estimate of model outcomes. In the sea-
sonal influenza example, the 95% CI for QALYs gained from vaccination in
elderly individuals was 140–​525. The 95% CI for costs was $282,000–​$939,000.
Net monetary benefit (NMB) is a more useful measure than the ICER for
representing uncertainty in cost-​effectiveness estimates. ICERs are not con-
tinuously defined. If no change in outcome is observed, the denominator is
0 and the ICER is undefined. In addition, two ICERs may have very different
interpretations. Negative ICERs can represent cost-​saving or dominated
treatment strategies. Finally, when analyzing multiple treatment strategies,
the relative ranking of strategies may vary in ICER calculations. NMB has the
benefit of being linear and continuously defined. Additionally, when com-
paring the NMB of multiple treatment strategies, the option with the highest
value is always the most cost-​effective.

4.3.2.5.2 Cost-​effectiveness scatter plots

Cost-​effectiveness scatter plots are a visual means of presenting the results
from a PSA. These are cost-​effectiveness planes with the outcomes of all PSA
iterations presented together. It is common to present all results incremental
to a standard comparator, which is represented at the origin of the figure.
Fig. 4.3.3 shows the cost-​effectiveness scatter plot for the seasonal influenza
vaccine example.
When two interventions are being considered, we can directly calculate
the probability that an intervention is cost-​effective through observation
 4.3.2 Probabilistic sensitivity analysis 303

of the cost-​effectiveness scatter plot. First, we plot a cost-​effectiveness

threshold on a scatter plot of incremental costs and incremental QALYs
for an intervention versus a comparator. Next, we count the number of
PSA iterations that lie below the cost-​effectiveness threshold. In Fig. 4.3,
81% of PSA iterations produced cost-​effectiveness estimates below a
threshold of $3,040/​QALY. We can infer that the treatment has an 81%
probability of being cost-​effective. While the process is slightly more com-
plicated for multiple treatment options, the general process is the same.
In this situation, using NMB to represent cost-​effectiveness simplifies the
calculation.
There is typically some relationship between costs and effects in a de-
cision modelling study. For example, if a treatment successfully prevents
disease-​related events it may also reduce costs associated with these events.
This means that we cannot infer the uncertainty in cost-​effectiveness di-
rectly from the confidence intervals for costs and effects. Instead, we need
to represent the joint density of costs and effects. Visually, uncertainty
in cost-​effectiveness can be represented with confidence ellipses around
points on a cost-​effectiveness scatter plot, fitted using a system of equations
that account for potential correlations between costs and effects (Briggs,
O’Brien, & Blackhouse, 2002; Nixon, Wonderling, & Grieve, 2010).

4.3.2.5.3 Cost-​effectiveness acceptability curves

Suppose that the cost-​ effectiveness threshold in Fig. 4.3.3 was in-
creased to $5,000/​QALY. More PSA iterations would now fall below the
threshold on the cost-​effectiveness scatter plot. Our estimated probability
of the treatment being cost-​effectiveness would increase. Intuitively, this
makes sense: the more a decision maker is willing to pay for an increase
in health, the likelier a novel treatment which improves health will be
cost-​effective.
It is often informative to show the likelihood that a treatment is cost-​
effective over a range of different cost-​effectiveness thresholds. We can
count the number of PSA iterations for a treatment that are cost-​effective
at a range of different thresholds and present these graphically. This graph
is called a cost-​effectiveness acceptability curve. Fig. 4.3.5 shows the cost-​
effectiveness acceptability curve for vaccinating elderly individuals in
South Africa against seasonal influenza. A threshold of $3,040/​QALY is
represented on the cost-​effectiveness acceptability curve by a dashed line.
As noted previously, vaccinating this population has an 81% probability of
being cost-​effective at this threshold.
304 Probabilistic sensitivity analysis and value of information analysis

100%
90%
80%
Probability cost-effective

70%
60%
50%
40%
30%
20% Seasonal influenza
vaccination for elderly
10%
South Africa CE threshold
0%
0

10

20

30

40

50

60

70

80

90

10
0

,
00
0

0
Cost-effectiveness threshold ($/QALY)

Fig. 4.3.5 Cost-​effectiveness acceptability curve. CE, cost-​effectiveness.

4.3.3 Value of information analysis

The results from a PSA can be used to estimate the value of further re-
search. There will always be some uncertainty surrounding decisions to
implement new health technologies. Reducing this uncertainty lowers
the likelihood that a decision maker will implement a treatment that is
not cost-​effective or reject one that is cost-​effective. Value of information
analysis quantifies the gains in expected health benefit associated with re-
duced uncertainty.
Three measures are fundamental in value of information analysis. These
are expected value of perfect information (EVPI), expected value of perfect
parameter information (EVPPI), and expected value of sample information
(EVSI). This section will describe each of these measures and how they can
be used to inform healthcare decision-​making.
As a motivating example, consider a football team that needs to score more
goals and decides to recruit a new striker. They have two potential signings on
their shortlist. Club statisticians use data from previous seasons to estimate
how many goals each striker will score in the upcoming season. Some uncer-
tainty will surround each of these estimates, driven by variables which can be
studied (e.g., age, fitness, form). If the club recruited a clairvoyant manager,
they could spend their money on the player who will definitely score the most
 4.3.3 Value of information analysis 305

goals.1 There may be some benefit in resolving uncertainty in one specific

parameter. For example, perfect information regarding each player’s future
injury risk would be beneficial. Finally, even without perfect information,
collecting additional information on each player’s fitness would be useful as
it would ultimately reduce uncertainty in expected goals scored. These three
options align with the concepts of EVPI, EVPPI, and EVSI, respectively.

4.3.3.1 Expected value of perfect information

The EVPI refers to the amount of money a decision maker would pay for all
uncertainty in a decision to be resolved. A clairvoyant decision maker will
always make correct implementation decisions. Hence, they avoid any losses
associated with incorrect decision-​making.
When choosing which of multiple mutually exclusive treatments to imple-
ment, a decision maker must consider the expected cost-​effectiveness of these
treatments. With current, imperfect information, it is possible that an incor-
rect decision will be made. This would occur if the most cost-​effective treat-
ment was not implemented. A decision maker with perfect information will
always implement the most cost-​effective treatment.
When assessing multiple interventions, there is uncertainty around model
parameters. A set of different eventualities exists. Each combination of model
parameters reflects a different eventuality. Only one of these eventualities is
the true state of the world. Depending on the eventuality modelled, the rela-
tive cost-​effectiveness of the interventions will vary. In each eventuality, only
one intervention will be optimal. A decision maker with perfect information
knows which eventuality will occur and will always choose to implement the
optimal treatment strategy for this situation. A decision maker with imper-
fect information will choose to implement the treatment with the highest
expected cost-​effectiveness, weighted by the probability of each eventuality
occurring.
The mathematical procedure for estimating EVPI is outlined in Appendix
6. It involves probabilistically running a model multiple times and simulating
the decisions that would be made by decision makers with current and perfect
information.

1 Such a clairvoyant manager has not been on the job market since Martin O’Neill managed the Celtic

football team in Glasgow, Scotland, from 2000 to 2005 taking the team to win 75% of its games.
306 Probabilistic sensitivity analysis and value of information analysis

125,000

100,000

75,000
EVPI ($)

50,000

25,000

0
0

10

20

30

40

50

60

70

80

90

10
00

00

00

00

00

00

00

00

00

,0
00
Cost-effectiveness threshold ($/QALY)

Fig. 4.3.6 Expected value of perfect information versus cost-​effectiveness threshold.

Fig. 4.3.6 shows the EVPI for the South African seasonal influenza vaccine
program at different cost-​effectiveness thresholds. A threshold of $3,040/​
QALY is represented by a dashed line. The EVPI is near 0 when the threshold
is near 0 because there is strong certainty that the vaccination program will
generate incremental costs. At low thresholds, QALY gains are assigned little
value and it is unlikely that an expensive new treatment will be cost-​effective.
Therefore, the expected NMB of not vaccinating, E(NMBno vaccine ), is greater
than expected NMB of the vaccination program, E(NMBvaccine ). The decision
maker with current information will reject the program and current informa-
tion is adequate to support this decision.
We know that the vaccine likely produces health benefits and as the
threshold increases, these benefits are assigned higher value. Hence, the
probability that the vaccination program is cost-​effective will increase as the
threshold increases. While the threshold remains low (below $2,090/​QALY)
but increases, the decision maker with current information will still choose
not to implement the vaccination program. As the threshold grows, health
gains are assigned more value and the possibility that this decision is incorrect
grows. Accordingly, the value of obtaining more information grows.
An inflection point occurs at a threshold of $2,090/​QALY, the ICER associ-
ated with the vaccination program. At this threshold, E(NMBvaccine ) becomes
 4.3.3 Value of information analysis 307

larger than E(NMBno vaccine ) and the decision maker with current information
will choose to implement the vaccination program. As the threshold increases
beyond the inflection point, health benefits are assigned progressively greater
value and the probability that implementing the vaccination program was
cost-​ineffective falls. Hence the possibility of an incorrect decision with cur-
rent information declines alongside the value of additional research.
The EVPI can help determine whether a decision maker should fund addi-
tional research. Future research will reduce the gap between expected NMB
with perfect information and expected NMB with current information.
Spending more than EVPI on this research will lead to an expected net loss in
population health. If EVPI is greater than the proposed cost of a study, then
the research funds would be better spent elsewhere in the healthcare budget.
We can therefore regard EVPI as an upper bound that a healthcare decision
maker should spend on a research project regarding a treatment decision.
The EVPI associated with a healthcare decision does not imply a cost at
which research should be pursued. It is necessary but not sufficient that fur-
ther research costs less than the EVPI. For the seasonal influenza vaccine in
South Africa, a decision maker may desire additional information on the cost-​
effectiveness of vaccinating elderly individuals. According to Fig. 4.3.6, the
EVPI is approximately $31,000 at a cost-​effectiveness threshold of $3,040/​
QALY. If the cost required to conduct a proposed trial is below this value, a
decision maker cannot not rule out conducting the research. The sample size
for a trial of that cost may be too small to plausibly change a treatment deci-
sion. Further information, namely the EVSI of the proposed study, is required
before pursuing this study.

4.3.3.2 Expected value of perfect parameter

information

Clinical research generally focuses on a small set of parameters rather than

the entire uncertainty in a treatment decision. Randomized controlled trials
of new vaccines generally examine the efficacy or effectiveness of a vaccine.
Some studies focus on alternative parameters (e.g., costs, adverse events).
Hence, research typically reduces uncertainty in a limited number of param-
eters. The EVPPI quantifies the expected value associated with removing one
source of uncertainty from the decision-​making calculus.
Conceptually, EVPPI is very similar to EVPI; it is equal to the difference
between expected NMB with perfect parameter information and the expected
308 Probabilistic sensitivity analysis and value of information analysis

NMB with current information. The expected NMB with perfect parameter
information is equal to the expected NMB for a decision maker with perfect
information regarding some proper subset of the decision-​making param-
eters. The mathematical formulation of EVPPI is presented in Appendix 6.
If there are multiple sources of uncertainty in a healthcare decision, calcu-
lating the EVPPI for various uncertain parameters may help guide future re-
search. The parameter with the highest EVPPI is the parameter which would
be most beneficial to have complete information on. Of course, it may be more
costly to study some parameters than others.
As before, the EVPPI associated with a decision provides an upper bound
to the resources that should be expended on future research. If a randomized
controlled trial which assesses the effectiveness of a vaccine will cost less than
the EVPPI, a decision maker should not necessarily fund this trial. Estimating
EVSI would help to determine whether the research should be pursued.

4.3.3.3 Expected value of sample information

The EVPI and EVPPI associated with a decision estimate the maximum amount
of money that should be spent on new research. Simply costing less than EVPI
and EVPPI is not sufficient to justify funding a new study. Collecting additional
information will not eliminate uncertainty in a decision but will increase the
likelihood that the most cost-​effective treatment strategy will be adopted. The
EVSI estimates the value of pursuing a specific new research study, dependent
on an expected reduction in uncertainty. This estimate can be used to deter-
mine how much money should be spent on a new study.
The EVSI represents the gain in NMB that a decision maker expects to
achieve with new information. It is equal to the difference between the ex-
pected NMB with current information and the expected NMB with sample
information. The process for calculating expected NMB with sample infor-
mation is outlined in Appendix 6. It accounts for existing levels of uncertainty
and the type of data being collected by a proposed study.

References
Arbyn, M., Xu, L., Simoens, C., & Martin-​Hirsch, P. P. (2018). Prophylactic vaccination against
human papillomaviruses to prevent cervical cancer and its precursors. Cochrane Database of
Systematic Reviews, 5(5), CD009069. doi:10.1002/​14651858.CD009069.pub3
Arrow, K., & Lind, R. C. (1970). Uncertainty and the evaluation of public investment decisions.
American Economic Review, 60(3), 364–​378. https://​Eco​nPap​ers.repec.org/​RePEc:aea:aec​
rev:v:60:y:1970:i:3:p:364-​78
 4.3.3 Value of information analysis 309

Bojke, L., Claxton, K., Sculpher, M., & Palmer, S. (2009). Characterizing structural uncertainty
in decision analytic models: A review and application of methods. Value in Health, 12(5),
739–​749. doi:10.1111/​j.1524-​4733.2008.00502.x
Briggs, A. (2000). Handling uncertainty in cost-​effectiveness models. Pharmaco­Economics,
17(5), 479–​500. doi:10.2165/​00019053-​200017050-​00006
Briggs, A., O’Brien, B. J., & Blackhouse, G. (2002). Thinking outside the box: Recent advances
in the analysis and presentation of uncertainty in cost-​effectiveness studies. Annual Review
of Public Health, 23, 377–​401. doi:10.1146/​annurev.publhealth.23.100901.140534
Claxton, K. (1999). The irrelevance of inference: A decision-​making approach to the sto-
chastic evaluation of health care technologies. Journal of Health Economics, 18(3), 341–​364.
doi:10.1016/​s0167-​6296(98)00039-​3
Edoka, I., Kohli-​Lynch, C. N., Fraser, H., Hofman, K., Tempia, S., McMorrow, M.,... Cohen, C.
(2021). A cost-​effectiveness analysis of South Africa’s seasonal influenza vaccination pro-
gramme. Vaccine, 39(2), 412–​422. doi:10.1016/​j.vaccine.2020.11.028
Edoka, I., & Stacey, N. K. (2020). Estimating a cost-​effectiveness threshold for health care
decision-​making in South Africa. Health Policy and Planning, 35(5), 546–​555. doi:10.1093/​
heapol/​czz152
Greenhalgh, T. (1997). How to read a paper: Getting your bearings (deciding what the paper is
about). BMJ, 315(7102), 243. doi:10.1136/​bmj.315.7102.243
Hatswell, A. J., Bullement, A., Briggs, A., Paulden, M., & Stevenson, M. D. (2018). Probabilistic
sensitivity analysis in cost-​effectiveness models: Determining model convergence in co-
hort models. PharmacoEconomics, 36(12), 1421–​1426. https://​Eco​nPap​ers.repec.org/​
RePEc:spr:pha​rme:v:36:y:2018:i:12:d:10.1007​_​s40​273-​018-​0697-​3
International Decision Support Initiative. (2017). IDSI reference case for economic evaluation.
https://​www.ids​ihea​lth.org/​resou​rce-​items/​idsi-​refere​nce-​case-​for-​econo​mic-​eva​luat​ion
Kohli-​ Lynch, C. N., & Briggs, A. H. (2019). Heterogeneity in cost-​ effectiveness analysis.
Oxford: Oxford University Press.
Leal, J., Wordsworth, S., Legood, R., & Blair, E. (2007). Eliciting expert opinion for ec-
onomic models: An applied example. Value in Health, 10(3), 195–​ 203. doi:10.1111/​
j.1524-​4733.2007.00169.x
National Institute for Health and Clinical Excellence. (2013). Guide to the methods of technology
appraisal 2013. https://​www.nice.org.uk/​proc​ess/​pmg9/​resour​ces/​guide-​to-​the-​meth​ods-​
of-​tec​hnol​ogy-​apprai​sal-​2013-​pdf-​200797​5843​781
Nixon, R., Wonderling, D., & Grieve, R. (2010). Non-​parametric methods for cost-​effectiveness
analysis: The central limit theorem and the bootstrap compared. Health Economics, 19,
316–​333. doi:10.1002/​hec.1477
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M.,... Ganiats, T. G.
(2016). Recommendations for conduct, methodological practices, and reporting of cost-​
effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and Medicine. JAMA,
316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
Sculpher, M. (2008). Subgroups and heterogeneity in cost-​effectiveness analysis. Pharmaco­
Economics, 26(9), 799–​806. doi:10.2165/​00019053-​200826090-​00009
4.4
Economic evaluation reference case
with Markov model
William V. Padula, Shreena Malaviya, Natalie M. Reid, Jonothan Tierce,
and G. Caleb Alexander

Sometimes the best way to understand the use of a classical infectious di-
sease Markov model such as the Susceptible–​Exposed–​Infected–​Recovered
(SEIR) framework is to see it applied to an actual example.
This chapter draws material from an earlier technical report based on
a model developed in the early phases of the coronavirus disease 2019
(COVID-​19) outbreak in the US to better under the value of vaccines rel-
ative to treatments. The data in the model were calibrated to reach the
date of September 22, 2020 when the US surpassed the grim milestone of
200,000 deaths from COVID-​19. Although this infectious disease has had
a prolonged and profound fatal impact on societies worldwide, these early
epidemiologic data combined with our understanding of the economics
of vaccines provide some valuable information on the opportunity costs
avoided with the introduction of a vaccine.
This chapter illustrates some of the conceptual elements of a typical cost-​
effectiveness analysis of a vaccine being considered for adoption. The full
technical report is available on SSRN as report number 358664 entitled,
“Economic value of treatment and vaccine to address the COVID-​19 pan-
demic: A U.S. cost-​effectiveness and budget impact analysis.”

4.4.1 Introduction

The severe acute respiratory syndrome coronavirus 2 (SARS-​CoV-​2) virus

causing COVID-​19 rapidly transmitted through nearly 200 countries, infected
hundreds of millions of individuals, and resulted in millions of deaths since its
onset in China in late 2019. While dozens of countries were heavily impacted,

William V. Padula, Shreena Malaviya, Natalie M. Reid, Jonothan Tierce, and G. Caleb Alexander, Economic evaluation reference
case with Markov model In: Handbook of Applied Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and
William V. Padula, Oxford University Press. © Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0025
 4.4.1 Introduction 311

the US has experienced one of the highest levels of early-​onset morbidity and
mortality from the pandemic, with over 4 million individuals infected and
over 200,000 succumbing from COVID-​19 as of September 22, 2020 (Johns
Hopkins Coronavirus Resource Center, 2020). Widely cited projections avail-
able in September 2020 suggested that circumstances would worsen before
they improved (Institute for Health Metrics & Evaluation, 2020).
The magnitude of morbidity and mortality from the pandemic, as well as
its economic impact and effect on all sectors of society, galvanized the sci-
entific community and unleashed enormous activity devoted to identifying
treatments that reduce viral replication or host response, as well as vaccines
that ultimately diminish viral transmissibility. A review of World Health
Organization and US clinical trial registries revealed over 300 clinical trials
testing the therapeutic benefits of 92 drugs or plasma against COVID-​19,
including 64 in monotherapy and 28 different combinations (J. M. Sanders,
Monogue, Jodlowski, & Cutrell, 2020) by late 2020. Vaccines to protect against
COVID-​19 infection (Park, 2020), were presumed an important tool in
COVID-​19 control.
Treatments and vaccines offer different value propositions. Treatments re-
duce the downstream disease burden on the health system and are only taken
on an as-​needed basis. Yet, they are only useful among those who become in-
fected, they do not prevent the initial spread of infection, especially among
asymptomatic carriers, and they typically require a provider’s prescription.
Vaccines are valued differently. Vaccines should be given to as many
who have access and are highly vulnerable. As the number of vaccinated
people grows, non-​vaccinated individuals benefit through herd immunity
(Mauskopf et al., 2018). Vaccines have the benefit of preventing the down-
stream effects of disease, keeping societies more productive and worry free,
and reducing the need for isolating social measures. However, vaccines are
often not as efficiently targeted as treatments in the sense that they must be
given to many individuals who never would have contracted the disease,
leading to high upfront costs for governments and payers that invest in
their coverage (Carvalho, Jit, Cox, Yoong, & Hutubessy, 2018).
Despite the catastrophic nature of the COVID-​19 pandemic and the un-
precedented scientific effort to address it, the relative value of a treatment or
vaccine for COVID-​19 had not been immediately characterized (Apuzzo &
Kirkpatrick, 2020). We quantified this value using economic modelling based
on rates of COVID-​19 infection in the US as of September 22, 2020. This ap-
proach can be used to generate fundamental new knowledge regarding the
312 Economic evaluation reference case with Markov model

economic value of vaccines for infectious disease outbreak with respect to

commonly assumed alternatives (e.g., acute treatment, social distancing
measures).

4.4.2 Methods

4.4.2.1 Study design

We developed a Markov model to evaluate the cost-​effectiveness of disruptive

strategies developed in response to the COVID-​19 pandemic. We used the
model to compare two innovative technologies, a treatment or vaccination,
to current status quo for most communities in the US: doing nothing or social
distancing. The model was designed using a SEIR structure, which has been
used to construct a number of epidemiological and mathematical models
for the spread of COVID-​19 (Hou et al., 2020; Iwata & Miyakoshi, 2020; Wei
et al., 2020).
In accordance with guidelines established by the Second Panel on Cost-​
Effectiveness in Health & Medicine, we used a US societal perspective
to measure resource utilization, financial burden, and health effects (G.
D. Sanders et al., 2016). The model’s time horizon was 365 days with 1-​day
cycles. We used a discounting rate of 3% where applicable, and we adjusted
all costs for inflation to represent 2020 US dollars, measuring effectiveness in
units of quality-​adjusted life years (QALYs).
We sought several findings from the model results. By applying the eco-
nomic model to the roughly 330 million US residents according to the 2020
census, we were able to both calibrate model outcomes with current real-​
world data on disease outcomes as of September 22, 2020. We simulated the
impact of COVID-​19 and potential solutions on the US population in terms
of total cost, number of infections, number of hospitalizations, and mortality.
The model was also purposed to measure the incremental cost-​effectiveness
ratio (ICER) and budget impact. ICERs were interpreted at a willingness-​to-​
pay threshold of $50,000/​QALY.

4.4.2.2 Model structure

Our SEIR model simulated the transition of the US population through

ten mutually exclusive health states (Fig. 4.4.1). All patients began in a
 4.4.2 Methods 313

Recovered
R

Susceptible Exposed Infected

S E I

(Vaccination) (Social distancing)

(Treatment)

Death

Fig. 4.4.1 Markov model of disease progression with coronavirus disease 2019
(COVID-​19). Patients progressed through a modified “SEIR” process (Susceptible–​
Exposed–​Infected–​Recovered). The infection phases were staged from 0 through 4 in
terms of increasing escalation, including use of critical care services. Patients who did
not recover from COVID-​19 died. Model alternatives to doing nothing included social
distancing, treatment during the I1 infected phase, or a vaccination to avoid entry into
the susceptible phase.

susceptible (S) population. Over the course of time, patients would then
become exposed (E) to COVID-​19 infection. Upon exposure, a propor-
tion of the population would become infected (I). We assumed five sepa-
rate infected states (I0–​I4) based on current information about COVID-​19
(Cascella, Rajnik, Cuomo, Dulebohn, & Di Napoli, 2020; Shi et al., 2020).
The states that did not require hospitalization included (I0) asymptomatic,
(I1) uncomplicated or mild symptoms, and (I2) moderate symptoms without
signs of severe pneumonia. Patients who required hospitalization to manage
symptoms were broken up into two states: (I3) severe or (I4) critical (Shi
et al., 2020). Any patient who survived infection transitioned to recovered
(R). Patients in I3 or I4 also risked death.

4.4.2.3 Mitigation strategies

We customized the comparator arms of the model to capture health state

transitions as appropriate in reference to the base case. The treatment arm
considered a therapy at the onset of mild symptoms during the I1 phase to
increase the probability of transition to R prior to any chance of escalation.
In particular, we assumed an antiviral medication similar to zanamivir
314 Economic evaluation reference case with Markov model

(brand name: Relenza; manufacturer: GlaxoSmithKline) in terms of list

price, efficacy, and course of treatment for the base-​case analysis (Monto
et al., 2002). We also conducted uncertainty tests to evaluate scenarios
where a hypothetical treatment differed from zanamivir (e.g., higher price,
lower efficacy, etc.).
The vaccination arm separated a large proportion of the population
out of the susceptible pool, contributing to an assumed herd immunity
(Fine, Eames, & Heymann, 2011). In this scenario, we assumed that 60% of
the population would be vaccinated, consistent with recommendations by
Gavi, the Vaccine Alliance (Gavi, 2020e).

4.4.2.4 Model parameters

4.4.2.4.1  Costs
We applied micro-​costing methods to a literature review to calculate the costs
associated with each symptomatic infection state and prevention or treatment
strategy (Table 4.4.1) (Xu, Grossetta Nardini, & Ruger, 2014). The model state
costs were obtained from published literature. Individuals in health states S,
E, I0, and R were expected to have no additional costs. Infected patients were
assessed the cost of a lost work day, based on methods recommended by the
US Bureau of Labor Statistics. Cost of a lost work day was calculated from the
US gross domestic product of $21 trillion distributed among 330 million indi-
viduals, averaging 261 work days each to represent an average of $243.81 per
person per day (Bureau of Labor Statistics, 2008).
The societal costs of a new treatment or vaccine were assumed from the
US list prices for similar technologies, zanamivir or the influenza vaccine, in
accordance with World Health Organization methods guidance on costing
(World Health Organization, 2015).

4.4.2.4.2 Probabilities
Most model probabilities were obtained from official reports of COVID-​
19 outcomes, which were current as of September 22, 2020, through these
sources: the Johns Hopkins Coronavirus Resource Center (https://​coro​navi​
rus.jhu.edu/​, Baltimore, MD, USA); the US Centers for Disease Control
and Prevention (https://​www.cdc.gov/​coro​navi​rus/​2019-​ncov/​, Atlanta,
GA, USA); the World Health Organization (https://​www.who.int/​, Geneva,
Switzerland); and the Institute for Health Metrics and Evaluation (https://​covi​
d19.hea​lthd​ata.org/​uni​ted-​sta​tes-​of-​amer​ica, Seattle, WA, USA). In a small
number of instances where data were not available through these sources,
Table 4.4.1 Model parameters based on infection statistics and assumptions as of September 22, 2020

Parameter Expected value Range for sensitivity analysis Source

Lower bound Upper bound

Infection population statistics

US population (2020) 330 million
Probability of social contact (b) 0.053 Yang et al. (2020)
Social contacts per person (c) 7.00 5.00 9.00 Eames, Tilston, Brooks Pollock, &
Edmunds (2012)
Inverse average infectious period (ɣ) 0.14 Yang et al. (2020)
Transmission rate (β) 0.37 (b × c)
Basic reproduction number (R0) 2.57 1.95 3.25 (β/​ɣ)
Social contacts per person with social 3.00 Assumed
distancing

Transition probabilities
S to E 0.307 0.000 1.00 Yang et al. (2020)
E to I 0.005 0.004 0.006 Johns Hopkins Coronavirus Resource
Center (2020); Yang et al. (2020)
I to I0 0.250 −0.429 0.929 Whitehead (2020)
I0 to R* 1.000 Assumed
I to I1* 0.750
I1 to I2* 0.191
I1 to R 0.809 0.000 1.000 Chinese Center for Disease Control and
Prevention (2020)
(continued)
Table 4.4.1 Continued

Parameter Expected value Range for sensitivity analysis Source

Lower bound Upper bound
I2 to I3 0.314 0.000 10.000 Centers for Disease Control and
Prevention (2020)
I2 to R* 0.686
I3 to I4 0.115 0.000 0.531 Centers for Disease Control and
Prevention (2020)
I3 to R* 0.859
I3 to death 0.026 0.000 0.084 Johns Hopkins Coronavirus Resource
Center (2020); Centers for Disease
Control and Prevention (2020)
I4 to R* 0.974
I4 to death 0.026 0.000 0.100 Johns Hopkins Coronavirus Resource
Center (2020); Centers for Disease
Control and Prevention (2020)
Effect modifiers of interventions
Treatment efficacy 0.820 0.000 1.000 Monto et al. (2002)
Vaccine efficacy 0.900 0.000 1.000 Assumed
Proportion vaccinated 0.600 0.000 1.000 Gavi (2020)
S to E with social distancing 0.157 0.000 1.000 Anderson, Heesterbeek, Klinkenberg, &
Hollingsworth (2020)
Risk reduction with social distancing 0.600 0.000 1.000 Anderson et al. (2020)
Risk reduction from herd immunity 0.500 0.000 1.000 Assumed
Daily costs (USD 2020)
I1 costs
Sick day $122.86 $92.15 $153.58 Bureau of Labor Statistics (2008)
Medications $0.95 $0.71 $1.19 Drugs.com (2020a, 2020c)
I2 costs
Sick day $252.87 $189.65 $316.09 Bureau of Labor Statistics (2008)
Medications $5.07 $3.80 $6.34 Drugs.com (2020a, 2020b)
Primary care visit and tests $1,248.70 $936.53 $1,560.88 Castlight Health (2020)
Urgent care visit and tests $1,553.38 $1,165.04 $1,941.73 Castlight Health (2020)
Emergency care visit and tests $9,253.25 $6,939.94 $11,566.56 Castlight Health (2020)
I3 and I4 costs
Primary care visit, tests, and X-​ray $8,850.19 $6,637.64 $11,062.74 Castlight Health (2020)
Urgent care visit, tests, and X-​ray $9,143.51 $6,857.63 $11,429.39 Castlight Health (2020)
Emergency care visit, tests, and X-​ray $16,843.31 $12,632.48 $21,054.14 Castlight Health (2020)
Hospital bed (I3) $1,432.57 $1,074.43 $1,790.71 Rae, Claxton, Kurani, McDermott, &
Cox (2020)
ICU bed (I4) $3,898.85 $2,924.14 $4,873.56 Rae et al. (2020)
Intervention costs
Treatment $1,000.00 $750.00 $1,250.00 Assumed
Vaccine $100.00 $75.00 $125.00 Assumed
(continued)
Table 4.4.1 Continued

Parameter Expected value Range for sensitivity analysis Source

Lower bound Upper bound

Utilities (QALYs)
Utility of susceptible (S) 0.880 0.018 1.000 Khan, Muennig, Gardam, & Zivin,
2005; Sullivan & Ghushchyan (2006)
Utility of exposed (E) 0.880 0.018 1.000 Khan et al. (2005)
Utility of infected (I) 0.833 0.017 1.000 Khan et al. (2005)
Utility of asymptomatic (I0) 0.833 0.017 1.000 Khan et al. (2005)
Utility of mild symptoms (I1) 0.614 0.012 1.000 Yang et al. (2020)
Utility of moderate symptoms (I2) 0.500 0.010 0.990 Khan et al. (2005)
Utility of severe symptoms (I3) 0.250 0.005 0.495 Khan et al. (2005)
Utility of critical care (I4) 0.050 0.001 0.099 Khan et al. (2005)
Utility of death 0.000 Anchor
Utility of recover (R) 0.880 0.018 1.000 Khan et al. (2005); Sullivan &
Ghushchyan (2006)
Utility with vaccine 0.900 0.018 1.000 Khan et al. (2005); Lee et al. (2015);
Sullivan & Ghushchyan (2006)
 4.4.2 Methods 319

such as for historical outcomes of comparators based on influenza treatment

and vaccination, we obtained model probabilities from the peer-​reviewed sys-
tematic review and clinical trials literature.

4.4.2.4.3 Health utilities

We measured utilities in units of QALYs based on five-​dimensional EuroQol
(EQ-​5D) index scores collected from the US Medical Expenditure Panel
Survey (MEPS) (Shaw, Johnson, & Coons, 2005; Sullivan & Ghushchyan,
2006; Sullivan, Lawrence, & Ghushchyan, 2005). In general, QALYs range
from 0.0 (death) to 1.0 (full health). We assumed the US population average
was 0.88 QALYs (Sullivan & Ghushchyan, 2006). Patients infected and symp-
tomatic had utilities that reflected less desirable health states. In the hypothet-
ical state with the availability of a vaccine for COVID-​19, we assumed a higher
population average of 0.90 QALYs since utilities are a reflection of prefer-
ences, and individuals would likely have preferred the availability of vaccines
to forgo the threat of an infectious disease (Lee et al., 2015).

4.4.2.5 Sensitivity analysis

We conducted sensitivity analyses to test model uncertainty by varying the ex-

pected parameter values (Briggs, 2000). Parameter value uncertainty ranged
within the reported 95% confidence interval or standard deviation. In rare cir-
cumstances where no uncertainty range was reported, we varied the param-
eter by ±25% of its mean/​median in order to assess the impact on the model
results. One-​way sensitivity analyses were done first. This approach produced
a tornado diagram to evaluate parameters that had the greatest univariate im-
pact on model results (i.e., that a single parameter could change the ICER in-
sofar as to adjust our reporting of overall findings between interventions as
“cost-​effective” or “cost-​ineffective”).
A Bayesian multivariate probabilistic sensitivity analysis was performed
using 10,000 Monte Carlo simulations with no intervention as the baseline
approach (Briggs, Goeree, Blackhouse, & O’Brien, 2002). The probabilistic
sensitivity analysis applied a distribution for each variable to characterize the
impact of uncertainty on all parameters simultaneously. We used beta distri-
butions for variables with values ranging between 0.0 and 1.0 such as prob-
abilities and utilities, and gamma distributions for positive values greater than
1.0, such as costs.
320 Economic evaluation reference case with Markov model

4.4.3 Results

4.4.3.1 Estimated costs of the pandemic

Using this economic model, we were able to estimate the economic impact
of the COVID-​19 pandemic across 330 million people in the US. The base-
line approach, doing nothing, could be associated with about $699 billion in
financial impact to healthcare and labor sectors based on a combination of
direct medical costs and lost wages due to sick leave (Table 4.4.2). These costs
were mostly represented by about 67 million hospital days and over 200,000
deaths.
By contrast, considering our early understanding of the impact of social
distancing, COVID-​19 would have a lessened economic impact of about
$548.6 billion. Thus, social distancing presented a potential savings of nearly
$150 billion. This approach reduced the number of simulated outcomes across
sectors, including 14.5 million fewer hospital days and about 50,000 deaths
averted.

4.4.3.2 Effect of treatments and vaccines

Both treatments and vaccines were associated with large reductions in ex-
pected morbidity and costs. For example, the availability of a treatment pre-
sented a total cost of only $66.6 billion, a 90% cost reduction compared with
doing nothing. This approach was associated with 31 million hospital days,
and about 106,000 deaths.
By contrast, the availability of a vaccine reduced societal costs to only $9.9
billion, a 98% cost reduction compared with doing nothing. The presence of
a vaccine and herd immunity for many improved outcomes to only about
30 million hospital days, and 104,000 deaths.

4.4.4 Cost-​effectiveness findings

Under the base-​case findings of our cost-​effectiveness analysis, each proposed

intervention offered greater value to US society than the prospect of doing
nothing. In particular, all three options (social distancing, treatment, and vac-
cination) dominated doing nothing by lowering societal costs and increasing
QALYs. On an individual basis, treatment and vaccination were more efficient
Table 4.4.2 Cost-​effectiveness analysis comparing mitigation interventions to no intervention to address US COVID-​19 pandemic

Comparator Economic impact Epidemiologic impact

Cost Δ cost Effectiveness Δ effectiveness ICER Program cost Budget impact Hospital-​days Deaths
($) ($) (QALYs) (QALYs) ($/​QALY) ($ billions) ($ per person per (N) (N)
month)

Do nothing 2,115 0.874 697.83 176.22 67,166,963 227,513

Social distancing 1,738 −377 0.875 0.001 Dominates 548.85 138.60 52,687,745 177,981
Treatment 1,299 −885 0.877 0.003 Dominates 66.56 16.81 31,349,944 106,186
Vaccination 999 −1115 0.892 0.019 Dominates 9.90 2.50 30,784,605 104,265

ICER, incremental cost-​effectiveness ratio; QALY, quality-​adjusted life year.

Dominates indicates that the alternative is preferred at a lower cost and higher effectiveness relative to “do nothing.”
322 Economic evaluation reference case with Markov model

than social distancing alone, although a vaccination program would provide

greater QALY gains and reduce costs further than a treatment option based on
currently available data about COVID-​19 outcomes.
In terms of budget impact, both treatment and vaccination represent
lower-​cost alternatives to US society and payers, who would have to front
the initial costs of technologies compared with long-​run costs of disease
burden on health systems. If we assume that the costs of technology can be
financed across a risk pool of 330 million Americans in a 12-​month period,
then doing nothing comes at a societal budget impact of $176 per person per
month. By comparison, the budget impacts of treatment, $16.81 per person
per month, and vaccination, $2.50 per person per month, are much more
affordable.

4.4.5 Sensitivity analysis

In one-​way sensitivity analysis, uncertainty intervals around parameter es-

timates did not change the results of the model from the main conclusions
(Fig. 4.4.2). Parameters with the greatest impact on model results included
treatment efficacy, treatment cost, societal cost of social distancing, vaccination

Incremental cost-effectiveness ratio

(Icer, $/QALY)
200,000 150,000 100,000 50,000 0 50,000 100,000 150,000

Treatment efficacy

Treatment cost

Social distancing cost

Rate of vaccination

Vaccine efficacy

Probability of infectivity

Vaccine cost

Number of social contacts

ICER−25% ICER+25%

Fig. 4.4.2 Tornado diagram of one-​way sensitivity analysis for parameter variabilities
with greatest impact on study results.
 4.4.6 Discussion 323

Cost effectiveness acceptability curve

1
Probability cost-effective 0.9
0.8
0.7
0.6
0.4
0.5
0.3
0.2
0.1
0
$0 $50,000 $100,000 $150,000 $200,000 $250,000
Willingness-to-pay threshold ($/QALY)

Do nothing Social distancing Treatment Vaccine

Fig. 4.4.3 Cost-​effectiveness acceptability curve of the probability that a comparator is

cost-​effective at a given willingness-​to-​pay threshold, based on results of a probabilistic
sensitivity analysis of 10,000 Monte Carlo simulations.

rate, and vaccine efficacy. The probabilistic sensitivity analysis confirmed

findings from the base-​case cost-​effectiveness analysis. At a willingness-​to-​
pay threshold of $100,000/​QALY, the vaccination option was cost-​effective
in 71% of simulations (Fig. 4.4.3). Treatment was also cost-​effective at this
threshold in 21% of simulations. Social distancing was cost-​effective in 8% of
simulations.

4.4.6 Discussion

The COVID-​19 pandemic has resulted in enormous morbidity, mortality,

and societal upheaval worldwide. Given this, as well as uncertainty regarding
the pandemic’s future course, tremendous scientific effort was devoted to
identifying treatments and vaccines that reduced the likelihood or conse-
quences of coronavirus infection. We used economic modeling to explore the
relative value of such technologies. We found that both alternatives offered
more efficient options compared to doing nothing or social distancing.
While treatments represent efficient technology to address individual
COVID-​19 cases, a vaccination program comes at a lower cost, offers greater
effectiveness, and averts more deaths and hospital days (Armstrong, 2007; Van
Damme & Beutels, 1996). Perhaps these data imply the extended dominance
324 Economic evaluation reference case with Markov model

of a vaccine over treatment for COVID-​19 in particular. One reason for this is
that treatments introduce additional burdens on the healthcare system since,
presumably, payers would expect providers to conduct appropriate testing
and write prescriptions for the treatment to authorize reimbursement. In ad-
dition, treatments do not minimize the asymptomatic spread of infection,
which we now know represents a substantial proportion of infected cases; nor
is it clear how treatments would impact people’s behavior to self-​isolate (Bai
et al., 2020; Salathe et al., 2020). Furthermore, the treatment options do not
sufficiently alleviate the disproportionate impact on the most vulnerable of
our population, including the elderly, those who are chronically ill, and racial
and ethnic minorities (Devakumar, Shannon, Bhopal, & Abubakar, 2020; Liu,
Chen, Lin, & Han, 2020; McMichael et al., 2020). Thus, compared to vacci-
nation, treatment would prove to be a much larger burden on the healthcare
system in the long run if additional measures are not implemented to stop the
spread of infection.
As with all models, ours had several limitations. First, our model did not
completely demonstrate the time dependency of the COVID-​19 epidemic
based on a limited understanding of the virus at this time. Transition prob-
abilities are usually regressed over several time periods of seasonal data,
whereas our understanding of COVID-​19 infection globally was based on a
few short months of initial reports between March and September of 2020.
That being said, we ranged all model parameters within credible uncertainty
intervals for the sensitivity analysis to provide a realistic outlook for multiple
scenarios. Second, our model presumes a static rather than dynamic popu-
lation, and thus, we do not consider the entry or exit of individuals due to
birth, or death from other causes. Third, our approach did not account for
the different proportions and corresponding risks of age groups in the US
population, yet the risks of COVID-​19 infection may vary considerably by
sociodemographic factors. As more data became available, this model pro-
vided a useful template for subgroup analyses. Fourth, our results varied
depending on the actual costs and efficacies of treatments and vaccines that
undergo development. Fifth, the model assumed availability of important re-
sources that payers often leverage as gatekeepers to authorize treatments, such
as diagnostic tests. We acknowledge that global shortages of COVID-​19 diag-
nostic tests biased our estimates since the cost of disease management without
technological intervention would be more efficient. Sixth, the health utilities
for this study were based on values of previous studies about SARS infection.
Finally, we did not capture complete information on the costs to payers and
the healthcare system at this time.
 4.4.6 Discussion 325

In conclusion, this economic model represented an attempt to understand

the burden of COVID-​19 on the US economy, particularly on the healthcare
and labor sectors. A treatment or vaccine for COVID-​19 presented excep-
tional potential value to the US healthcare system and economy even early on
in the outbreak before many other countries faced similarly concerning rates
of infection and mortality. Vaccine-​based COVID-​19 control options likely
could save 90% or more on COVID-​19 costs to US society while significantly
improving clinical benefit for millions of individuals. While many treatments
and vaccines are undergoing clinical development, there are two practical is-
sues that governments, health systems, and payers should begin to address
now. One concern is the budget impact that a single technology could have
if uptake is high. The other concern is that this demand will necessitate ca-
pacity building to properly allocate these technologies to the right individuals.
Governments should begin mapping out policies for cross-​subsidies and an
expanded healthcare infrastructure with other healthcare stakeholders so that
business transactions with manufacturers are not the hold up for patients to
receive medically necessary and beneficial technology to address the COVID-​
19 pandemic.

References
Anderson, R. M., Heesterbeek, H., Klinkenberg, D., & Hollingsworth, T. D. (2020). How will
country-​based mitigation measures influence the course of the COVID-​19 epidemic?
Lancet, 395(10228), 931–​934.
Apuzzo, M., & Kirkpatrick DD. (2020, April 1). COVID-​19 changed how the world does sci-
ence, together. New York Times. https://​www.nyti​mes.com/​2020/​04/​01/​world/​eur​ope/​coro​
navi​rus-​scie​nce-​resea​rch-​coop​erat​ion.html
Armstrong, E. P. (2007). Economic benefits and costs associated with target vaccinations.
Journal of Managed Care Pharmacy, 13(7 Suppl B), 12–​15. doi:10.18553/​jmcp.2007.13.
s7-​b.12
Bai, Y., Yao, L., Wei, T., Tian, F., Jin, D. Y., Chen, L., & Wang, M. (2020). Presumed asymp-
tomatic carrier transmission of COVID-​ 19. JAMA, 323(14), 1406–​1407. doi:10.1001/​
jama.2020.2565
Briggs, A. (2000). Handling uncertainty in cost-​effectiveness models. Pharmaco­Economics,
17(5), 479–​500. doi:10.2165/​00019053-​200017050-​00006
Briggs, A., Goeree, R., Blackhouse, G., & O’Brien, B. J. (2002). Probabilistic analysis of cost-​
effectiveness models: Choosing between treatment strategies for gastroesophageal reflux di-
sease. Medical Decision Making, 22(4), 290–​308. doi:10.1177/​0272989x0202200408
Bureau of Labor Statistics. (2008). Technical information about the BLS major sector productivity
and costs measure. Major sector productivity and costs. https://​www.bls.gov/​lpc/​lpc​meth​
ods.pdf
Carvalho, N., Jit, M., Cox, S., Yoong, J., & Hutubessy, R. (2018). Capturing budget impact con-
siderations within economic evaluations: A systematic review of economic evaluations of
326 Economic evaluation reference case with Markov model

rotavirus vaccine in low-​and middle-​income countries and a proposed assessment frame-

work. PharmacoEconomics, 36(1), 79–​90. doi:10.1007/​s40273-​017-​0569-​2
Cascella, M., Rajnik, M., Cuomo, A., Dulebohn, S. C., & Di Napoli, R. (2020). Features, evalua-
tion and treatment coronavirus (COVID-​19). Treasure Island, FL: StatPearls Publishing
Castlight Health. (2020). The costs of COVID-​19. https://​www.cast​ligh​thea​lth.com/​wp-​cont​
ent/​uplo​ads/​2020/​03/​Costs-​of-​COVID-​19.pdf
Centers for Disease Control and Prevention. (2020). Severe outcomes among patients with co-
ronavirus disease 2019 (COVID-​19). Morbidity and Mortality Weekly Report, 69(12), 343–​
346. doi:10.15585/​mmwr.mm6912e2
Chinese Center for Disease Control and Prevention. (2020). [The epidemiological characteris-
tics of an outbreak of 2019 novel coronavirus diseases (COVID-​19) in China.] Zhonghua Liu
Xing Bing Xue Za Zhi, 41(2), 145–​151. doi:10.3760/​cma.j.issn.0254-​6450.2020.02.003
Devakumar, D., Shannon, G., Bhopal, S. S., & Abubakar, I. (2020). Racism and discrimination
in COVID-​19 responses. Lancet, 395(10231), 1194. doi:10.1016/​S0140-​6736(20)30792-​3
Drugs.com. (2020a). Acetaminophen prices, coupons and patient assistance programs. Price guide.
https://​www.drugs.com/​price-​guide/​acetam​inop​hen#oral-​tab​let-​325-​mg
Drugs.com. (2020b). Benzonatate prices, coupons and patient assistance programs. Price guide.
https://​www.drugs.com/​price-​guide/​benz​onat​ate
Drugs.com. (2020c). Tussin expectorant. Price guide. https://​www.drugs.com/​otc/​102​841/​tus​
sin-​expe​ctor​ant.html
Eames, K. T., Tilston, N. L., Brooks Pollock, E., & Edmunds, W. J. (2012). Measured dynamic
social contact patterns explain the spread of H1N1v influenza. PLoS Computational Biology,
8(3), e1002425.
Fine, P., Eames, K., & Heymann, D. L. (2011). “Herd immunity”: A rough guide. Clinical
Infectious Diseases, 52(7), 911–​916. doi:10.1093/​cid/​cir007
Gavi. (2020). What is herd immunity? https://​www.gavi.org/​vacci​nesw​ork/​what-​herd-​
immun​ity
Hou, C., Chen, J., Zhou, Y., Hua, L., Yuan, J., He, S.,... Jia, E. (2020). The effectiveness of
the quarantine of Wuhan city against the corona virus disease 2019 (COVID-​ 19):
Well-​mixed SEIR model analysis. Journal of Medical Virology, 92(7), 841–​848. doi:10.1002/​
jmv.25827
Institute for Health Metrics & Evaluation. (2020). COVID-​19 projections assuming full social
distancing through May 2020: Total deaths. https://​covi​d19.hea​lthd​ata.org/​
Iwata, K., & Miyakoshi, C. (2020). A simulation on potential secondary spread of novel coro-
navirus in an exported country using a stochastic epidemic SEIR model. Journal of Clinical
Medicine, 9(4), 944. doi:10.3390/​jcm9040944
Johns Hopkins Coronavirus Resource Center. (2020). COVID-​19 map. Coronavirus COVID-​19
global cases by the Center for Systems Science and Engineering. https://​coro​navi​rus.jhu.edu/​
Khan, K., Muennig, P., Gardam, M., & Zivin, J. G. (2005). Managing febrile respiratory
illnesses during a hypothetical SARS outbreak. Emerging Infectious Diseases, 11(2), 191–​200.
doi:10.3201/​eid1102.040524
Lee, B. Y., Bartsch, S. M., Brown, S. T., Cooley, P., Wheaton, W. D., & Zimmerman, R. K. (2015).
Quantifying the economic value and quality of life impact of earlier influenza vaccination.
Medical Care, 53(3), 218–​229. doi:10.1097/​mlr.0000000000000302
Liu, K., Chen, Y., Lin, R., & Han, K. (2020). Clinical features of COVID-​19 in elderly pa-
tients: A comparison with young and middle-​aged patients. Journal of Infection, 80(6), e14–​
e18. doi:10.1016/​j.jinf.2020.03.005
Mauskopf, J., Standaert, B., Connolly, M. P., Culyer, A. J., Garrison, L. P., Hutubessy,
R.,... Severens, J. L. (2018). Economic analysis of vaccination programs: An ISPOR good
practices for outcomes research task force report. Value in Health, 21(10), 1133–​1149.
doi:10.1016/​j.jval.2018.08.005
 4.4.6 Discussion 327

McMichael, T. M., Currie, D. W., Clark, S., Pogosjans, S., Kay, M., Schwartz, N. G.,... Duchin, J. S.
(2020). Epidemiology of Covid-​19 in a long-​term care facility in King County, Washington.
New England Journal of Medicine, 382(21), 2005–​2011. doi:10.1056/​NEJMoa2005412
Monto, A. S., Pichichero, M. E., Blanckenberg, S. J., Ruuskanen, O., Cooper, C., Fleming, D. M.,
& Kerr, C. (2002). Zanamivir prophylaxis: An effective strategy for the prevention of influ-
enza types A and B within households. Journal of Infectious Diseases, 186(11), 1582–​1588.
doi:10.1086/​345722
Park, A. (2020, February 25). COVID-​19 vaccine shipped, and drug trials start. Time. https://​
time.com/​5790​545/​first-​covid-​19-​vacc​ine/​
Rae, M., Claxton, G., Kurani, N., McDermott, D., & Cox, C. (2020). Potential costs of coro-
navirus treatment for people with employer coverage. Health System Tracker. https://​www.
heal​thsy​stem​trac​ker.org/​brief/​potent​ial-​costs-​of-​coro​navi​rus-​treatm​ent-​for-​peo​ple-​with-​
emplo​yer-​cover​age/​
Salathe, M., Althaus, C. L., Neher, R., Stringhini, S., Hodcroft, E., Fellay, J.,... Low, N. (2020).
COVID-​19 epidemic in Switzerland: On the importance of testing, contact tracing and isola-
tion. Swiss Medical Weekly, 150, w20225. doi:10.4414/​smw.2020.20225
Sanders, G. D., Neumann, P. J., Basu, A., Brock, D. W., Feeny, D., Krahn, M.,... Ganiats, T. G.
(2016). Recommendations for conduct, methodological practices, and reporting of cost-​
effectiveness analyses: Second Panel on Cost-​Effectiveness in Health and Medicine. JAMA,
316(10), 1093–​1103. doi:10.1001/​jama.2016.12195
Sanders, J. M., Monogue, M. L., Jodlowski, T. Z., & Cutrell, J. B. (2020). Pharmacologic treat-
ments for coronavirus disease 2019 (COVID-​19): A review. JAMA, 323(18), 1824–​1836.
doi:10.1001/​jama.2020.6019
Shaw, J. W., Johnson, J. A., & Coons, S. J. (2005). US valuation of the EQ-​5D health
states: Development and testing of the D1 valuation model. Medical Care, 43(3), 203–​220.
Shi, Y., Wang, Y., Shao, C., Huang, J., Gan, J., Huang, X.,... Melino, G. (2020). COVID-​19 infec-
tion: The perspectives on immune responses. Cell Death & Differentiation, 27(5), 1451–​1454.
doi:10.1038/​s41418-​020-​0530-​3
Sullivan, P. W., & Ghushchyan, V. (2006). Preference-​based EQ-​5D index scores for chronic
conditions in the United States. Medical Decision Making, 26(4), 410–​420. doi:10.1177/​
0272989x06290495
Sullivan, P. W., Lawrence, W. F., & Ghushchyan, V. (2005). A national catalog of preference-​
based scores for chronic conditions in the United States. Medical Care, 43(7), 736–​749.
doi:10.1097/​01.mlr.0000172050.67085.4f
Van Damme, P., & Beutels, P. (1996). Economic evaluation of vaccination. PharmacoEconomics,
9(3), 8–​15. doi:10.2165/​00019053-​199600093-​00005
Wei, Y. Y., Lu, Z. Z., Du, Z. C., Zhang, Z. J., Zhao, Y., Shen, S. P.,... Chen, F. (2020). [Fitting and
forecasting the trend of COVID-​19 by SEIR(+​CAQ) dynamic model.] Zhonghua Liu Xing
Bing Xue Za Zhi, 41(4), 470–​475. doi:10.3760/​cma.j.cn112338-​20200216-​00106
Whitehead, S. (2020). CDC director on models for the months to come: “This virus is going to be with us.”
NPR. https://​www.npr.org/​secti​ons/​hea​lth-​shots/​2020/​03/​31/​824155​179/​cdc-​direc​tor-​on-​mod​
els-​for-​the-​mon​ths-​to-​come-​this-​virus-​is-​going-​to-​be-​with-​us
World Health Organization. (2015). WHO guideline on country pharmaceutical pricing policies.
https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​153​920/​978924​1549​035_​eng.pdf
Xu, X., Grossetta Nardini, H. K., & Ruger, J. P. (2014). Micro-​costing studies in the health and
medical literature: Protocol for a systematic review. Systematic Reviews, 3, 47. doi:10.1186/​
2046-​4053-​3-​47
Yang, Z., Zeng, Z., Wang, K., Wong, S.-​S., Liang, W., Zanin, M.,... He, J. (2020). Modified SEIR
and AI prediction of the epidemics trend of COVID-​19 in China under public health inter-
ventions. Journal of Thoracic Disease, 12(3), 165–​174. doi:10.21037/​jtd.2020.02.64
 5
FINANCING AND RESOURCE TRACKING
OF VACCINATION PROGRAMS
Edited by Logan Brenzel and Shreena Malaviya
5.0
Section introduction: financing
and resource tracking of
vaccination programs
Logan Brenzel and Shreena Malaviya

The ability of national immunization programs to introduce and deliver

new vaccines, and reach their target populations, particularly those children
and communities who are hard to reach, is related to the resources available
to provide services. Sustaining adequate financing is a critical function of
immunization programs and is reflected as one of the pillars of the global
Immunization Agenda 2030. This section of the handbook focuses on var-
ious dimensions of immunization financing. Chapter 5.1 examines histor-
ical immunization spending by governments, development partners, and
households/​private sector for the total program, and split between routine
and campaign services, as well as vaccines and delivery. Chapter 5.2 elab-
orates on the current sources of financing, and how these might be better
utilized to improve adequacy and sustainability of the program, particularly
as coronavirus disease 2019 (COVID-​19) has affected economies and health
budgets. Finally, Chapter 5.3 explores the development partner landscape
for immunization, particularly focusing on Gavi, the Vaccine Alliance and
some of the innovative financing instruments that have been developed.
Various resources exist that provide further details on health and immuni-
zation financing, and these are provided for readers.

Logan Brenzel and Shreena Malaviya, Section introduction: financing and resource tracking of vaccination programs In: Handbook
of Applied Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0026
5.1
Introduction to immunization financing
and expenditure
Logan Brenzel and Shreena Malaviya

Immunization offers a first line of protection for households against immi-

nent catastrophic direct and indirect costs in treating vaccine-​preventable
illnesses and infections. Immunization programs have been successful in
reaching children worldwide. In 2018, 86% of children received all three doses
of the diphtheria, pertussis, and tetanus (DPT3) vaccine. The high coverage
of immunization programs forms a critical backbone of the universal health
coverage expansion strategy.
Immunization programs need sustainable financing to remain effective and
efficient. Each year a new cohort of children must be reached at high levels
of coverage which requires reliable resources. Immunization programs must
continue indefinitely to ensure a high coverage of vaccines, unless and until a
disease is eradicated. Well-​planned immunization programs have the poten-
tial to reach remote and geographically constrained settings that are less linked
to the formal health system, offering opportunities to those communities to
access basic and critical healthcare interventions. A drop in coverage could re-
sult in disease outbreaks that can undo all the years of investment. Moreover,
in some countries, access to immunization is considered a basic health right
and a public responsibility to be funded by the government. As a public good,
governments have a unique role in funding vaccination programs. Trends in
immunization expenditure highlight both the large financial requirement for
the program and the challenges that remain to ensure sustainability.
A recent comprehensive review and analysis estimated government, pri-
vate, and donor spending for immunization in 135 low-​and middle-​income
countries (Ikilezi et al., 2021b). Data were also disaggregated between vaccine
and delivery cost spending, as well as between routine immunization delivery
and supplementary campaigns. Immunization spending increased threefold
from $3.2 billion in 2000 to $10.2 billion in 2017, for a total of $112.4 bil-
lion over the period (Fig. 5.1.1). Increases in spending occurred in 90% of

Logan Brenzel and Shreena Malaviya, Introduction to immunization financing and expenditure In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0027
 Introduction to immunization financing and expenditure 333

12 Prepaid private spending

Out-of-pocket spending
Development assistance for health
10 Government spending
Billions of 2019 US$

0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Year

Fig. 5.1.1 Immunization spending by source (2000–​2017).

Source: Ikilezi et al. (2021b).

countries, while eight countries with primarily fragile and conflict settings ex-
perienced declines.
Government spending accounted for the largest share of total spending on
immunization throughout the period, ranging between 60% and 79% of total
spending each year. Development assistance for health represented nearly a
third of all spending at $31.7 billion (28%), with more than $13 billion chan-
neled through Gavi, the Vaccine Alliance. Development assistance for health
represented 49% and 48% of total immunization spending in sub-​Saharan
Africa and Asia, respectively. While immunization services are usually pro-
vided free of charge, the study estimated that $4 billion had been spent out of
pocket over the period. Finally, spending on vaccines nearly tripled over this
period, while actual spending on delivery remained relatively stable and ac-
counted for 44% of total spending in 2017 (Fig. 5.1.2).
Nearly 90% of government spending was allocated to routine immuniza-
tion. More than half of development assistance for health (58%) was allocated
to routine immunization activities. Immunization spending averaged $40 per
surviving infant in lower-​income countries. Spending was highest in Latin
America and the Caribbean at $208 per surviving infant. While higher DPT3
coverage was associated with increased government spending on immuniza-
tion, there were exceptions.
Analysis of JRF data also shows expenditure per surviving infant varies by
region and country income level. Two percent of global health expenditure
is spent on immunization, ranging from 2.8% of current health spending in
334 Introduction to immunization financing and expenditure

Fig. 5.1.2 Spending patterns on immunization (2000–​2017).

Source: Ikilezi et al. (2021b).

low-​income countries to 0.3% of current health spending in high-​income

countries. Higher income countries spent much more on immunization per
surviving infant ($1,000) in 2017 than lower income countries ($29.50) ac-
cording to the World Health Organization (WHO, 2020b).
Variation in expenditure levels between countries is related to some extent
by unit price differences of vaccines in each country. Pooled procurement
mechanisms tend to offer countries lower prices for a majority of vaccines.
Self-​procuring middle-​income countries pay twice as much as countries that
procure through UNICEF or the PAHO Revolving Fund. Longer contracts
are also associated with lower vaccine unit prices, as well as higher volumes
procured (WHO: Immunization Vaccines and Biologicals, 2019). Higher
income countries, particularly those in the Region of the Americas and the
European Region, represent the largest share of the global market value of
vaccines (68%).
Gavi-​eligible countries benefit from lower vaccine prices because of the
larger volumes of doses. The WHO 2019 Global Vaccine Market Report high-​
income countries introduce more vaccines (74% more than low-​income
countries and 23% more than middle-​income countries) in their routine im-
munization programs than lower-​income countries (WHO, 2020b).
Lower income countries that are eligible for support from Gavi, the Vaccine
Alliance may have a higher share of donor support for their immunization
program because of the initial subsidy for each new vaccine introduced.
Gavi support is tiered on the basis of the gross national income (GNI) of the
country. For the lowest income countries, the subsidy level doesn’t change
 5.1.1 Benefits of resource tracking in immunization financing 335

until a country’s GNI crosses a threshold, in which case the subsidy level is
ramped down. Every time a new vaccine is introduced with Gavi support, the
share of donor support would be expected to rise. Gavi’s co-​financing policy
has so far been a successful strategy to build in sustainable financing for im-
munization (Gavi, 2019; Kallenberg et al., 2016).
An increase in government funding would be expected as countries in-
crease their co-​financing responsibilities and transition out of Gavi support.
As countries transition from primarily external aid funded immunization
programs to domestically-​funded, they face a variety of financial and insti-
tutional challenges. Countries experience financial obstacles such as high
total vaccine costs and the need to create additional fiscal space in govern-
ment health budgets within a relatively short period of time. There are also
programmatic and institutional challenges such as limited procurement and
regulatory capacities. Of Gavi-​transitioned countries (16) have experienced
increases in domestic expenditures for vaccines up until 2018 (Ikilezi et al.,
2021b).
Additional expenditures will be required to scale-​up immunization cov-
erage, add in new vaccines, technologies, or delivery strategies, and reach
zero dose children and communities, Estimates suggest it could cost $54 USD
on average to fully immunize a child against ten antigens at high coverage
levels (Sim et al., 2020). An additional US $38.5 billion will be needed in 64
countries for child immunization between 2016 and 2030, representing 30%
more than what these countries are currently spending (Stenberg et al., 2017).
Countries will be challenged to fill the gap with domestic resources, which are
at risk following COVID-​19 health and economic impacts.

5.1.1 Benefits of resource tracking in

immunization financing

The Immunization Agenda 2030—​A Global Strategy to Leave No One

Behind—​encourages countries to consistently track and evaluate expend-
itures to achieve national immunization program goals (WHO, 2020a). The
cornerstone of a robust immunization program is reliable and sustainable
financing. The regular collection and analysis of immunization expenditure
data can provide the necessary data for better planning and budgeting of the
national program, and for advocating for additional resource mobilization.
evidence required to plan and secure sufficient funding for immunization
activities.
336 Introduction to immunization financing and expenditure

Resource tracking captures the sources and uses of resources (e.g. finan-
cial, material, and human) for the health sector and immunization programs.
The types of policy questions that can be addressed with resource tracking
include:

• How much do governments, households, partners spend per year on vac-

cines, routine services, campaigns?
• What percentage of spending for routine services comes from govern-
ment (domestic) sources or external sources?
• What percentage of total and government health expenditure is spent on
routine immunization?
• How have trends in 1-​3 above changed over time and how do they com-
pare to the aspirations of the country on sustainable financing?

5.1.2 Expenditure tracking for program planning

and budgeting

Analyzing a country’s immunization expenditure estimates over the years and

comparing it to other countries, allows Expanded Program on Immunization
managers and policymakers to gain insight into a number of critical issues
(Abt Associates, 2018).

• Sustainability of immunization financing: Data on immunization ex-

penditure by various sources—​ internal and external—​can provide
further insights into the government’s contribution to immunization fi-
nancing compared to that of donors. It is important to understand pat-
terns of funding for different sources and how these relate to goals of
self-​sufficiency and transition from donor support.
• Equitable access to immunization services: Out-​of-​pocket immunization
expenditure can be evaluated alongside income levels of households to
understand the potential financial burden and impact on the access to
immunization services (Abt Associates, 2018).
• Accountability of decision makers on spending: Actual spending
on immunization from Health Accounts exercises can be compared
to resource requirements as outlined in national plans, such as the
Comprehensive Multi-​Year Plan (cMYP), to understand any potential
funding gaps (Abt Associates, 2018). These data, when accurately cap-
tured, can also be used to monitor the country’s progress as well as to
compare budgets against spending (budget execution rate).
 5.1.3 Types and sources of data 337

• Efficiency of immunization spending: Comparing immunization

spending per live birth between countries can provide insight into how
their immunization spending and coverage compares to other programs
(Abt Associates, 2018).
• Impact: comparing total spending for immunization with coverage
levels and other measures of program output or outcomes can provide a
measure of the performance of the program.

5.1.3 Types and sources of data

The main sources of cross-​country and relatively comparable information

on immunization expenditures are the Joint Reporting Form (JRF) and the
National Health Accounts (NHA). The JRF was developed by the WHO and
UNICEF and tracks self-​ reported spending information from countries
on both government and total spending through seven indicators (WHO,
2020b). These expenditures pertain to immunization-​specific, noncapital
expenditures of the program. Hence, these data exclude spending on shared
resources or large investments like cold chain. JRF expenditure data are re-
ported and evaluated by WHO annually since 2010. Data can be obtained
through a request to WHO.
The NHA is a robust expenditure tracking methodology that allows
countries to evaluate and monitor total health spending across different
disease classifications funded from various sources. This approach yields
consistent and comprehensive data on health spending in a country.
Within this system, countries can track changes in policy priorities and
understand if the introduction of reforms and new programs resulted in
changes in resource allocation and expenditure. In this methodology, im-
munization spending relates to the function of preventive services (Abt
Associates, 2018; WHO, 2017). Data are posted in the public domain in
the Global Health Expenditure Database (GHED) which currently con-
tains estimates on immunization spending for nearly 60 countries, with
half of those estimates from countries in sub-​Saharan Africa (WHO,
2017). Higher-​income countries may report data into the OECD Health
Expenditure Database (Organization of Economic Cooperation and
Development, 2021).
Gavi and UNICEF regularly monitor country-​ level procurement and
spending on new and underused vaccines as part of co-​financing of new vac-
cines. These data are updated at least twice per year and exclude data on vac-
cines administered during supplementary activities. These reports are used to
338 Introduction to immunization financing and expenditure

assess the trajectories of Gavi countries in increasing the ownership of vaccine

financing and sustainability.
The cMYP is a comprehensive planning framework developed by
WHO and UNICEF for integrating immunization program strategies
and identifying the resources required to implement and achieve them
(WHO: Immunization Vaccines and Biologicals, 2014). An associated
Excel-​based tool is used to estimate the resource requirements and financing
needed for a 5-​year period, along with a funding gap. While the cMYP is a
planning tool and not a costing study per se, estimates of the resources used
in the first or year (baseline) has been used as a proxy for immunization pro-
gram costs. WHO and UNICEF have developed a new strategic framework
for national immunization planning (National Immunization Strategy)
which builds off of the country experience with the cMYP.
Comparisons between various sources of immunization spending re-
veal data quality issues and inconsistencies though some improvements are
occurring over time (Ikilezi et al., 2021b). Vaccine spending by governments
is reported more completely than routine immunization spending. This may
be due in part to the complexity of estimating immunization-​specific rou-
tine program spending. Some consistency is observed in the various sources
on vaccine spending. Based on cMYP data, routine immunization spending
was nearly twice as much as that reported through the JRF, and almost four
times higher than that reported through the NHA on average Both the JRF
and cMYP reported higher spending than was reported through Gavi co-​
financing (Ikilezi et al., 2021a).
Thus, it is important to consider these factors when utilizing the data for
immunization financing development and planning. All stakeholders should
be encouraged to collaborate and work with the government to ensure better
reporting of vaccine and immunization spending.

5.1.4 Conclusion

Immunization is a highly cost-​effective intervention whose effectiveness and

efficiency benefits from robust and sustainable financing. Resource tracking
on a frequent basis is a powerful tool for understanding the sources and uses of
funding, as well as to benchmark country immunization spending relative to
neighboring countries achieving similar levels of performance. Expenditure
data can be sourced from a variety of sources, and factors such as the com-
pleteness, reliability, and validity of expenditure data need to be considered
when interpreting trends.
 5.1.4 Conclusion 339

References
Abt Associates. (2018). Understanding and using immunization expenditure data:
A primer for immunization program managers. https://​www.abtass​ocia​tes.com/​insig​hts/​
publi​cati​ons/​rep​ort/​unders​tand​ing-​and-​using-​immun​izat​ion-​expe​ndit​ure-​data-​pri​mer-​for
Gavi. (2019). Annual program report. https://​www.gavi.org/​progr​ess-​rep​ort
Ikilezi, G., Bachmeier, S. D., Cogswell, I. E., Maddison, E. R., Stutzman, H. N., Tsakalos,
G., . . . Micah, A. E. (2021a). Tracking government spending on immunization: The joint re-
porting forms, national health accounts, comprehensive multi-​year plans and co-​financing
data. Vaccine, 39(25), 3410–​3418. doi:10.1016/​j.vaccine.2021.04.047
Ikilezi, G., Micah, A. E., Bachmeier, S. D., Cogswell, I. E., Maddison, E. R., Stutzman, H.
N., . . . Dieleman, J. L. (2021b). Estimating total spending by source of funding on rou-
tine and supplementary immunisation activities in low-​and middle-​income countries,
2000–​2017: A financial modelling study. Lancet Global Health, 398(10314), 1875–​1893.
doi:10.1016/​S0140-​6736(21)01591-​9
Kallenberg, J., Mok, W., Newman, R., Nguyen, A., Ryckman, T., Saxenian, H., & Wilson, P.
(2016). Gavi’s transition policy: Moving from development assistance to domestic financing
of immunization programs. Health Affairs, 35(2), 250–​258. doi:10.1377/​hlthaff.2015.1079
Organization of Economic Cooperation and Development. (2021). OECD health expenditure
database. https://​data.oecd.org/​health​res/​hea​lth-​spend​ing.htm
Sim, S. Y., Watts, E., Constenla, D., Brenzel, L., & Patenaude, B. N. (2020). Return on investment
from immunization against 10 pathogens in 94 low-​and middle-​income countries, 2011–​30.
Health Affairs, 39(8), 1343–​1353.
Stenberg, K., Hanssen, O., Edejer, T. T., Bertram, M., Brindley, C., Meshreky, A., . . . Soucat,
A. (2017). Financing transformative health systems towards achievement of the health
Sustainable Development Goals: A model for projected resource needs in 67 low-​income
and middle-​ income countries. Lancet Global Health, 5(9), e875–​ e887. doi:10.1016/​
s2214-​109x(17)30263-​2
World Health Organization. (2017). Expenditure on prevention activities under SHA
2011—​supplementary guidance. https://​apps.who.int/​nha/​datab​ase/​Docu​ment​atio​nCen​
tre/​Index/​en
World Health Organization. (2020a). Immunization Agenda 2030: A global strategy to leave no
one behind. https://​www.who.int/​publi​cati​ons/​m/​item/​immun​izat​ion-​age​nda-​2030-​a-​glo​
bal-​strat​egy-​to-​leave-​no-​one-​beh​ind
World Health Organization. (2020b). Situation analysis of immunization expenditure. https://​
www.who.int/​docs/​defa​ult-​sou​rce/​immun​izat​ion/​financ​ing/​situat​ion-​analy​sis-​key-​facts.
pdf?sfv​rsn=​8c6​5d92​2_​2
World Health Organization: Immunization Vaccines and Biologicals. (2014). Comprehensive
multi-​ year plan costing tool. https://​apps.who.int/​iris/​bitstr​eam/​han​dle/​10665/​128​051/​
WHO​_I​ VB​_1​ 4.06_​eng.pdf
World Health Organization: Immunization Vaccines and Biologicals. (2019). MI4A: Market
information for access to vaccines. https://​www.who.int/​teams/​immun​izat​ion-​vacci​nes-​and-​
biol​ogic​als/​vacc​ine-​acc​ess/​mi4a
5.2
Financing of immunization
programs
Logan Brenzel

With a return on investment of $21 for every $1 invested, vaccination is a

good investment in health (Ozawa et al., 2016). Because of the public goods
nature of vaccination, governments have a significant role to play in financing
services. However, many countries struggle to fully fund their programs
to be able to reach 100% coverage. Global immunization coverage for the
third dose of diphtheria, tetanus, and pertussis vaccine has hovered at 85%
for several years and countries face challenges to fully reach their population
with life-​saving vaccines. More than 22 million children a year miss out on
life-​saving vaccines (World Health Organization, 2020b). The vision of the
Immunization Agenda 2030—​“a world where everyone, everywhere, at every
age, fully benefits from vaccines for good health and well-​being” (World
Health Organization, 2020b)—​requires adequate and sustainable financing
for immunization.

5.2.1 Sources of financing for immunization

This chapter provides an overview of how countries are currently financing

their national immunization programs. Prospects for sustainable financing
cannot be disconnected from the reliability and adequacy of various sources
of financing. A key message is that in fiscally constrained settings, resources
available for the immunization program are a function of funding available
for the entire health sector and how those resources are prioritized and al-
located. Second, the mechanisms of health system financing can influence
the availability and predictability of financing for immunization. The three
main sources of immunization financing are (1) government, (2) private, and
(3) external. The mechanisms for each are discussed in turn below. Several
toolkits have been developed which describe country experiences with

Logan Brenzel, Financing of immunization programs In: Handbook of Applied Health Economics in Vaccines. Edited by:
David Bishai, Logan Brenzel and William V. Padula, Oxford University Press. © Oxford University Press 2023.
DOI: 10.1093/oso/9780192896087.003.0028
 5.2.1 Sources of financing for immunization 341

various sources of financing which will provide further details (Results for
Development, 2017; World Bank & Gavi, 2010).

5.2.1.1 Government (domestic) financing for

immunization

On average, $8.5 trillion, equivalent to 10% of global gross domestic product

was spent on healthcare in 2019. Health spending in lower-​income countries
primarily came from out-​of-​pocket and external sources. Primary healthcare
spending accounted for almost half of total spending, ranging from $12 in DR
Congo to $3,800 in Switzerland, highlighting the huge disparities in health-
care spending around the world. Cross-​country inequality is rising over time.
Government health financing accounted for less than 2% of gross domestic
product in 2019 in lower-​income countries (World Health Organization,
2020a, 2021).
While government allocations to health have generally grown over time,
overall economic growth, rather than reallocation between sectors has driven
this growth (Barroy & Gupta, 2020; Tandon, Cain, Kurowski, & Postolovska,
2018). The dependence upon economic growth means that government
health sector financing is vulnerable to economic shocks, such as during the
coronavirus 2019 (COVID-​19) pandemic.

5.2.1.1.1 National level financing for immunization

Immunization should be a government financial responsibility as benefits ex-
tend beyond the individuals vaccinated through herd immunity. Ensuring
reliable, predictable funding to immunize all newborn children every year
with the vaccines in the national schedule and with potentially new vaccines
to be introduced requires considerable resources. Health services in most
countries are a financed through tax and revenue collection from individuals,
corporations, and other sources. These are pooled and allocated across sec-
tors, including health, as part of the annual budget cycle and planning pro-
cess. Government health spending compared to total government spending
reflects the priority of health in the country. This estimate ranges from 5.6%
in low-​income countries, to 14.3% in high-​income countries (World Health
Organization, 2020).
Government health financing at the national level can be allocated directly
from the national Treasury to the Ministry of Health which will allocate funds
across preventive and promotive health programs and services, as well as to
342 Financing of immunization programs

specific service providers, such as hospitals and clinics. Health funding allo-
cated to the Ministry of Health may also be transferred to subnational units or
programs. The Treasury may also direct health funding to subnational units
through block grants, or to facilities through direct facility financing. Direct
facility financing is an approach that attempts to solve the delays and bottle-
necks in resources reaching frontline facilities.
The share of financing allocated for immunization from government
budgets depends upon competing priorities for those resources. Often this
share is based on historical budget levels rather than evidence of the cost of
providing those services. Incorporating the actual costs of vaccine delivery
into annual and multi-​year health budget processes can improve the align-
ment of budgets to actual resource needs, and ensure greater predictability
of allocations (Lydon et al., 2008). Having a line item for vaccines and immu-
nization may assist in creating greater visibility for the recurrent needs of the
program. A recent evaluation of African government budgets for immuniza-
tion revealed that countries with input-​based budgeting had several line items
for vaccine purchase, vaccine co-​financing, campaigns, and other immuniza-
tion program inputs, with an average of nine line items, ranging from 0 to 42
(Griffiths et al., 2020).
Greater advocacy on the return on investment to immunization and the
role of vaccination in reducing poverty and generating economic benefits
may aide program manager and ministry staff in effectively garnering addi-
tional funding. Recent evidence suggests that vaccination can protect house-
holds in the poorest countries from slipping into poverty, due to the high costs
of treating measles, pneumonia, and other illnesses (Chang et al., 2018). In
Bangladesh, the average out-​of-​pocket cost to households for treating a case
of childhood measles was $48 in public facilities and over $80 in private facil-
ities (de Broucker, Ahmed, et al., 2020). In Uganda, caregivers were respon-
sible for $23 in payments for measles treatment and paid the equivalent of
30% of the household’s monthly income to treat it (De Broucker, Ssebagereka,
et al., 2020).
Governments may allocate resources to immunization programs through
earmarks which are guaranteed allocations. Earmarking, or setting aside ei-
ther a specific amount or a proportion of health funding each budget cycle,
for immunization may be mandated in public law. Nine countries in Asia,
Africa, and Latin America have tried to institute or have such laws. Bolivia fi-
nances its program through allocation of a certain percentage of government
health funding. While there are obvious short-​term benefits, earmarking
introduces rigidities in the government budget which may, in the long run,
 5.2.1 Sources of financing for immunization 343

be detrimental to services. For instance, earmarks may cover the cost of pro-
curing vaccines, but support for delivery and operating the program may
be cut to accommodate increasing costs of vaccines. In economic shocks, if
health budgets are cut, the proportion going to the program may also falter
(Results for Development, 2017).

5.2.1.1.2 Subnational level financing

Government funding for healthcare is generated at subnational levels through
local tax and revenue collection. Funds generated at the local level and those
that are transferred from national sources through inter-​fiscal transfers form
the basis for government subnational financing. Allocation and use of lo-
cally generated and managed healthcare resources are critical to cover the
operating costs of immunization programs. This requires working with local
mayors and decision makers over how best to support service delivery. This
source of financing is often overlooked in relation to mobilizing additional re-
sources for the immunization program.
In a decentralized context, roles, responsibilities, and authority for im-
munization program functions are spread between the national and
subnational level. Local government health officers may not report directly
to the Ministry of Health, but to ministries of local government or other
agencies. The national government is usually responsible for policy guid-
ance, regulation, and vaccine procurement, while decentralized units must
cover the cost of delivery, often based upon national recommendations and
standards. In Indonesia, districts are fully responsible for implementing
the immunization program and must finance key components. In Pakistan,
the central level procures vaccines, but financing is supported by the pro-
vincial level as well (Learning Network for Countries in Transition, 2021).
Immunization programs at subnational level face challenges with respect
to coordination, budgeting, and management capacity. Managing and en-
suring adequate subnational funding flows to facilities and frontline pro-
viders for immunization services is critical.

5.2.1.1.3 National health insurance

National government health financing may also be allocated to social insur-
ance mechanisms that aim to cover the healthcare needs of the population.
Many countries in Europe, Asia, the UK, and Canada provide health serv-
ices to their population through a single-​payer system (i.e., the government).
Several lower-​income countries also committed to national health insurance
(NHI) as a mechanism for supporting universal health coverage such as such
344 Financing of immunization programs

Medical costs: what proportion

of the cost is covered?

Reduce cost sharing

and fees to users

Include additional
healthcare services
Extend to people who
are not covered Current pooled funds
Healthcare services: which
services are covered?

Population: who is covered?

Fig. 5.2.1 Services and coverage for universal healthcare.

Source: Adapted from Maeda et al. (2014).

as Ghana (Grépin & Dionne, 2013), Kenya (Union for International Cancer
Control, 2020), and Côte d’Ivoire (Dagnan, 2018).
Universal health coverage is based on three components as represented
in Fig. 5.2.1): (1) service coverage (expanding scope and quality of essential
services), (2) population coverage (expanding population groups that benefit
from essential packages of services, and (3) financial risk protection (Maeda
et al., 2014).
Making positive gains in these three domains may require countries to un-
dertake significant reforms and policy choices related to the financing and
delivery of health services. The additional cost of achieving universal health
coverage and the Sustainable Development Goals targets by 2030 have been
estimated to reach $371 billion annually or $58 per person (Stenberg et al.,
2017). Prioritizing preventive health services can result in potential health
systems savings from reduced utilization of expensive curative services and
hospital care. NHI relies heavily on public/​government financing and is
a way of pooling different health risks and providing a benefits package of
quality health services. A “health benefits package” is essentially a selection
of services to be delivered to the covered population (Glassman, Giedion, &
Smith, 2017).
A recent review found that maternal and child health services were the
services most often included in the benefits packages of 26 countries. Nearly
half of those specifically included immunization services, and 20% of them
partially reflected immunization in their benefits packages (Regan, Wilson,
Chalkidou, & Chi, 2021).
 5.2.1 Sources of financing for immunization 345

Fig. 5.2.2 Funding of vaccines and immunization through NHI schemes. MOH, Ministry
of Health.
Source: Adapted from Learning Network for Countries in Transition (2020).

While including immunization in the benefits packages for NHI may

bring benefits in terms of coverage, there may be some unanticipated impacts
on the national immunization program that should be considered. NHI does
not necessarily translate into additional resources for the health sector or for
the immunization program, as reductions in allocations to other Ministry of
Health programs and services may occur in order to finance NHI. NHI can
coexist alongside Ministry of Health funding which may create fragmented
funding flows to the program. NHIs may focus more on curative services,
potentially crowding out the financing and provision of preventive services
such as immunization. If NHI population coverage is limited, the immuniza-
tion program may not be able to vaccinate newborns and infants at the same
level of coverage. The cost of introducing new vaccines is higher than those
currently in the program, and this may pose greater challenges for financing
and sustaining NHI (Learning Network for Countries in Transition, 2020).
Fig. 5.2.2 shows the range of country experiences with financing vaccines
through NHI and other sources.

5.2.1.1.4 Health taxes

Taxes on alcohol, tobacco, and sugar have been a useful mechanism for raising
additional revenues for public spending. Evidence suggests that these types of
taxes also help to reduce the consumption of unhealthy commodities (Wright,
Smith, & Hellowell, 2017). Health taxes have also been shown to raise a sig-
nificant amount of revenue for additional resources for the health sector.
A 50% increase in the prices of sugar-​sweetened drinks, alcohol, and tobacco
346 Financing of immunization programs

can achieve $24.7 billion in 54 low-​and middle-​income countries (LMICs)

by 2030 (Marquez & Dutta, 2020). There are a few instances of health taxes
being used for immunization, including the Philippines. Ghana’s NHI scheme
is funded through a value-​added tax, which also supports the country’s co-​
financing requirements for Gavi (Results for Development, 2017). Health
taxes may be one way of expanding budgetary space for health services in the
future, given the impacts of COVID-​19 vaccines on government budgets.

5.2.1.2 Private financing

The private sector may play a range of roles in the national immunization
program by either extending service coverage in low-​income countries or
aiding the introduction of new vaccines in middle-​income countries (Levin
& Kaddar, 2011). Public and private sector collaboration is essential, aided by
regulations and oversight to ensure quality vaccination (Ahmed, DeRoeck, &
Sadr-​Azodi, 2019; Mitrovich, Marti, Watkins, & Duclos, 2017).
Most LMICs have official policies to provide vaccination free of charge,
though households may incur some costs related to seeking care. Country-​
level studies have shown a wide variation in both the volume of private sector
immunization services as well as the cost to patients paying out of pocket for
those services. For instance, in Benin, 64% of clients in facilities reported
paying for vaccination, down to 14% in Malawi. Clients reported paying for
vaccination cards, vaccination services, and syringes, among other services.
The average out-​of-​pocket cost per vaccination was $0.50 in Malawi and up to
$14 in Georgia (Levin et al., 2019).
There are very few health systems that are primarily financed from private
sources and where vaccination services are financed 100% out of pocket.
In mixed systems with public and private financing for health, private in-
surance may be a source of financing for vaccination and immunization
services. This tends to be the case for higher-​income countries. Private, vol-
untary insurance and employer-​based insurance may cover most of the cost
of vaccination, but households are responsible for monthly premiums, de-
ductibles, and co-​pays at the time of service. In the US, the Affordable Care
Act mandated that priority preventive vaccinations must be covered by pri-
vate insurance.
Trust funds may be financed through private donations and government
allocations which are invested and managed to fund a share of the cost of run-
ning health and immunization programs. For instance, Bhutan has established
 5.2.2 Improving budgetary space for immunization 347

a trust fund for health services which covers immunizations. This is a long-​
running trust fund with significant political commitment and support. Other
countries have examined setting up trust funds, but the complexity and cost
of operating and maintaining the fund may outweigh the additional resources
generated (Results for Development, 2017).
Lotteries generated through out-​of-​pocket purchases of tickets may be an-
other source of financing for immunization programs. For instance, the gov-
ernment of Costa Rica dedicates net funds from the November draw of the
national lottery which covers approximately 1% of vaccine purchasing re-
quirements (Results for Development, 2017; World Bank & Gavi, 2010).

5.2.1.3 External financing

External financing for immunization has been used for both procurement of
vaccines as well as support for the delivery of immunization services in many
LMICs. In general, the share of external financing is inversely related to the
country’s gross domestic product. Sustainable financing to meet immuniza-
tion program needs is critical, particularly as countries transition away from
donor support. Chapter 5.3 examines in greater depth the various external
sources for immunization financing, which include subsidies for new vac-
cines and systems strengthening from Gavi, the Vaccine Alliance; bilateral de-
velopment assistance; and foundation support.

5.2.2 Improving budgetary space for immunization

The level of funding available for the health sector can be increased through
various mechanisms. Fiscal space is related to economic growth; allo-
cating more resources to primary health care and immunization services;
earmarking; utilizing immunization resources more effectively and effi-
ciently; and funding through donor support (Tandon et al., 2018). However,
expanding fiscal space is not the only approach for garnering more resources
for health, and for primary health care and immunization programs by ex-
tension. Improving budgetary space through better public financial man-
agement (PFM) will likely lead to additional health sector outcomes. A study
examining the relationship between PFM and child mortality found that
a 1% improvement in PFM translated into a 14% reduction in child mor-
tality (Piatti-​Fünfkirchen & Smets, 2019). Four PFM interventions enhance
348 Financing of immunization programs

budgetary space for health: (1) reducing unnecessary expenditures through

more flexibility in budget structures, (2) a results-​ based approach to
budgeting, (3) improving budget execution, and (4) influencing future alloca-
tions by good budget performance (Barroy & Gupta, 2020).
Improving budgetary space for health will have positive benefits for pri-
mary health care and immunization programs. There are specific actions that
can be taken by programs themselves to further expand budgetary space.
First, budgeting and planning should be related to achieving results both in
the medium term and annually. The World Health Organization has recently
launched the National Immunization Strategy (NIS) which aims to bring a
greater strategic lens to planning and budgeting for the Immunization Agenda
2030 (World Health Organization: Immunization Vaccines and Biologicals,
2021). Second, immunization budgets need to be fully utilized and executed
within the fiscal year. There are often challenges in full utilization of resources
given delays in receipt of disbursements, particularly at the subnational level.
Third, greater efficiency in the use of immunization resources, through lower-​
cost procurement options, better planning and organization of campaigns,
reducing wastage, and using resources in more cost-​effective ways, among
others. These approaches can help to achieve greater performance and cov-
erage of the program.

5.2.3 Conclusion

This chapter has explored the range of financing sources available for ensuring
adequate and sustainable financing for immunization programs. Because of the
public goods nature of vaccination, governments have a unique role in financing
these programs, complemented by other sources. Providing vaccinations free
of charge or at highly subsidized costs can ensure that no one is deterred from
seeking immunization services if they want them. Government commitment is
key to mobilizing resources. However, countries, particularly LMICs, face con-
straints and competing priorities for government spending. Greater attention to
better financial management and use of program resources may provide addi-
tional budgetary space, particularly during economic crises.

References
Ahmed, N., DeRoeck, D., & Sadr-​Azodi, N. (2019). Private sector engagement and contribu-
tions to immunisation service delivery and coverage in Sudan. BMJ Global Health, 4(2),
e001414. doi:10.1136/​bmjgh-​2019-​001414
 5.2.3 Conclusion 349

Barroy, H., & Gupta, S. (2020). From overall fiscal space to budgetary space for health: Connecting
public financial management to resource mobilization in the era of COVID-​19. Center for
Global Development. https://​www.cgdev.org/​sites/​defa​ult/​files/​over​all-​fis​cal-​space-​budget​
ary-​space-​hea​lth-​con​nect​ing-​pub​lic-​financ​ial-​man​agem​ent.pdf
Chang, A. Y., Riumallo-​Herl, C., Perales, N. A., Clark, S., Clark, A., Constenla, D., . . . Verguet,
S. (2018). The equity impact vaccines may have on averting deaths and medical impover-
ishment in developing countries. Health Affairs (Millwood), 37(2), 316–​324. doi:10.1377/​
hlthaff.2017.0861
Dagnan, S. (2018). Health system reforms to accelerate universal health coverage in Côte
d’Ivoire. Health Systems & Reform, 4(2), 69–​71. doi:10.1080/​23288604.2018.1446123
de Broucker, G., Ahmed, S., Hasan, M. Z., Mehdi, G. G., Martin Del Campo, J., Ali, M.
W., . . . Patenaude, B. (2020). The economic burden of measles in children under five in
Bangladesh. BMC Health Services Research, 20(1), 1026. doi:10.1186/​s12913-​020-​05880-​5
De Broucker, G., Ssebagereka, A., Apolot, R. R., Aloysius, M., Ekirapa Kiracho, E., Patenaude,
B., & Constenla, D. (2020). The economic burden of measles in children under five in
Uganda. Vaccine: X, 6, 100077. doi:https://​doi.org/​10.1016/​j.jvacx.2020.100​077
Glassman, A., Giedion, U., & Smith, P. C. (2017). What’s in, what’s out? Designing benefits for
universal health coverage. Center for Global Development. https://​www.cgdev.org/​sites/​defa​
ult/​files/​whats-​in-​whats-​out-​design​ing-​benef​i ts-​final.pdf
Grépin, K. A., & Dionne, K. Y. (2013). Democratization and universal health coverage: A case
comparison of Ghana, Kenya, and Senegal. Global Health Governance, VI(2). http://​blogs.
shu.edu/​ghg/​files/​2014/​02/​GHGJ-​Vol​ume-​V1-​No.-​2-​Sum​mer-​2013.pdf#page=​5
Griffiths, U. K., Asman, J., Adjagba, A., Yo, M., Oguta, J. O., & Cho, C. (2020). Budget line items
for immunization in 33 African countries. Health Policy and Planning, 35(7), 753–​764.
doi:10.1093/​heapol/​czaa040
Learning Network for Countries in Transition. (2020). LNCT brief: Considerations for man-
aging immunization programs within national health insurance. https://​lnct.glo​bal/​wp-​
cont​ent/​uplo​ads/​2020/​02/​Con​side​rati​ons-​for-​Immun​izat​ion-​Progr​ams-​wit​hin-​NHI_​FI​
NAL.pdf
Learning Network for Countries in Transition. (2021). Mobilizing resources in decentralized
health systems: A desk review of LNCT country experiences. https://​lnct.glo​bal/​wp-​cont​ent/​
uplo​ads/​2021/​03/​Decen​tral​izat​ion-​resou​rce-​doc-​format​ted-​FINAL.pdf
Levin, A., & Kaddar, M. (2011). Role of the private sector in the provision of immunization
services in low-​and middle-​income countries. Health Policy and Planning, 26(Suppl 1), i4–​
i12. doi:10.1093/​heapol/​czr037
Levin, A., Munthali, S., Vodungbo, V., Rukhadze, N., Maitra, K., Ashagari, T., & Brenzel, L.
(2019). Scope and magnitude of private sector financing and provision of immunization in
Benin, Malawi and Georgia. Vaccine, 37(27), 3568–​3575. doi:https://​doi.org/​10.1016/​j.vacc​
ine.2019.05.023
Lydon, P., Levine, R., Makinen, M., Brenzel, L., Mitchell, V., Milstien, J. B., . . . Landry, S.
(2008). Introducing new vaccines in the poorest countries: What did we learn from the
GAVI experience with financial sustainability? Vaccine, 26(51), 6706–​6716. doi:10.1016/​
j.vaccine.2008.10.015
Maeda, A., Araujo, E., Cashin, C., Harris, J., Ikegami, N., & Reich, M. R. (2014). Universal health
coverage for inclusive and sustainable development: A synthesis of 11 country case studies.
Washington, DC: World Bank.
Marquez, P. V., & Dutta, S. (2020). Taxes on tobacco, alcohol, and sugar-​sweetened beverages re-
duce health risks and expand fiscal space for universal health coverage post-​COVID 19. World
Bank. https://​blogs.worldb​ank.org/​hea​lth/​taxes-​toba​cco-​alco​hol-​and-​sugar-​sweete​ned-​
bevera​ges-​red​uce-​hea​lth-​risks-​and-​exp​and-​fis​cal
Mitrovich, R., Marti, M., Watkins, M., & Duclos, P. (2017). A review of the private sector’s con-
tribution to immunization service delivery in low, middle, and high-​income countries. World
350 Financing of immunization programs

Health Organization. https://​www.who.int/​immun​izat​ion/​sage/​meeti​ngs/​2017/​april/​2_​

Review_​private_​sector​_e​ ng​agem​ent_​Mitr​ovic​h_​et​_​al.pdf?ua=​1
Ozawa, S., Clark, S., Portnoy, A., Grewal, S., Brenzel, L., & Walker, D. G. (2016). Return on
investment from childhood immunization in low-​and middle-​income countries, 2011–​20.
Health Affairs, 35(2), 199–​207. doi:10.1377/​hlthaff.2015.1086
Piatti-​Fünfkirchen, M., & Smets, L. (2019). Public financial management, health financing and
under-​five mortality: A comparative empirical analysis. Washington, DC: Inter-​American
Development Bank.
Regan, L., Wilson, D., Chalkidou, K., & Chi, Y. L. (2021). The journey to UHC: How well are
vertical programmes integrated in the health benefits package? A scoping review. BMJ
Global Health, 6(8), e005842. doi:10.1136/​bmjgh-​2021-​005842
Results for Development. (2017). Immunization financing: A resource guide for advocates, pol-
icymakers, and program managers. https://​r4d.org/​resour​ces/​immun​izat​ion-​financ​ing-​
resou​rce-​guide-​advoca​tes-​polic​ymak​ers-​prog​ram-​manag​ers/​
Stenberg, K., Hanssen, O., Edejer, T. T., Bertram, M., Brindley, C., Meshreky, A., . . . Soucat,
A. (2017). Financing transformative health systems towards achievement of the health
Sustainable Development Goals: a model for projected resource needs in 67 low-​income
and middle-​ income countries. Lancet Global Health, 5(9), e875–​ e887. doi:10.1016/​
s2214-​109x(17)30263-​2
Tandon, A., Cain, J., Kurowski, C., & Postolovska, I. (2018). Intertemporal dynamics of public fi-
nancing for universal health coverage: Accounting for fiscal space across countries. World Bank.
https://​doc​umen​ts1.worldb​ank.org/​cura​ted/​en/​639​5415​4528​1356​938/​pdf/​133​115-​19-​12-​
2018-​14-​44-​10-​Accounting​forF​isca​lSpa​cein​Heal​thFI​NAL.pdf
Union for International Cancer Control. (2020). Kicking off UHC in Kenya. https://​www.uicc.
org/​case-​stud​ies/​kick​ing-​uhc-​kenya
World Bank, & Gavi. (2010). Immunization finance toolkit: A resource for policy-​makers and
program managers. http://​hdl.han​dle.net/​10986/​20784
World Health Organization. (2020a). Global spending on health: Weathering the storm. https://​
www.who.int/​publi​cati​ons/​i/​item/​978924​0017​788
World Health Organization. (2020b). Immunization Agenda 2030: A global strategy to leave no
one behind. https://​www.who.int/​publi​cati​ons/​m/​item/​immun​izat​ion-​age​nda-​2030-​a-​glo​
bal-​strat​egy-​to-​leave-​no-​one-​beh​ind
World Health Organization. (2021). Global expenditure on health: Public spending on the rise?
https://​www.who.int/​publi​cati​ons/​i/​item/​978924​0041​219
World Health Organization: Immunization Vaccines and Biologicals. (2021). National immu-
nization strategy (NIS). https://​www.who.int/​teams/​immun​izat​ion-​vacci​nes-​and-​biol​ogic​
als/​vacc​ine-​acc​ess/​plann​ing-​and-​financ​ing/​nis-​(natio​nal-​immun​izat​ion-​strat​egy)
Wright, A., Smith, K. E., & Hellowell, M. (2017). Policy lessons from health taxes: A systematic
review of empirical studies. BMC Public Health, 17(1), 583. doi:10.1186/​s12889-​017-​4497-​z
5.3
Donor architecture
for immunization financing
Grace Chee, George Pariyo, and Shreena Malaviya

5.3.1 The role of Gavi and other institutions in

immunization financing

5.3.1.1 Origins and structure of Gavi, the Vaccine Alliance

Building on the successful eradication of smallpox, the Expanded Program

on Immunization was launched by the World Health Organization (WHO)
in 1974 targeting six diseases—​diphtheria, tetanus, whooping cough, polio,
measles, and tuberculosis. The program achieved a significant increase in im-
munization rates up to 60% in the lowest income countries (Zerhouni, 2019).
In 1984, a collaboration with UNICEF initiated the Universal Childhood
Immunization by 1990 initiative to help countries reach 80% immunization
coverage. In 1990, UNICEF declared that the target of the initiative had been
achieved (Hardon & Blume, 2005).
During the 1990s, as donor financing waned, progress in immunization rates
stagnated and declined in some countries (Clemens, Holmgren, Kaufmann,
& Mantovani, 2010; Muraskin, 2005). To counter this, the Children’s Vaccine
Initiative was launched in 1990, led by UNICEF, the United Nations (UN)
Development Programme, WHO, the World Bank, and the Rockefeller
Foundation. The Children’s Vaccine Initiative’s aims were to develop new vac-
cines to improve performance of the Expanded Program on Immunization,
such as combination and heat-​stable vaccines, as well as to ensure an adequate
supply of vaccines in lower-​income countries. Despite some initial success,
the Children’s Vaccine Initiative was unable to mobilize enough resources to
fulfill its goals (Clemens et al., 2010; Muraskin, 2005; Zerhouni, 2019). By the
late 1990s, it had become clear that despite these efforts, there were still many
childhood deaths from vaccine-​preventable diseases. For instance, measles
outbreaks were still commonplace even though a highly effective measles

Grace Chee, George Pariyo, and Shreena Malaviya, Donor architecture for immunization financing In: Handbook of Applied
Health Economics in Vaccines. Edited by: David Bishai, Logan Brenzel and William V. Padula, Oxford University Press.
© Oxford University Press 2023. DOI: 10.1093/oso/9780192896087.003.0029
352 Donor architecture for immunization financing

vaccine had been around for decades (Zerhouni, 2019). Additionally, vaccines
that were routinely used in high-​income countries, most notably hepatitis B,
were not widely provided in low-​income countries.
The Global Alliance for Vaccines and Immunization (GAVI) was formed
on January 31, 2000, at the World Economic Forum in Davos. GAVI’s focus
was to improve the disparities in access to life-​saving vaccines between
wealthy and lower-​income countries. GAVI aimed to increase the availability
and affordability of new and underused vaccines, building on a partnership
between countries, donors, the private sector, and multilateral and civil so-
ciety organizations (Zerhouni, 2019). GAVI’s founding partners include
UNICEF, WHO, the World Bank, and the Bill & Melinda Gates Foundation.
The first phase of GAVI (2000–​2005) focused on introducing new and
underused vaccines in the poorest countries. GAVI’s second phase (2006–​
2010) included a new focus on strengthening health systems to sustainably
deliver immunization and other services, accelerating the uptake of new and
underused vaccines, and promoting sustainable financing for immunization.
With the progress made previously, GAVI’s third strategy (2011–​2015) con-
tinued to focus on increasing the use of new vaccines and strengthening of
health systems, and added a focus on market shaping to increase the supply
of high-​quality and affordable vaccines. GAVI 4.0 (2016–​2020), built on the
success of previous periods, with renewed focus on sustainability. Over this
period, GAVI’s name changed to Gavi, the Vaccine Alliance (Gavi). Gavi’s
current Phase (2021–​2025), Gavi 5.0, has a new orientation toward reducing
the number of zero-​dose and under-​immunized children.
Gavi is governed by a Board responsible for strategic direction and policies.
The Board is comprised of recipient country governments, donor agencies,
vaccine industry and pharmaceutical company representatives, civil society,
research and health institutes, and independent individuals (Zerhouni, 2019).
The four founding partners hold permanent seats on Gavi’s Board while other
partners rotate on a time-​limited basis. Each partner brings a complemen-
tary perspective and expertise to the complex multifaceted field of vaccines
and immunization, as well as well representing the interests of their respective
constituencies, and providing technical support This complex partnership
and governing arrangement has stood the test of time and ensures that all key
players have an opportunity to express their concerns and priorities.
The Gavi Secretariat is based in Geneva, with an office in Washington,
DC. The Secretariat is responsible for the operation of Gavi. Gavi has an in-
ternational institution status in Geneva, Switzerland, and a public charity
status in the US. Gavi has no in-​country presence and relies on its core
 5.3.1 The role of Gavi and other institutions 353

partners (WHO, UNICEF, and the World Bank) as well as extended part-
ners, for technical support.
WHO is the UN agency charged with normative functions such as set-
ting standards for healthcare practices, products, and safety, and pre-​
certification of healthcare and pharmaceutical products. The Department
of Immunization, Vaccines and Biologicals (IVB) plays a key role in en-
suring state-​of-​the art scientific practices and safety of vaccines supported
by Gavi. The agency also plays an important role in advising Ministries of
Health of member states, and through the World Health Assembly devel-
oping global consensus through resolutions on pressing health matters such
as immunization.
UNICEF is the UN agency charged with protecting the health of women
and children. Through its country offices in every member state, UNICEF
provides critical programmatic and funding support to countries on
implementing immunization programs. UNICEF Supply Division acts as the
main procurement agency for vaccines and immunization cold chain logistics
and equipment for Gavi-​eligible countries. This function is particularly im-
portant in helping Gavi-​eligible countries procure quality and effective vac-
cines efficiently and at subsidized prices.
The World Bank is an international institution that provides financial assis-
tance, such as loans and grants, to governments of low-​and middle-​income
countries for development projects. The Bank was involved in the develop-
ment of two innovative financing mechanisms—​the International Finance
Facility for Immunization (IFFIm) and the Advance Market Commitment
(AMC)—​that raise additional resources for vaccines. The Bank acts as a
steward for the IFFIm program and manages the AMC, while serving as a de-
velopment partner to Gavi (World Bank, 2015).
Extended partners include international and country-​level civil society or-
ganizations and consulting firms with expertise and experience in providing
technical support.

5.3.1.2 Donor funding to Gavi

Every 5 years, Gavi holds a donor pledging conference and raises funds to
finance a new phase of Gavi with a 5-​year strategy. By March 2021, Gavi had
received nearly US $21 billion through donor contributions (Table 5.3.1).
Approximately 57% of the contributions are from direct donations provided by
governments. The UK, the US, and Norway are the top three donors, followed
354 Donor architecture for immunization financing

Table 5.3.1 Donor contributions to Gavi, 2000–​March 2021

Donors Contributions % of total

(in US $, millions) contributions

Government contributions 12,018 57.3

UK 3,000 14.3
US 2,760 13.2
Norway 1,908 9.1
Germany 876 4.2
Canada 681 3.2
Netherlands 606 2.9
Sweden 587 2.8
Others 1,600 7.6
Private contributions 4,300 20.5
Bill & Melinda Gates Foundation 4,051 19.3
Others 249 1.2
Innovative financing mechanisms 4,661 22.2
International Finance Facility for 3,348 16.0
Immunization (IFFIm)
Advance Market Commitment (AMC) 1,313 6.3
Total 20,979 100.0

by Germany. Private donations account for 24% of total contributions, with

the Bill & Melinda Gates Foundation accounting for 19% of total donations to
Gavi. Other notable private donors include TikTok, Mastercard, Comic Relief,
and the International Federation of Pharmaceutical Wholesalers. Around
22% of the funding comes from the proceeds of IFFIm and AMC. IFFIm is
projected to ramp up significantly in the 2021–​2037 period.
For the next phase of Gavi from 2021 to 2025, an additional US $8.8 billion
was pledged by world leaders at the Global Vaccine Summit 2020, surpassing
the target of US $7.4 billion (Gavi 2021a). This included US $2 billion from the
UK, US $1.6 billion from the Bill & Melinda Gates Foundation, and US $1 bil-
lion from Norway (Gavi, 2020d).

5.3.1.3 Bilateral donors

In addition to the main central donor funds that support Gavi’s strategy and
vaccine and programmatic grants to countries outlined above, other funding
for countries’ immunization activities comes in the form of bilateral donor
support provided for immunization or to the broader health sector directly
 5.3.2 Gavi policies and support to countries 355

in-​country through development cooperation agencies. Some of this sup-

port, for instance, to strengthen health workforce, infrastructure, information
systems, or transport and logistics, ends up also benefitting or strengthening
immunization activities. Some of the agencies that have historically provided
direct in-​country support to Gavi-​eligible countries in the form of bilateral aid
to governments or through nongovernmental organizations include USAID,
the Japan International Cooperation Agency, and the UK Department for
International Development. For example, USAID investments help build and
sustain routine immunization systems to deliver Gavi-​supported vaccines. In
Tanzania, USAID supported the pilot of an innovative Vaccine Information
Management System to improve access and use of data at all levels and to drive
improvements in routine immunization, which is now being supported with
resources from Gavi (USAID, 2018). In Mozambique, USAID funded health
worker training, supply chain procurement, and system strengthening to sup-
port the distribution of new vaccines (USAID, 2020).

5.3.2 Gavi policies and support to countries

5.3.2.1 Types of Gavi support

Gavi offers two main types of support to eligible countries:

1. New vaccine support. This is support for new and underutilized vac-
cines. As of 2020, Gavi supports vaccines that protect against 17 infec-
tious diseases.
2. Health system and immunization strengthening (HSIS) support. This
support bundles several types of cash support to countries. HSIS grants
support priorities to improve delivery of immunization services. New
vaccine introduction grants support some of the introduction activ-
ities for new vaccine introduction. Gavi supports vaccines and opera-
tional costs of immunization campaigns. Beginning in 2021, support
for cold chain equipment previously provided through the Cold Chain
Equipment Optimization Platform was rolled into HSIS support.

Additionally, Gavi supports immunization partners to provide targeted tech-

nical assistance at country level to improve coverage rates and reduce equity
barriers.
Since its inception through July 2019 (last data available), Gavi has
disbursed over US $1.6 billion to support 76 low-​and middle-​income
356 Donor architecture for immunization financing

Table 5.3.2 Gavi country support by type, 2000–​July 2019 (latest available data)

Types of support Funding in US $ millions

(percentage)

New and under-​utilized vaccines 10,432.93 (70.9)

Health system and immunization strengthening (HSIS) 2,156.8 (14.7)
Vaccine introduction grants 196.0 (1.3)
Other cash support 121.7 (0.8)
Operational costs for immunization campaigns 642.9 (4.4)
Outbreak support and response 1.055.16 (7.2)
Other programs 99.4 (0.7)
Total 14,704.89

countries (Table 5.3.2). About 67% of the disbursements were towards new
and underutilized vaccines, followed by cash support including HSIS, intro-
duction grants, and other (16.9%). HSIS alone accounted for 14.7% of Gavi
disbursements (Gavi 2021c).
To support countries during the coronavirus disease 2019 (COVID-​19)
pandemic, Gavi provided new flexibility to allow up to 20% of HSIS funding
to be reallocated to the COVID-​19 response, and to repurpose technical assis-
tance implemented by partners to the COVID-​19 response.

5.3.2.2 Gavi eligibility and transition policy

Since its creation, Gavi has been committed to working with countries to
achieve financial sustainability of their immunization programs. Through its
eligibility and transition policy and its co-​financing policy, Gavi aims to en-
courage countries to expand their immunization programs with vaccines of
public health importance, increase domestic ownership of vaccine financing,
and fully self-​finance vaccines when Gavi support ends. As of the end of 2019,
16 countries have successfully transitioned from Gavi support (Gavi, 2019a).
In the beginning, Gavi aimed to support the world’s poorest countries, and
selected 75 countries with gross national income (GNI) per capita below US
$1,000 as the focus for eligibility. In 2011, Gavi’s GNI per capita eligibility
threshold was reset at US $1,500, with the threshold updated annually to ac-
count for inflation. In 2015, the eligibility policy was revised to determine
eligibility based on average GNI per capita over a 3-​year period, with some ad-
justments for large single-​year increases in GNI per capita, in order to smooth
 5.3.2 Gavi policies and support to countries 357

out annual fluctuations. As of 2019, the threshold stood at US $1,580, with 58

countries eligible for Gavi support (Gavi, 2018). Once a country’s GNI per
capita exceeds the eligibility threshold, it begins a transition process, during
which Gavi support is phased out over 5 years.
Due to disruptions caused by COVID-​19, Gavi announced that countries
would retain their eligibility status from 2020 to 2021. Vaccine co-​financing in
2021 would also remain at the 2020 levels.
In 2020, the Gavi Board approved an approach for engaging with former
and never-​eligible middle-​income countries, with GNI per capita up to
US $4,000.

5.3.2.3 Gavi co-​financing policy

To help countries on a pathway to financial sustainability for new vaccines,

Gavi developed the concept of co-​financing in 2006. This was a novel co-​
procurement mechanism whereby a country, depending upon its GNI per
capita, would be responsible for procuring a complementary portion of vac-
cine doses to those procured by Gavi. Country contributions are phased over
time based on national income. For countries whose GNI per capita is below
the World Bank threshold for a low-​income country, their initial co-​financing
was set at US $0.20 per dose for all new vaccines (Gavi, 2016). These countries

Low-income threshold:$995 GNI/capita (2019)

Variable duration Variable duration 5 years

Vaccine price Eligibility threshold: $1,580 GNI per capita (2019)

Gavi support
level

Country
co-financing level
Initial self- Preparatory Accelerated Fully self-
financing transition transition financing

Fig. 5.3.1 How the Gavi co-​financing model works.

Source: Adapted from https://​www.gavi.org/​vacci​nesw​ork/​how-​gavis-​co-​financ​ing-​model-​works
358 Donor architecture for immunization financing

are in Gavi’s initial phase of transition. Once a country’s income surpasses this
threshold, its co-​financing obligation on a per dose basis increases by 15%
per year—​this is referred to as the preparatory transition phase. The dura-
tion of this phase is entirely dependent on the country’s income growth rate.
Once its income exceeds the Gavi eligibility threshold, a country enters ac-
celerated transition. During this period, co-​financing obligations are increase
linearly each year, in order to reach full self-​financing at the end of 5 years. If
a country’s GNI per capita falls below the threshold amount after it enters ac-
celerated transition, it would regain its eligibility status.
A Board-​approved modification to Gavi’s co-​financing policy aims to
simplify the calculation of co-​financing requirements to a percentage of
doses. Operationalization of this policy decision is still under development
(Gavi, 2019b).

5.3.3 Transition from Gavi support

5.3.3.1 Transition planning

The Gavi Secretariat and partners work with countries to plan for both the
financial and programmatic aspects of transition from support. Transition
assessments were conducted and plans were developed approximately
2–​3 years before a country was projected to enter accelerated transition.
Based on the transition plan, Gavi supported activities deemed critical for
successful transition.
Unlike the preparatory transition phase, which is not time-​bound, the ac-
celerated transition phase is limited to a 5-​year period. During the accelerated
transition phase, Gavi continues to support already introduced vaccines and
maintains its existing commitments for HSIS support. In 2018, Gavi revised
its eligibility and transition policy to allow countries to apply for new vaccine
support while in the accelerated transition phase.
Gavi launched the Learning Network for Countries in Transition in 2017 to
support successful country transitions. The Learning Network for Countries
in Transition uses a peer-​learning approach to provide countries with prac-
tical solutions to address transition challenges. The Learning Network for
Countries in Transition facilitates access to peer countries’ experiences, tech-
nical resources, and global expertise to support countries to develop solu-
tions to transition challenges (Learning Network for Countries in Transition,
2021). Currently, the peer learning platform has been renamed to the Linked
Learning Action Network.
 5.3.3 Transition from Gavi support 359

5.3.3.2 Transition challenges

The highest priorities for most countries as they prepare for transition are
often domestic resource mobilization and vaccine procurement. In addition
to securing adequate funding for vaccines, countries must decide how best to
manage vaccine procurement. While countries have the option of procuring
through UNICEF Supply Division, which manages procurement of Gavi-​
supported vaccines, conflicts between domestic procurement regulations and
UNICEF procedures can be challenging to overcome. Some countries with
small populations face challenges securing competitive pricing with inde-
pendent procurement.
Transitioning countries also face challenges to their immunization
programs posed by the broader health financing context. Countries that are
expanding social health insurance are in the midst of decisions regarding
whether immunization and other preventative services are included as part
of insurance benefits or funded separately. If immunization is funded through
health insurance, key program functions may need to be transferred to the in-
surance administrator. Countries with decentralized health systems may need
to mobilize funding from multiple entities, including subnational govern-
ments or subnational health offices, in order to fund necessary immunization
functions. Transitioning countries may also face growing vaccine hesitancy
among their population that threatens the achievements of the immunization
program.
Some countries face more significant programmatic and institutional chal-
lenges to scaling-​up and sustaining immunization coverage, even as they
prepare for transition away from Gavi support. For the countries that have
more deep-​rooted obstacles to achieving high coverage, Gavi works with the
country to develop tailored plans to support successful transition. In rare
cases where a country faces a high risk of unsuccessful transition, specific
flexibilities, including time-​limited extension of the accelerated transition
phase, may be approved (Gavi, 2019b).

5.3.3.3 Post-​transition engagement

Gavi continues to engage with countries after they transition away from Gavi
support, including monitoring vaccine coverage rates and new vaccine intro-
duction. Immunization program sustainability is facilitated by affordable vac-
cine prices. Gavi has negotiated with vaccine manufacturers to secure pricing
commitments so that transitioned countries can access prices for pentavalent,
360 Donor architecture for immunization financing

pneumococcal, rotavirus, and human papillomavirus vaccines at prices sim-

ilar to those paid by Gavi for 5–​10 years after transition (Gavi, 2020c).
In addition, the Gavi Board allocated US $30 million to mitigate any risks
for successful transition (Gavi, 2020c). These risks include those related to
supply chain and data system weaknesses, as well as leadership, manage-
ment, and coordination. This post-​transition support has encouraged gov-
ernment commitment to long-​term financial sustainability and innovation
of the program (Gavi, 2020c).

5.3.3.4 Middle-​income country engagement

In December 2020, the Gavi Board approved engagement with middle-​

income countries, including former-​Gavi (transitioned) countries and never
Gavi-​eligible (previously ineligible) countries with per capita GNI under US
$4,000. The objectives of the engagement are to prevent backsliding in former
Gavi countries and to drive the sustainable introduction of key missing vac-
cines in former and select never Gavi-​eligible countries.
The primary focus over the first 18 months is to prevent and mitigate back-
sliding in former-​Gavi countries, prioritizing those countries with large num-
bers of children who have received no vaccine doses. Gavi engagement will
focus on building political will and advocating for improving equity in rou-
tine immunization, proving technical assistance to re-​establish routine im-
munization services, targeted interventions to restore coverage and catch up
missed children, and fostering peer-​to-​peer learning for disseminating best
practices. Gavi will lay the groundwork for future introductions of pneumo-
coccal conjugate, rotavirus, and human papillomavirus vaccines by leveraging
its country-​level engagement with COVID-​19 vaccines (Gavi, 2020a). More
guidance from Gavi on the practical implementation of this decision will be
needed, particularly as it relates to countries that are currently in accelerated
transition.

5.3.4 Innovations for immunization financing

5.3.4.1 Advance market commitments

AMCs have been shown to be an effective mechanism for mobilizing re-

sources and incentivizing development of new vaccines for lower-​income
 5.3.4 Innovations for immunization financing 361

markets. By announcing a donor commitment to purchase vaccines at a set

price with predetermined specifications and level of demand, this would in-
centivize manufacturers’ and producers’ interest in developing a vaccine.
Greater reliability and certainty around price and demand would motivate
vaccine manufacturers to invest in development. Based on the initial con-
cept developed by the Center for Global Development, the World Bank and
Gavi convened experts in public health, health economics, and contract
law to develop key terms and structure of a pilot AMC (Levine, Kremer, &
Albright, 2005).
The pilot AMC for pneumococcal vaccine was announced in February 2007.
The Governments of Italy, the UK, the Russian Federation, Canada, Norway, and
the Bill & Melinda Gates Foundation pledged a total of US $1.5 billion to fund
the AMC. By March 2010, GlaxoSmithKline and Pfizer both agreed to provide
30 million doses per year for up to 10 years. While both the GlaxoSmithKline
and Pfizer vaccines were already in advanced stages of development when the
AMC was announced, the long-​term demand stimulated by the AMC led to
investments to expand capacity and to create presentations specifically for Gavi
countries. The AMC also contributed to bringing in two donors who had previ-
ously not provided funding to Gavi.
In June 2020, Gavi launched the COVAX AMC led by Gavi, the Coalition for
Epidemic Preparedness (CEPI), and WHO. The Facility is designed to guar-
antee rapid, fair, and equitable access to COVID-​19 vaccines for 92 low-​and
middle-​income countries. Another 81 self-​financing countries have also indi-
cated interest in COVAX. The Facility pools resources from many countries,
shares risks by investing in a portfolio of vaccine candidates, and will pro-
vide vaccines for low-​income countries. The AMC was established to mobilize
resources from Organisation for Economic Co-​operation and Development
countries with a goal of providing 2 billion doses of COVID-​19 vaccine by the
end of 2021 to low-​and middle-​income countries. The AMC enters advanced
purchase agreements with suppliers prior to vaccine licensure and WHO
prequalification. As of May 10, 2021, COVAX has shipped 58 million doses of
COVID-​19 vaccine to 122 countries (Gavi, 2021b).

5.3.4.2 International Finance Facility for

Immunization vaccine bonds

The IFFIm is a global development finance tool launched by Gavi in 2006. Based
on a concept developed by the UK Treasury and Goldman Sachs, pledges from
362 Donor architecture for immunization financing

donor governments are used to issue bonds, often referred to as “vaccine bonds,”
on capital markets around the world to finance immunization programs. IFFIm
bonds are backed by irrevocable, legally binding funding pledges from ten sov-
ereign governments, including the UK, France, the US, and Norway. More in-
formation is available at https://​iffim.org/​don​ors. Since its inception, IFFIm
has attracted more than US $6.5 billion in sovereign pledges and disbursed US
$2.6 billion (as of October 2018) to support immunization, or approximately
one-​quarter of Gavi’s program funding since 2006. The World Bank acts as the
treasury manager for IFFIm.
In July 2020, IFFIm issued NOK 2 billion (US $224 million) in vaccine
bonds to finance COVID-​19 vaccine research and development, against a 10-​
year commitment from the Kingdom of Norway. The IFFIm mechanism al-
lowed a 10-​year pledge to be converted into cash in order to maximize urgent
investments in vaccine candidates and vaccine trials.

5.3.5 Partnerships with private sector

Since its inception, private sector partners have played a key role in supporting
Gavi’s mission. Gavi has mobilized the private sector to contribute both fi-
nancially and technically to improve immunization program operations. Gavi
continues to seek new private sector partnerships, aiming to drive innovation,
cut costs, and increase operational efficiency to achieve its goals.
With the UPS Foundation, Gavi engaged in a vaccine delivery network
innovation in three districts of Uganda reaching more than 150 clinics that
serve 3 million people. The pilot employed a range of supply chain tools and
innovations. A stock management system was used to help reduce vaccine
stock-​outs. A “Strategic Training Executive Program” leadership development
program helped build management supply capacity.
Deutsche Post DHL Group was another Gavi partner whose expertise in
transportation and logistics was applied to immunization. Using DHL’s global
network, the Ministry of Health of Kenya, DHL, and Gavi developed a trans-
portation management innovation. A transport support hub coordinated
third-​party transport carriers to improve the reliability and speed of vaccine
deliveries to points of service delivery.
The Zipline drone network in Ghana works with the national supply chain
to prevent vaccine stockouts in facilities and for campaigns. The current ser-
vice of 30 drones covers 2,000 health facilities. UPS also provides technical
guidance and consultancy services as needed. Zipline drones are also used in
 5.3.6 Conclusion 363

delivery of COVID-​19 vaccines to the hardest-​to-​reach areas of Ghana. The

drone vaccine network builds on Zipline’s experience in Rwanda to provide
blood products to remote clinics in Rwanda.
To identify and fund potential innovations, Gavi established the INFUSE
platform (Innovation for uptake, scale and equity in immunization) to con-
nect private sector partners and funders with high-​tech innovators (Gavi,
2020b). One innovation to improve cold chain performance relies on Nexleaf
sensor technology to provide real-​time data paired with targeted analytics to
analyze the root causes of cold chain failures in the field, which allows for pri-
oritization of resources, efficient repairs, and informed procurement. Also
identified through INFUSE, Zenysis provides its software platform, analytical
training, and IT skills development to integrate data from their fragmented
information systems and help decision makers see where children are not re-
ceiving vaccines. Advanced analytics will then help countries decide how to
target their limited resources for maximum impact.
Gavi is also collaborating with Airtel to improve the quality of reporting
on immunization coverage through the E-​Reporting Project through the
digitization of paper-​based reports, and provision of digital tools to primary
healthcare workers across Nigeria.
These advanced logistics, transport, and technology are innovations that have
potential to strengthen vaccine supply chains, identify under-​immunized chil-
dren, and improve immunization data. Whether they are true breakthroughs in
resolving longstanding challenges will depend on whether they can be imple-
mented sustainability at scale.

5.3.6 Conclusion

This chapter provides an overview of the various public–​private partnerships

that have served to support immunization coverage, new vaccine introduc-
tion, and strengthened immunization systems since the 1980s. Gavi has and
continues to provide a significant level of funding and technical support to
countries, as well as to lead on innovations related to financing and sustaina-
bility of the program. It was one of the first partnerships that developed a cost-​
sharing/​co-​procurement relationship with countries and was the basis for
innovations such as the AMC and IFFIm which are being explored for other
life-​saving commodities. Innovative partnerships with the private sector on
data and supply chain can promote delivery of life-​saving vaccines to the re-
motest populations.
364 Donor architecture for immunization financing

References
Clemens, J., Holmgren, J., Kaufmann, S., & Mantovani, A. (2010). Ten years of the Global
Alliance for Vaccines and Immunization: Challenges and progress. Nature Immunology, 11,
1069–​1072. doi:10.1038/​ni1210-​1069
Gavi. (2016). Gavi, the Vaccine Alliance co-​financing policy. https://​www.gavi.org/​sites/​defa​ult/​
files/​docum​ent/​gavi-​co-​financ​ing-​policy​pdf.pdf
Gavi. (2018). Eligibility and transition policy. https://​www.gavi.org/​pro​gram​mes-​imp​act/​progr​
amma​tic-​polic​ies/​elig​ibil​ity-​and-​transi​tion​ing-​pol​icy
Gavi. (2019a). Annual program report. https://​www.gavi.org/​sites/​defa​ult/​files/​pro​gram​mes-​
imp​act/​our-​imp​act/​apr/​Gavi-​Progr​ess-​Rep​ort-​201​9_​1.pdf
Gavi. (2019b). Gavi 5.0: Funding policy review. https://​www.gavi.org/​sites/​defa​ult/​files/​board/​
minu​tes/​2019/​4-​dec/​09%20-​%20G​avi%205.0%20Fund​ing%20Pol​icy%20Rev​iew.pdf
Gavi. (2020a). Gavi’s approach to engagement with former and never-​eligible middle-​income
countries (MICs). https://​www.gavi.org/​sites/​defa​ult/​files/​board/​minu​tes/​2020/​15-​dec/​
07%20-​ % 20G​ avi%27s%20a​ p pro​ a ch%20to%20eng ​ a gem ​ e nt%20w ​ i th%20for ​ m er%20
and%20ne​ver-​eligi​ble%20M​ICs.pdf
Gavi. (2020b). INFUSE. https://​www.gavi.org/​invest​ing-​gavi/​inf​use
Gavi. (2020c). Transitioning out of Gavi support. https://​www.gavi.org/​types-​supp​ort/​sus​tain​
abil​ity/​tra​nsit​ion
Gavi. (2020d). World leaders make historic commitments to provide equal access to vaccines for
all. https://​www.gavi.org/​news/​media-​room/​world-​lead​ers-​make-​histo​ric-​comm​itme​nts-​
prov​ide-​equal-​acc​ess-​vacci​nes-​all
Gavi. (2021a, March 31). Cash receipts 31 March 2021. https://​www.gavi.org/​news/​docum​ent-​
libr​ary/​cash-​recei​pts-​31-​march-​2021
Gavi. (2021b). COVAX. https://​www.gavi.org/​covax-​facil​ity
Gavi. (2021c). Disbursements and commitments. https://​www.gavi.org/​pro​gram​mes-​imp​act/​
our-​imp​act/​disbur​seme​nts-​and-​comm​itme​nts
Hardon, A., & Blume, S. (2005). Shifts in global immunisation goals (1984–​ 2004):
Unfinished agendas and mixed results. Social Science & Medicine, 60(2), 345–​356. doi: 10.1016/​
j.socscimed.2004.05.008
Learning Network for Countries in Transition. (2021). Homepage. https://​lnct.glo​bal/​
Levine, R., Kremer, M., & Albright, A. (2005). Making markets for vaccines: Ideas to action.
https://​w ww.cgdev.org/​sites/​defa​ult/​f iles/​arch​ive/​doc/​b ooks/​vacc​ine/​Making​Mark​ets-​
compl​ete.pdf
Muraskin, W. (2005). Crusade to immunize the world’s children. Durham, NC: Lulu.com.
USAID. (2018). USAID’s investment to Gavi, the Vaccine Alliance. https://​www.usaid.gov/​sites/​
defa​ult/​files/​docume​nts/​1864/​GAVI​_​fac​tshe​et_​4​_​16-​508.pdf
USAID. (2020). Support community health interventions to address neonatal and child health—​
UNICEF. https://​www.usaid.gov/​moz​ambi​que/​fact-​she​ets/​supp​ort-​commun​ity-​hea​lth-​interv​
enti​ons-​addr​ess-​neona​tal-​and-​child
World Bank. (2015). Global program review: The World Bank’s partnership with the GAVI
Alliance. https://​openkn​owle​dge.worldb​ank.org/​bitstr​eam/​han​dle/​10986/​23612/​The0Worl​
d0Ba​nk0t​h0th​e0GA​VI0a​llia​nce.pdf
Zerhouni, E. (2019). GAVI, the Vaccine Alliance. Cell, 179(1), 13–​17. doi:10.1016/​j.cell.2019.08.026
 APPENDIX 1

Exercise: a case study on estimating the

total and unit routine immunization costs
from the facility to the national level
Ijeoma Edoka1

A.1.1 Background
The objectives of the exercise are to:
• Identify recurrent and capital cost components of an immunization program.
• Calculate average total costs at facility level using sampling weights.
• Generate a weighted national estimate of total and unit costs using the averaging method.
Accurate data on the costs of routine and new vaccines immunization can improve country-​
level planning and financing for immunization. These data can build a base of evidence to
provide inputs for policy and resource mobilization domestically and externally. Applying
standardized methods to analyze facility-​based data will make it easier to compare and apply
results more widely.
The Expanded Program on Immunization Costing and Financing (EPIC) project was an in-
itiative that collected facility-​based data on the costs of immunization across six countries. The
study used an approach which applied a bottom-​up, ingredients-​based costing methodology
from the perspective of a health service provider. The study created a unique pooled dataset of
316 sites to explore cross-​country determinants of costs (Brenzel, Young, & Walker, 2015). This
exercise below will focus on the data collected from the Uganda EPIC study.
Uganda had a population of approximately 34.5 million people in 2011. The types of health
facility in Uganda are health centers (HC; level II, III, and IV) and hospitals (general, re-
gional, and national hospitals). An estimated 72% of the population lived within 5 km of a
health facility in 2010, and coverage with most vaccines has remained above 80% but with
variations between districts (Guthrie et al., 2014).
The Uganda study applied a multistage, purposive, and stratified random sampling approach
(Brenzel et al., 2015). All ten regions were represented and one or two districts per region were
purposively sampled to represent a range of typical service contexts. Fifty-​two health facilities
were randomly sampled from the strata of health facilities in these districts (general hospitals
and HC II, III, and IV) and 49 of them were included in the study (Guthrie et al., 2014). The
costs were estimated retrospectively for 2011 and were captured in Ugandan shillings (UGX).

1. Open the excel document named “Costing Exercise”. Additional content on (Costing
Exercise) is available online, 10.1093/oso/9780192896087.012.0001. In Sheet A, you have
been provided with the raw data collected during the Uganda study. In the subsequent
sheets, you will be analyzing the data.

1 ‘Teaching Vaccine Economics Everywhere’ materials: https://​immuni​zati​onec​onom​ics.org/​

366 Appendix 1

2. Go to Sheet A and briefly go through the data provided. You will notice that column B
contains the cost and output items, and columns C–​AY contain the values of the items
corresponding to each facility. Hover on the cells in column B to see a more detailed defi-
nition of the item.

A.1.2 Categorize cost items and complete the

cost table
3. Cells B7–​B27 contain cost line items. In column A (the red box), mark the items with a ‘C’
if they are capital costs, or with a ‘R’ if they are recurrent costs.
4. Go to Sheet B and you will find Table B-​1, the cost table for the HC III facility Kiswa.
5. Complete column B of the table by filling in the values for the cost items. Column C will
automatically be filled in with the USD values. The values for Kiswa will be found in Sheet
A, column AA.
6. Calculate the total capital cost (cell B6), the total recurrent cost (cell B17), and the overall
total cost (cell B19) for Kiswa.
7. Complete column D of the table in Sheet B by calculating the share of the total cost for
each item.
8. What are your observations about the costs of the Kiswa facility?

A.1.3 Sampling weights

Analysts use sampling weights to get nationally representative value estimates of average
weighted facility costs. The sampling weights for each facility’s costs were calculated sepa-
rately based on inverse probabilities of sampling at each sampling stage.
• A sample of I regions was selected from a total of L regions in the country, and their prob-
ability of selection was =​I × L (IL).
• A sample of n districts was selected from a total of N districts in a region, and their proba-
bility of selection was thus =​n × N (nN).
• A sample of m facilities have the probability of being selected from a total of Mi similar
health units in the same district (i) was equal to m × Mi (mMi).
The overall probability of selection of a specific health unit in a district is therefore =​
IL × nN × mMi.
The weight of a sampled health unit was the reciprocal of its probability of being selected:
1/IL × nN × mMi = LNMiInm
In Uganda, there ten regions and 112 districts. All regions were included in the study and so
IL =​1, and that simplifies the sample weights to NMinm.
9. Lira is one of the two districts sampled from the region mid-​north of Uganda. This region
has 15 districts in total. Assuming Lira has been selected at random and that Lira is rep-
resentative of mid-​north Uganda, how much weight do we need to give to observations
from Lira when we include them in estimates that are intended to be nationally represen-
tative of Uganda?
10. Ober is one of the two HC III facilities sampled in Lira. There are 14 HC III facilities in
this district. Let us assume that Ober has been selected at random and that Ober is a rep-
resentative HC III facility in Lira.
a. How much weight do we need to give to observations from Ober when we include
them in estimates that are intended to be representative of Lira district?
 Appendix 1 367

b. What weight should we apply to make Ober’s contribution appropriate for a nationally
representative estimate?

A.1.4 Weighted estimate of total immunization costs

at facilities
We want to calculate the total cost of carrying out immunization at facilities in Uganda and to
break this total cost down by the different types of facilities from HC II, HC III, all the way to
regional hospitals.
11. Go to Sheet C. Here you will find data on the facility name, facility type, total facility cost,
and the adjusted sample weights of all facilities. In Sheet D, calculate the weighted esti-
mate of total cost for each health facility type that is nationally representative.
a. For each facility, multiply the total cost with their sample weight in column E to obtain
the weighted facility total cost.
b. Select row 1. Click on Sort & Filter button in the toolbar (on the right-​hand side).
Select Filter.
c. Click on the drop-​down button on the column B header, cell B1. Select Sort A to Z.
d. Go to Sheet D, where you will Table D-​1 set up to calculate average total cost by fa-
cility type.
e. In cell B2 of Sheet D, sum all the weighted total costs for HC II.
f. Similarly, in cell B3 of Sheet D, sum all the facility weights for HC II.
g. In cell B4 of Sheet D, divide the weighted total costs by the sum of the facility weights
to obtain the weighted average facility cost for HC II facilities.
h. Repeat steps e–​g for other facility types.

A.1.5 Total costs and unit costs of Ugandan

routine immunization
The handbook describes various options for aggregating costing data from a sample into a total
national estimate. These options vary by how much bias they introduce into the total cost estimate
for the country (Clarke-​Deelder, Vassall, & Menzies, 2019).
A relatively simple approach to aggregating costs is presented below. The total national cost
of routine immunization in Uganda can be calculated as:
Equation A.1.1:
Tcountry = THF + TDistrict + TNational

In the previous step, we calculated a weighted estimate of the total cost (including vaccines) for
each facility type. We apply this to all of the same type of facilities in the country. The average
total cost of the sampled districts, excluding vaccines, was then applied to all the districts in the
country. These aggregated facility-​and district-​level estimates were then added to the national-​
level spending (after excluding the national-​level vaccine costs).
To calculate nationally representative estimates of average cost, we simply divide the total
national cost by a denominator. Commonly used denominators are doses, infant population,
three-​dose diphtheria, tetanus, and pertussis (DTP3)-​vaccinated children yielding costs per
dose, cost per infant, and cost per DTP3.
12. Go to Sheet E. Table E-​1 is set up to calculate total national routine immunization cost. Table E-​2
is set up to calculate unit costs of national routine immunization.
368 Appendix 1

From prior calculations we happen to know that average cost per district-​level facility is 29,923
USD and that total central facility cost is 4,511,957 USD. We also know that there are 2215 HC
II facilities, 1180 HC III facilities, 185 HC IV facilities, 127 general hospitals, and 112 district-​
level facilities.
13. Complete column B for the facilities section using the values calculated in the previous step.
14. In column C, calculate the total facility cost for each row in the facility and district
sections. (Multiply the weighted average total costs with the number of facilities of the
same type present in Uganda.)
15. Calculate the total immunization costs in cell D12 (Tcountry =​THF +​TDistrict +​TNational).
16. Calculate the share of total cost for each row.
17. What do you observe?
The population of Uganda in 2011 was 32,939,800 and the infant population was 1,476,164.
The total child doses administered was 11,964,835 and the total DTP3 vaccinated children were
1,219,455 (estimates derived from the Comprehensive Multi-​Year Plan 2011).
18. Complete the table (Table E-​2) below the total cost table in Sheet D with the above
estimates.
19. Calculate the cost per dose, cost per infant, and cost per DTP3-​vaccinated child.

A.1.6 Analyzing the relationship between unit costs

and output

20. Go to Sheet F. For each facility, you are given the total facility cost, total doses adminis-
tered, and total DTP3-​vaccinated children.
21. In Table F-​1, calculate the cost/​dose for each facility. (Tip: enter the formula in cell B7, se-
lect the cell, and drag it horizontally to autofill the formula in the row.)
The graph below the table will plot each data point. To see the facilities corresponding to the
data point:
a. Click on the green cross and left click on “Add Data Labels.”
b. Click on the arrow and select click “More Options.”
c. A new window appears to the right, deselect X and Y Value.
d. Enable “Value from cells.”
e. Select cell range B3:AY3.
f. Click OK.
The graph next to the main graph provides a zoomed-​in look at the data points.
22. Can you locate the outliers? What do you observe?
23. In Table E-​2, calculate the cost/​DTP3-​vaccinated child. Repeat step 21.
24. Can you locate the outliers? What do you observe?
25. What can you say about the relationship between unit costs and output volume?

A.1.7 Discussion prompt

26. Carmen has analyzed the graphs of unit costs versus output, and she decided to close
down all clinics that have very high unit costs and minimal output. Would you and your
team advise Carmen to proceed with her decision? Explain your reasoning.
 Appendix 1 369

References
Brenzel, L., Young, D., & Walker, D. G. (2015). Costs and financing of routine immuniza-
tion: Approach and selected findings of a multi-​country study (EPIC). Vaccine, 33, A13–​
A20. doi:10.1016/​j.vaccine.2014.12.066
Clarke-​Deelder, E., Vassall, A., & Menzies, N. A. (2019). Estimators used in multisite healthcare
costing studies in low-​and middle-​income countries: A systematic review and simulation
study. Value in Health, 22(10), 1146–​1153. doi:10.1016/​j.jval.2019.05.007
Guthrie, T., Zikusooka, C., Kwesiga, B., Abewe, C., Lagony, S., Schutte, C.,... Kinghorn, A.
(2014). Costing and financing analyses of routine immunization in Uganda. Health &
Development Africa. https://​stat​ic1.squa​resp​ace.com/​sta​tic/​556de​b8ee​4b08​a534​b836​0e7/​
t/​5596f​a4ae​4b07​b7dd​a4dd​04d/​143595​7834​829/​UGA​NDA+​Immun​izat​ion+​Cost​ing+​Rep​
ort+​1+​Decem​ber+​14+​submit​ted+​FINAL+​upd​ate+​15+​12+​14+​err​ors.pdf
 APPENDIX 2

Exercise: estimating new vaccine

introduction costs
Susmita Chatterjee1

A.2.1 Introduction
Information on the cost of procuring vaccines and of running a national immunization pro-
gram is a key factor in planning and managing an effective immunization program. This infor-
mation is also needed to conduct a cost-​effectiveness analysis of a new vaccine.
Vaccine program costs are context specific and not generalizable from one country to
another—​vaccination schedules, target populations, vaccine introduction strategies, and prices
may vary.
In this exercise, you will be given a scenario and asked to estimate how much it would cost to
introduce a new vaccine into routine immunization, to produce information that can be used as
input into a cost-​effectiveness assessment, or to assist with planning and budgeting. The exer-
cise provides a practical demonstration on the application of concepts introduced throughout
this appendix.
Specific objectives of the exercise are to:
• Gain deeper understanding on the cost components of a cost-​effectiveness analysis and
the types and sources of data required.
• Practice some of the basic processes and logic of estimating immunization program costs.
• Learn some ways to generate cost estimates that can inform decisions about the adoption
of a new vaccine or a change in vaccine program strategies.
• To interpret cost data, and understand their usefulness, challenges, and limitations.

A.2.2 Case scenario

A ministry of health planner has been asked to investigate the introduction of pneumococcal
conjugate vaccine (PCV) into the national routine immunization program in the lower middle-​
income country Contagia. The country has a total population of 14,752,000 and a crude birth
rate of 38/​1000. An estimated 81% of children are fully immunized. The country is a lower
middle-​income country which has been receiving Gavi support but expects to have to fund
more of its immunization program over the next few years. The current routine childhood im-
munization schedule is given in Table A.2.1.
PCV will be in single-​dose vials and administered by injection at the same time as the diph-
theria, tetanus, and pertussis (DTP) vaccine at 6, 10, and 14 weeks of age. Using the information
provided, please answer the following questions:

1 ‘Teaching Vaccine Economics Everywhere’ materials: www.immuni​zati​onec​onom​ics.org

372 Appendix 2

Table A.2.1 Current routine immunization schedule

Antigen Age of administration
OPV0 At birth up to 13 days
BCG At birth or first contact
OPV1, DTP–​HepB–​Hib1 6 weeks of age
OPV2, DTP–​HepB–​Hib2 10 weeks of age
OPV3, DTP–​HepB–​Hib3 14 weeks of age
Measles—​MCV1 9 months of age
Measles—​MCV2 18 months of age

BCG, bacillus Calmette–​Guérin; HepB, hepatitis B; Hib, Haemophilus influenzae type B;

MCV, measles-​containing-​vaccine; OPV, oral polio vaccine.

A.2.3 Exercise: part 1

1. Using the routine immunization schedule (Table A.2.1), calculate the total number of
doses administered in Contagia per annum.
2. What do you think the percent wastage of the PCV vaccine should be based on? How could
you more accurately estimate the potential wastage of the PCV?
3. The cost of PCV, under an agreement with manufacturers to incentivize the development
of vaccines, is expected to be $7 per dose for the first 3 years. During this time, the vaccine
will be donor funded. After 3 years, the cost is expected to decline to $3.50 per dose and
will be funded by the country. What cost would you use when completing your total vac-
cine cost calculation and why?
4. Use your estimates of vaccine price from question 3 and wastage from question 2 to calcu-
late total PCV costs.

A.2.4 Exercise: part 2

5. Table A.2.2 shows the total costs of delivering all vaccines in the country’s routine im-
munization program before PCV introduction. How are the results useful from a pro-
gram planning and management perspective?
6. What are the major cost drivers of the existing immunization program in this country?
7. Which costs of introducing PCV will be most important to estimate accurately for
a cost-​effectiveness analysis? Would any of these be less important for program
budgeting?
8. For national programs, an important estimate is the cost per fully immunized child. The
simplest way to estimate this is to divide the total costs of the program by the number of
children with a complete vaccine schedule.
a. Using information from Table A.4.2, what is the cost per fully vaccinated child in this
country?
b. List one or two limitations on using estimates of the cost per fully immunized child.
9. Using information provided in Tables A.2.1 and A.2.2:
a. Estimate the total cost per dose of all vaccines currently in the immunization schedule
where total cost is cost of vaccine plus non-​vaccine (or delivery) costs. (For the two
vaccines given at birth, OPV0 and BCG, assume 100% coverage. For the other vac-
cines in the schedule, assume 81% coverage.)
 Appendix 2 373

Table A.2.2 Total costs of routine immunization program in Contagia before PCV
introduction (2016)
Cost ($, 000) % of cost
Capital costs
Cold chain 570 1.5
Vehicles 2,000 5.2
Buildings 1,085 2.8
Other 557 1.5
Subtotal 4,212 11.0

Recurrent costs
Vaccine and supplies 6,168 16.2
Vaccine injection and safety supplies 186 0.5
Paid labor 18,130 47.5
Volunteers/​community health workers 730 1.9
Allowances for travel/​subsistence 4,390 11.5
Cold chain (energy) 120 0.3
Vehicles (maintenance, fuel) 2,770 7.3
Communications 725 1.9
Building maintenance, utilities 350 0.9
Other supplies and printing 372 1.0
Subtotal 33,941 89.0
Total 38,153 100

b. Estimate the service delivery cost per dose, that is, non-​vaccine cost in the current im-
munization schedule.
c. Estimate the total cost per dose of PCV. Note any limitations of your estimate.
10. Do you think that the total cost per dose estimated in question 9 will vary within a
country? Where would you expect total cost per dose to be higher or lower?
11. PCV is usually administered alongside the DTP vaccine at 2, 4, and 6 months. Some
countries now use a “2 +​1” schedule (i.e. two doses alongside the DTP series plus one
booster later—​usually around 12 months). What would be the implications (financial and
logistical) of implementing the new schedule?
12. What are the most important components that need to be taken into consideration to cal-
culate the cost of introducing PCV into your own country’s routine immunization pro-
gram? List up to five major inputs of this program.

Further reading
Griffiths, U. K., Bozzani, F. M., Chansa, C., Kinghorn, A., Kalesha-​Masumbu, P., Rudd,
C.,... Schutte, C. (2016). Costs of introducing pneumococcal, rotavirus and second dose
measles vaccine into the Zambian immunization program: are expansions sustainable?
Vaccine, 34(35), 4213–​4220. doi:10.1016/​j.vaccine.2016.06.050
 APPENDIX 3

Immunization activities and line item costs

Logan Brenzel

Table A.3.1 illustrates a crosswalk between immunization line items (cost elements) and ac-
tivities (Brenzel, 2014). When analyzing immunization service costs, analysis can be done to
represent costs in these different ways.
Table A.3.1 Immunization activities and line item crosswalk for routine immunization
Line item Immunization activities
Routine Record Supervision Outreach Training Social Surveillance Cold chain Vaccine Program Other
facility-​based keeping & service mobilization maintenance collection, management
service delivery HMIS delivery & advocacy distribution,
storage

Salaried labor
Volunteer labor
Per diem & travel
allowances

Vaccines
Vaccine injection &
safety supplies

Other supplies
Transport/​fuel
Vehicle
maintenance

Cold chain energy

costs

Printing
 Building
overhead, utilities,
communication

Other recurrent
Cold chain
equipment

Vehicles
Lab equipment
Other equipment
Other capital
Buildings
TOTAL

Source: Brenzel (2013).

378 Appendix 3

Reference
Brenzel, L. (2014). Working Paper: Common approach for the costing and financing analyses
of routine immunization and new vaccine introduction costs (EPIC). Mimeograph. Bill &
Melinda Gates Foundation. www.immuni​zati​onec​onom​ics.org/​epic-​info
 APPENDIX 4

Markov decision processes

Emmanuel F. Drabo and William V. Padula

A.4.1 Designing a Markov decision process

A Markov model of the type described in Chapter 4.1 typically assumes that only one decision
is made, at the start of the model. From that point forward, a sequence of events unfolds ac-
cording to state-​dependent transition probabilities.
A Markov decision process (MDP) extends a simple Markov model by allowing for mul-
tiple decision points over time. Each decision period is defined by a set of possible health
states, S, a set of possible actions, A, a probability matrix, Ta, and a reward matrix, Ra. The
probability of transitioning from state s to s′ if action a is taken at time t is then defined as
Ta ( s, s ′ ) = Pr(st +1 = s ′ | st = s, at = a) ; the corresponding immediate reward associated with tran-
sitioning from state s to s′ if action a is taken at time t is Ra ( s, s ′ ) .
The solution to the MDP is calculated recursively and is defined by an optimal policy π , and
value function V, defined as follows:
Equation A.4.1:

π ( s ) = argmax ∑Pa ( s, s ′ )V ( s ′ ) st
a s’

Equation A.4.2:

V ( s ) = R ( s ) + β∑Pπ(s ) ( s, s ′ )V ( s ′ )
s’

where β denotes the one-​period discount factor. It should be noted that MDPs optimize in one
dimension (i.e., minimize costs or maximize health benefits), and can thus be unattractive in cost-​
effectiveness analysis. In MDPs, an optimal policy is first determined, and the value function as-
sociated with that optimal policy is calculated. Once these are determined, then the costs and
benefits of different optimal policies with different objective functions are compared using cost-​
effectiveness analysis.

A.4.2 Additional considerations about using

Markov models
At this juncture, it is worthwhile asking when is using Markov model appropriate? As we have
noted at the start of this appendix, Markov models are most useful for representing events that
are recurrent, when we want to model the “natural history” of a disease, when the time horizon
is long, and when the quantity and quality of time spent in different health states are outcomes
of interest, as they are in cost-​effectiveness analysis.
One additional important consideration when developing Markov models is the choice of
computational tool to implement the model. Most Markov models are developed using basic
380 Appendix 4

spreadsheet software such as Microsoft Excel, as illustrated in our example above, or commer-
cial packages such as TreeAge. For simple processes and most vaccine economics problems,
these tools will be sufficient. However, for more complex processes, these tools may prove
inadequate (Filipović-​Pierucci, Zarca, & Durand-​Zaleski, 2017; Williams, Lewsey, Briggs, &
Mackay, 2017).

References
Filipović-​Pierucci, A., Zarca, K., & Durand-​Zaleski, I. (2017). Markov models for health eco-
nomic evaluations: The R package heemod. ArXiv. https://​arxiv.org/​pdf/​1702.03252.pdf
Williams, C., Lewsey, J. D., Briggs, A. H., & Mackay, D. F. (2017). Cost-​effectiveness analysis
in R using a multi-​state modeling survival analysis framework: A tutorial. Medical Decision
Making, 37(4), 340–​352. doi:10.1177/​0272989X16651869
 APPENDIX 5

Derivation of the annual expected costs

associated with the Infected state
Emmanuel F. Drabo and William V. Padula

We calculated the annual expected cost of an acute hepatitis B infection to be $1,481. This
expected annual cost is the weighted average cost of outpatient care for acute symptomatic
infection, the cost of hospitalization for those with acute fulminant hepatitis B infection,
and the cost of hospitalization for non-​fulminant cases, as calculated below:

{
C Acute = pA × C A + (1 − pA ) × (1 − pH ) × CO + pH ×  pF × CF + (1 − pF ) × CH  }
where C denotes costs, and p denotes the probabilities of each event. Specifically, C Acute rep-
resents the annual cost per capita of an acute infection, C A denotes the cost of an asympto-
matic infection (assumed to be zero), CO is the outpatient costs for symptomatic patients,
CF represents the annual hospitalization cost for fulminant cases, and CH denotes the an-
nual hospitalization cost for non-​fulminant cases. Similarly, pA is the proportion of asymp-
tomatic cases among individuals with an acute infection, pH is the hospitalization risk
among symptomatic patients, and pF denotes the proportion of hospitalized patients that
are fulminant cases. Substituting the values of each parameter into the equation, we calcu-
late the health-​related quality of life (HRQoL) weight for acute infections as:

C Acute = 0.70 × $0+(0 − 0.70) × {(1 − 0.38) × $402 + 0.38

× [0.04 × $19, 481 + (1 − 0.04) × $12, 034]} = $1, 481.

The HRQoL utility weight associated with the Infected state is calculated in a similar
fashion as costs:

HRQoLAcute = pA × HRQoLA + (1 − pA ) × {(1 − pH ) × HRQoLO + PH

× [ pF × HRQoLF + (1 − pF ) × HRQoLH ]}

where C denotes costs, and p denotes the probabilities of each event, as defined above.
Specifically, C Acute represents the annual cost per capita of an acute infection, CO is the out-
patient costs for symptomatic patients, CF represents the annual hospitalization cost for
fulminant cases, and CH denotes the annual hospitalization cost for non-​fulminant cases.
Substituting the values of each parameter into the equation, we calculate the HRQoL weight for
acute infections as:

HRQoLAcute = 0.70 × 1.00 + (1 − 0.70) × {(1 − 0.38) × 0.99 + 0.38

×[0.04 × 0.994 + (1 − 0.04) × 0.99]} = 0.997.
382 Appendix 5

The derivation of the annual expected cost and HRQoL utility weights associated with a
chronic infection from the data in Hoerger, Young, and Walker (2013) requires a slightly
more complex analysis. The mathematical derivation of these expected costs goes beyond the
scope of this handbook. But it is enough for the reader to know that these costs are calculated
by accumulating the expected lifetime medical costs of an individual with a chronic hepa-
titis B infection, as they cycle through various stages of chronic hepatitis B disease progres-
sion, including chronic hepatitis, inactive carrier, compensated cirrhosis, decompensated
cirrhosis, hepatocellular carcinoma, liver transplantation, and death (see Figure 1.b in the
Supplementary Data of Hoerger et al. (2013)). From that model, we estimated the expected
annual cost of chronic hepatitis B infection to be approximately $4,688; the expected HRQoL
utility weight for chronic hepatitis B is approximately 0.939.
With these annual estimates, we are nearly ready to calculate the expected cost of spending
a year in the Infected state. But we need one more parameter, the prevalence of chronic (or
acute) hepatitis B infection in the population, so that we can calculate the proportions of
the Infected population who have acute and chronic hepatitis B infections. Fortunately, data
from the epidemiological literature allow us to calculate such a parameter. Data from 1999 to
2010 from the National Health and Nutrition Examination Survey reported that the relative
risk of chronic hepatitis B infection for diabetes patients in the US was 60% higher than the
risk for individuals without diabetes, with odds ratios (OR) of 1.7 (95% confidence interval
(CI) 1.3–​2.2) for persons aged 18 through 59 years, and 1.3 (95% CI 1.0–​1.6) for persons
aged 60 years and older. The reported prevalence of chronic hepatitis B infection among dia-
betics in the sample was 8.2% (95% CI 6.8–​9.8) (Schillie, Xing, Murphy, & Hu, 2012). We will
use this prevalence estimate in our model. Applying this prevalence and its complement as
weights to the annual costs and HRQoL utility weights of acute and chronic hepatitis B infec-
tion, we estimate an expected cost and HRQoL utility weight of $1,744 (i.e., 0.082 × $4,688
+​(1 − 0.082) × $1,481) and 0.992 (i.e., 0.082 × 0.939 +​(1 − 0.082) × 0.997), respectively, for
spending a year in the Infected state. These are the cost and HRQoL utility weight estimates
that we use in the simplified Markov model that was introduced in Chapter 4.1.

References
Hoerger, T. J., Schillie, S., Wittenborn, J. S., Bradley, C. L., Zhou, F., Byrd, K., & Murphy, T.
V. (2013). Cost-​effectiveness of hepatitis B vaccination in adults with diagnosed diabetes.
Diabetes Care, 36(1), 63–​69. doi:10.2337/​dc12-​0759
Schillie, S. F., Xing, J., Murphy, T. V., & Hu, D. J. (2012). Prevalence of hepatitis B virus infection
among persons with diagnosed diabetes mellitus in the United States, 1999–​2010. Journal of
Viral Hepatitis, 19(9), 674–​676. doi:10.1111/​j.1365-​2893.2012.01616.x
 APPENDIX 6

Making models probabilistic and estimating

the value of information
Ciaran N. Kohli-​Lynch

A.6.1 Making models probabilistic

The core parameters in most decision models relate to costs, utilities, probabilities, and treat-
ment effect sizes. Below is a list of candidate distributions for these data types and informa-
tion on how these distributions are constructed. Non-​normal distributions accept parameters
as inputs which help determine the shape and size of their curves. Using the method of mo-
ments, these can be derived from sample statistics like the mean and standard error of a dataset.
When a random variable has more than one candidate distribution, sensitivity analysis and
goodness-​of-​fit tests can be conducted to determine which distribution is most suitable (Nixon
& Thompson, 2004).

A.6.1.1 Costs
Healthcare costs are typically calculated as the product of resource counts and unit costs.
Uncertainty may exist in both these parameters. The distributions used to represent count
and cost data are both defined by the fact that values less than 0 are impossible. For count
data, a discrete probability distribution which can only assume whole numbers. Cost data
are typically represented by continuous probability distributions that can take on any value
greater than or equal to 0.
The unit cost of a healthcare resource can vary anywhere from 0 to positive infinity. Negative
unit costs are not possible. Cost data also tend to be highly skewed. Unlike the normal distri-
bution, they do not tend to be symmetrically distributed around a peak value. Typically, costs
are very low in a large proportion of observations but the mean is “skewed” higher by a small
number of very high costs in a dataset (Malehi, Pourmotahari, & Angali, 2015).
One candidate distribution to represent cost data is the gamma distribution. This distribu-
tion is constrained on the range 0 to positive infinity and can be highly skewed. If α and β are
parameters which describe the shape and scale of gamma distribution, then the expected value
and variance of the distribution are defined by the following equations:
Equation A.6.1:

EVSI = E  NMBsi ]− E[ NMBci 

Equation A.6.2:

var[ X ] = aβ2
384 Appendix 6

To produce a distribution which fits observed data, we substitute the expected value and
variance of the distribution for the sample mean and variance, respectively. After setting
E( X ) = x and var( X ) = s 2 and rearranging the above equations, the following equations are
derived:
Equation A.6.3:

x2
α=
s2

Equation A.6.4:

s2
β=
x

The gamma distribution may provide a reasonable fit for the seasonal influenza vaccine cost
data. We know that we need a distribution that is bounded by 0 and positive infinity. We have
additional information from our cost study: the arithmetic mean was $3.00 and the standard
error was $2.34. We require a probability distribution that exists on the domain 0 to positive
infinity and maintains these attributes. It is therefore reasonable to assume that the vaccine
x 2 3.002 s 2 2.342
cost data are gamma(α, β) distributed, where a = 2 = 2
≈ 1.68 and β = = ≈ 1.81 .
s 2.34 x 3.00
The blue curve in Fig. 4.3.4 shows a gamma distribution of the vaccine data which evidently
provides a better fit for the data than the red normal distribution.
The log-​normal distribution may also be assigned to costs in probabilistic sensitivity anal-
ysis (PSA). Like the gamma distribution, it is constrained on the interval 0 to positive infinity
and can be skewed. The log-​normal distribution assumes that the logarithm of a random var-
iable is normally distributed. In statistical software it is often parameterized as log-​normal(μ,
s), where μ and s are the mean and standard deviation of the log-​transformed dataset,
respectively.

A.6.1.2 Utilities
The utility associated with a health state (also known as the quality or disability weight) is usu-
ally bounded between 0 and 1. A candidate distribution for data that exist in this domain is the
beta distribution. If α and β are parameters which describe the shape and scale of beta distribu-
tion, then the expected value and variance of the beta distribution are defined by the following
equations:
Equation A.6.5:

α
E[ X ] =
α +β

Equation A.6.6:

aβ
var[ X ] =
(α + β)2 (α + β + 1)
 Appendix 6 385

As above, we set we E[ X ] = x and var[ X ] = s 2 and rearrange the equations. After rearrangement,
we arrive at the following equations:
Equation A.6.7:

 x (1 − x ) 
α=x 2
− 1
 s 

Equation A.6.8:

(1 − x )
β=α
x

Disability-​adjusted life years are always constrained between 0 and 1 (Burstein et al. 2015).
Therefore, the beta distribution should always be considered as a candidate to represent disa-
bility weights. For quality-​adjusted life years (QALYs), almost all health states have values be-
tween 0 and 1. In rare circumstances, health states can have a QALY value less than 0 (Lamers,
2007). A QALY less than 1 represents a health state deemed “worse than death.” In this sit-
uation, we need to use a distribution which is bounded by negative infinity and 1. A simple
workaround involves modelling health state utility as decrements to perfect health. Health state
utility, U, for health state, s, is now calculated using the function: U s = 1 − Ds. The random vari-
able Ds is the deviation of U s from full health. Since it is defined on the domain 0 to infinity, we
can represent it using the gamma and log-​normal distributions.

A.6.1.3 Probabilities
A clear constraint for probabilities is that they can only assume values in the domain 0 to 1. If we
have estimates for the mean and variance of a probability, we can use the methods described for
utilities to define an appropriate beta distribution.
In studies which produce binomial event data (i.e., event or no event), a beta distribution can
be constructed to predict the likelihood of an event occurring. Consider a study with n obser-
vations. Each observation records whether or not an event of interest occurs. If r events occur,
the beta(α, β) distribution with α = r and β = n − r will produce a distribution representative of
the observed data.
An additional constraint should be considered when we model probabilities for multiple
mutually exclusive events. The probabilities of a set of mutually exclusive events must sum
to 1. When there are only two possible events at some point in a decision model, we can
easily code the probability of one event as the probabilistic complement to the other. Hence,
including one event probability as a distribution will inevitably alter rates of the comple-
mentary event. Uncertainty for more than two mutually exclusive events can be represented
with a Dirichlet distribution. The Dirichlet distribution is the multivariate generalization of
beta distribution. Instead of outputting a single value, it accepts k parameters and outputs
k values.
Consider a clinical study of patients at risk of three mutually exclusive events: Event A, Event
B, and No Event. Five hundred patients are observed for the entirety of the study. At the end of
follow-​up, 300 patients have experienced Event A, 150 have experienced Event B, and 50 did
not have an event. The Dirichlet (300,150, 50) distribution describes the range of probabilities
that can be assigned to these events. Its output is a vector of three values which align with each
386 Appendix 6

of the study endpoints. More generally, the Dirichlet (α1 , α 2 ,…, α n ) distribution can be em-
ployed to predict the respective probabilities of n mutually exclusive events.

A.6.1.4 Treatment effects

Treatment effects are a key source of uncertainty in decision modelling studies. It is im-
portant to consider specific constraints on a treatment effect before assigning a distribu-
tion. Some treatments can be modeled through absolute changes in risk factor values. For
example, antihypertensive medications tends to drop blood pressure by a fixed amount
(Ogden, He, Lydick, & Whelton, 2000). As blood pressure can either increase or decrease
after a treatment, positive and negative treatment effects are possible. The treatment ef-
fect exists on the domain negative infinity to positive infinity and could be modeled with a
normal distribution.
Often, treatments lead to a relative reduction in an individual’s likelihood of experiencing an
event. We represent such treatment effects with statistical measures like the hazard ratio, odds
ratio, and relative risk. The lowest possible value for this parameter is 0 which occurs when
no events are observed in the treatment arm of a study. It is also possible that a treatment will
cause a large increase in events. Hence, the domain for this parameter is 0 to positive infinity. It
is commonly assumed that relative measures of event probability are log-​normally distributed
(Barendregt, 2010).

A.6.1.5 Defining distributions in Excel and running

Monte Carlo simulations
Distributions can be defined in many ways. Sometimes they are defined with the mean and var-
iance from a dataset, sometimes with descriptive statistics of transformed data, and sometimes
with shape and scale parameters. The parameters we use to define a distribution depends on the
type of distribution and the statistical software being used. The exercises in this handbook use
a Microsoft Excel model. Hence, the description of distributions and key parameters included
above generally align with the way that distributions are defined in Excel.
When conducting a PSA in Excel, we want to draw a value from each parametric distribution
employed in our decision model. Many distributions are available for use in Excel. We must pull
randomly selected values from these distributions, using inverse distribution functions.
In Fig 4.3.4, we fitted a gamma(1.68,1.81) distribution to approximate the unit cost of
vaccines in South Africa. In Excel, the function GAMMA.DIST ( x , alpha, beta, FALSE ) can
model this distribution if one sets alpha =​1.68 and beta =​1.81. When function type “cu-
mulative” is declared FALSE, the output of this function is the probability density function
and can be described mathematically as GAMMA.DIST (x , alpha, beta, FALSE ) = P ( X = x ).
When cumulative is declared TRUE, the output is the probability that the true value is less
than or equal to x. This is the cumulative density function and can be presented mathemati-
cally as GAMMA.DIST (x , alpha, beta, TRUE ) = P ( X ≤ x ).
When conducting a PSA in Excel, we wish to draw a number randomly from a prede-
fined distribution. To do so we use inverse cumulative density functions. Alongside
parameters like alpha and beta which describe the form of a distribution, inverse cumu-
lative density functions accept a probability, L, as input. They output a value, x, such that
the probability of random variable X is less than or equal to x equals L. Mathematically,
GAMMA.INV (L, alpha, beta) = x ; x : P ( X ≤ x ) = L.
 Appendix 6 387

In Fig. A.6.2, it appears that approximately 50% of the gamma distribution lies between
$2.00 and $3.00. In Excel, the function GAMMA.INV (0.50,1.68,1.81) predicts that the median
cost in the dataset is $2.46. This means that if we were to randomly select an observation from
the gamma(1.68,1.81) distribution, there is a 50% chance it would be less than $2.46.
In the PSA process, we randomly select parameters from a distribution multiple times. In
Excel, the function RAND() randomly selects a value between 0 and 1. Hence, it can be used
in the inverse cumulative density function to randomly select a value from a distribution. For
example, GAMMA.INV(RAND(),1.68,1.81) will select a random value from the vaccine cost
distribution. When repeated multiple times, this process will draw a range of plausible values
from across the entirety of the gamma(1.68,1.81) distribution.
Once distributions have been defined for each parameter in a decision model, a set of parameter
values are randomly selected, the model is run with these parameter values, and outcomes are re-
corded. At this point, this first iteration of the PSA is complete. Parameter values are subsequently
redrawn from their respective distributions and model outcomes are recorded again. This process is
repeated until the total number of desired iterations have been completed. These steps are outlined
below. In Excel, this process can be automated using Visual Basic for Applications (VBA) macros.
Steps to conduct a PSA in Excel for a decision model with random variables X and Y:
1. Select a random value L between 0 and 1 using RAND() function. Example: RAND() selects
0.60 from the range of possible values (Fig A.6.1).
2. Draw value from distribution of X such that P(X ≤x) equals random value drawn in step 1,
using function GAMMAINV(RAND(),alpha,beta). Example: 60th percentile of X’s
gamma(1.68,1.81) distribution is drawn. As shown, this value equals approximately $3.03.
3. Repeat Steps 1–​2 for random variable Y, selecting a random value from its distribution.
4. Compute and record model outcomes with inputs derived from Steps 1–​3.
5. Repeat Steps 1–​4 multiple times, using different values for random variables X and Y each
time to inform model parameters.
6. Average over results from all model iterations to obtain central cost-​effectiveness es-
timate. Range of observed results can be used to determine confidence interval for
estimate.

A.6.2 Value of information analysis

The processes required to estimate expected value of perfect information (EVPI), expected
value of perfect parameter information (EVPPI), and expected value of sample information
(EVSI) are described below. A summary of the subscripts and terms used henceforth to de-
scribe value of information calculations is provided in Table A.6.1.

A.6.2.1 Calculating expected value

of perfect information
Let NMBx , j equal net monetary benefit (NMB) for eventuality x and treatment j. Each eventu-
ality occurs with probability p(x ). The expected NMB for strategy j is calculated by averaging
across all possible eventualities:
Equation A.6.9:

E x [NMB j ] = ∑ p(x ) * NMBx , j

x
388 Appendix 6

1.1
1.0
Relative Frequency of Value Selected by 0.9
0.8
0.7
0.6
RAND()

0.5
0.4
0.3
0.2
0.1
0.0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Fig. A.6.1 Relative frequency of value selected by RAND().

When estimating the expected NMB of a treatment using PSA results, we assume that a
discrete number of eventualities are possible (i.e., x is a discrete random variable). In re-
ality, many sources of parametric uncertainty are continuous random variables, and it
would be mathematically more correct to estimate their expected value using integration.
Here, the process is discretized to help illustrate the processes involved in value of infor-
mation analysis. In addition, most decision models are too complex to easily evaluate using
integration. Monte Carlo simulation uses a discrete approach to simulate the process of
integration.

100%
90%
80%
Probability True Cost ≤ x

70%
60%
50%
40%
30%
20%
10%
0%
0.0 2.0 4.0 6.0 8.0 10.0
Cost (x)

Fig. A.6.2 Probability that the true cost is ≤x.

 Appendix 6 389

To maximize health benefits with current levels of information, a decision maker will
fund the intervention with the highest expected NMB. The expected NMB achieved by a de-
cision maker with current information (ci) can therefore be calculated as follows:
Equation A.6.10:

E[NMBci ] = max j (E x [NMB j ])

A decision maker with perfect information knows which eventuality will occur and will choose
the intervention with the highest NMB in this situation. Based on current information, we
do not know which eventuality will occur; the uncertainty in the decision is not resolved. We
simply know that some theoretical perfectly informed decision maker would have this infor-
mation. The various eventualities still have p(x ) probability of occurring. Hence, the expected
NMB with perfect information (pi) equals the probability-​weighted sum of the maximum
NMB across each eventuality:
Equation A.6.11:

E[NMB pi ] = E x [max j (NMBx , j )] = ∑ p(x ) × max j (NMBx , j )

x

Finally, the expected value of perfect information equals the expected gains that would be
achieved by obtaining perfect information compared to current information. This is the differ-
ence between the expected NMB with perfect information and the expected NMB with current
information:
Equation A.6.12:

EVPI = [ ENMB pi ] − E[NMBci ]

The following steps describe how to calculate EVPI using Monte Carlo simulation with a
decision-​analytic model and predefined probability distributions for parameters:
1. Simulate the “true value” of random variable x by randomly drawing a value, xs, from its
distribution S times.
2. Estimate E[NMBci ] by computing the treatment strategy with maximum E x [NMB j ] aver-
aged across the S simulations of x.
3. Estimate E[NMB pi ] by computing the treatment strategy with maximum NMBx , j in each
of the S simulations of x then averaging across all S results.
4. Calculate EVPI by subtracting E[NMBci ] from E[NMB pi ].

A.6.2.2 Calculating expected value of perfect

parameter information
The EVPPI is calculated in a similar way to EVPI. To calculate EVPPI, we must consider the
range of uncertain parameters in a decision problem. Let us denote the parameter of interest
as xref and the remaining sources of parametric uncertainty as xoth. For simplicity, the following
example assumes that xref and xoth are independent. This means that eliminating uncertainty
regarding one parameter provides no information on the true value of the other. We will also
assume that xref is a single parameter. Neither of these assumptions is fundamental to the esti-
mation approach for EVPPI presented below.
390 Appendix 6

Table A.6.1 Parameters for value of information analysis

Value Definition
ci Current information
E[NMBci ] Expected net monetary benefit with current information

E[NMB pi ] Expected net monetary benefit with perfect information

E[NMB ppi ] Expected net monetary benefit with perfect parameter information

E[NMBsi ] Expected net monetary benefit with sample information

Ex[ . . . ] Expected value across all possible values of x

j Treatment under consideration
maxj(NMBx,j) Value of NMBx,j with for treatment with the highest net monetary benefit
among a set of alternative options under eventuality x
NMBx,j Net monetary benefit of intervention j under eventuality x
p(x) Likelihood that a given value of x is the true state of the world
pi Perfect information
ppi Perfect parameter information
si Sample information
x Eventuality, or state of the world, represented by parameter or set of
parameters in a decision model
xoth Remaining parameter(s) in EVPPI analysis
xref Parameter(s) of interest in EVPPI analysis
z Summary statistics obtained from new sample information

Now, consider a decision maker who has perfect information regarding the true value of xref.
Each possible value of xref occurs with probability p(x ref ). Let NMBxref ,xoth , j equal the NMB for
treatment j with parameters xref and xoth. To estimate the expected NMB for strategy j when we
know the value of xref, we must average over all remaining sources of decision uncertainty (see
Equation A.6.13).
Equation A.6.13:

E xoth [NMB j | x ref ] = ∑ p(xoth ) * NMBxoth ,xref , j

xoth

The decision maker with perfect parameter information will know the true value of xref. At this
value, they will choose to implement the treatment with the highest value of E xoth [NMB j | x ref ].
With current information, the true value of x ref is unknown. To calculate the expected NMB
given perfect parameter information, we must average across all possible values of xref (see
Equation A.6.14).
 Appendix 6 391

Equation A.6.14:

E[NMB ppi ] = E xref [max j (E xoth [NMB j | xref ])] = ∑ p(xref ) × max j (E xoth [NMB j | x ref ])
xref

To calculate EVPPI, we subtract the expected NMB given current information from the ex-
pected NMB with perfect parameter information (see Equation A.6.15).
Equation A.6.15:

EVPPI = E[NMB ppi ][− E[NMBci ]

When estimating E[NMB ppi ] with a decision model, it is necessary to conduct a PSA with two
loops. The outer loop randomly samples the parameter of interest, xref. The inner loop samples
all remaining parameters, xoth. An additional PSA is required to estimate E[NMBci ].
The following steps describe how to calculate EVPPI using Monte Carlo simulation with a
decision-​analytic model and predefined probability distributions for parameters:
1. Simulate a set of random variables x = {x ref , x oth } by randomly drawing two values,
x s = {x ref ,s , xoth ,s }, from their distributions S times.
2. Estimate E[NMBci ] by computing the treatment strategy with maximum E x [NMB j ]
averaged across the S simulations of x.
3. Outer loop: simulate xref by randomly drawing a value, xref,o, from its distribution O times.
Inner loop: for each value of x ref ,o = {x ref ,1 ,…, xref ,O }, simulate xoth by randomly drawing a
value, xoth,i, from its distribution I times.
a. . . . compute NMBxoth ,i ,xref ,o , j for each combination of value values xref,o and xoth,i.
b. . . . estimate E xoth [NMB j | x ref ,o ] for each of xref,o by averaging NMBx ,x , j across the
oth ,i ref ,o
I simulations of xoth.
4. Estimate E[NMB ppi ] by recording the treatment strategy with maximum E xoth [NMB j | x ref ,o ]
in each of the O simulations of xoth and averaging across all O results.
5. Calculate EVPPI by subtracting E[NMBci ] from E[NMB ppi ].
It may be desirable to relax the assumptions required for the above framework. In some
cases, we cannot assume independence between xref and xoth. Without this assumption we
must account for the conditional distribution of xoth dependent on xref (Brennan, Kharroubi,
O’Hagan, & Chilcott, 2007). Theoretically this involves replacing p(x oth ) with a probability
distribution which conditions xoth on xref. Similarly, the simulation of xoth in Step 3a of the
Monte Carlo simulation would be drawn from a conditional probability distribution of xoth
dependent on xref,o. We may also wish to estimate the value of reducing uncertainty in mul-
tiple parameters concurrently. In this situation,x ref can be replaced by a vector of parameters,
with x oth representing all remaining sources of parametric uncertainty in the decision.
The extensive process required to estimate EVPPI with Monte Carlo simulation can be
computationally demanding. Dependent on the complexity of a decision model, traditional
PSAs may take multiple days to complete. As technology develops, it will be possible for most
researchers to quickly run these analyses. In the meantime, computational limitations will
continue to inhibit the widespread use of EVPPI analysis. Algorithms have been developed
which reduce the computational burden of EVPPI analysis (Brennan et al., 2007; Sadatsafavi,
Bansback, Zafari, Najafzadeh, & Marra, 2013).
392 Appendix 6

A.6.2.3 Calculating expected value of sample information

Any future research will be conducted in a sample of size n, which predicts a new state of the
world z ∈ Z, where z represents summary statistics taken from the new sample and Z repre-
sents the set of all possible samples of size n in the population of interest. If the population of
N!
interest is of size N, there are such samples of size n. This new sample enables us
(N − n)!n !
to update our prior belief, p(x ), for each value x. The updated or joint posterior distribution,
p(x | z ), assigns a likelihood to each eventuality x influenced by both our prior belief and the
sample information provided by z. The degree to which sample information updates our prior
belief depends on the type of data being observed, the effect size, and the variability of the ob-
served effect in the sample.
Expected NMB with a new sample z is constructed in a similar fashion to E[NMBci ], with be-
liefs about the true value of x updated based on the information provided by z.
Equation A.6.16:

 
E[NMB | z ] = max j (E x [NMBx|z , j ]) = max j  ∑ p ( x | z ) × NMBx , j 
 x 

Some samples will produce estimates of p(x ) similar to our prior belief and some will vary.
Before conducting a study, we do not know which sample will be observed. Therefore, we must
average across all possible samples when estimating the expected NMB with sample informa-
tion. This is similar to the estimation approach required for EVPPI, where the true value of xref
is unknown.
Equation A.6.17:

E[NMBsi ] = ∑ p(z ) × E[NMB | z ]

z

To estimate Equation A.6.17, we need information on the distribution of z. This distribu-

tion is defined by the relationship p(z , x ) = p(x ) × p(z | x ). The function p(x ) represents our
prior belief about the value of x. The function p(z | x ) is called the sampling distribution of the
data. The structure of this function is determined by the type of data being collected (Kunst
et al., 2019).
Finally, we calculate EVSI by subtracting expected NMB given current information from ex-
pected NMB with sample information.
Equation A.6.18:

EVSI = E[NMBsi ] − E[NMBci ]

The following steps describe how to calculate EVSI using Monte Carlo simulation with a
decision-​analytic model and predefined probability distributions for parameters:
1. Outer loop: simulate the “true value” of random variable x by randomly drawing a value,
xs, from its distribution S times. Inner loop: for each value x s = {x1 ,…, x S } , simulate a
sample of n estimates z s = {x s ,1 ,…, x s ,n }. The sampling distribution of zs,n will depend on
the structure of the parameter being analyzed and the sample size.
a. . . . estimate E x [NMBx|z , j ] for each sample zs. This will involve updating the base esti-
mate of x with the sample information provided by zs.
b. . . . record E[NMB | z s ], the maximum E x|z s [NMB j ] for each sample zs.
 Appendix 6 393

2. Estimate E[NMBci ] by computing the maximum E x [NMB j ], averaged across the S simula-
tions of x.
3. Estimate E[NMBsi ] by computing the average value of E[NMB | z s ] across the S
simulations.
4. Calculate EVSI by subtracting E[NMBci ] from E[NMBsi ].
The framework above assumes we are only interested in one value of n. If we are only con-
sidering one sample size, a proposed study which recruits n patients and costs less than EVSI
should be funded. To estimate the optimal sample size for a study, EVSI could be estimated
for multiple different values of n. Intuitively, as n increases, we get closer to sampling the en-
tire population and sample information tends towards perfect information. Hence, the upper
bound for EVSI is EVPI.
For illustrative purposes, the example above did not disaggregate parameters into those
being studied and those which will maintain existing levels of uncertainty. As with EVPPI, the
framework can be extended to distinguish between sources of parametric uncertainty that will
be studied and others which will remain unexamined. When a proposed study will focus on
a limited set of model parameters, the upper bound for its EVSI is the corresponding EVPPI.

References
Barendregt, J. J. (2010). The effect size in uncertainty analysis. Value in Health, 13(4), 388–​391.
doi:10.1111/​j.1524-​4733.2009.00686.x
Brennan, A., Kharroubi, S., O’Hagan, A., & Chilcott, J. (2007). Calculating partial expected
value of perfect information via Monte Carlo sampling algorithms. Medical Decision Making,
27(4), 448–​470. doi:10.1177/​0272989X07302555
Burstein, Roy, Tom Fleming, Juanita Haagsma, Joshua A. Salomon, Theo Vos, and Christopher
JL. Murray. (2015). Estimating distributions of health state severity for the global burden
of disease study. Population Health Metrics, 13(1), 31. https://​doi.org/​10.1186/​s12​
963-​015-​0064-​y
Kunst, N., Wilson, E., Alarid-​Escudero, F., Baio, G., Brennan, A., Fairley, M., . . . Heath, A.
(2019). Computing the expected value of sample information efficiently: Expertise and skills
required for four model-​ based methods. ArXiv. https://​arxiv.org/​ftp/​arxiv/​pap​ers/​1910/​
1910.03368.pdf
Lamers, L. M. (2007). The transformation of utilities for health states worse than
death: Consequences for the estimation of EQ-​5D value sets. Medical Care, 45(3), 238–​244.
doi:10.1097/​01.mlr.0000252166.76255.68
Malehi, A. S., Pourmotahari, F., & Angali, K. A. (2015). Statistical models for the anal-
ysis of skewed healthcare cost data: A simulation study. Health Economics Reviews, 5, 11.
doi:10.1186/​s13561-​015-​0045-​7
Nixon, R. M., & Thompson, S. G. (2004). Parametric modelling of cost data in medical studies.
Statistics in Medicine, 23(8), 1311–​1331. doi:10.1002/​sim.1744
Ogden, L. G., He, J., Lydick, E., & Whelton, P. K. (2000). Long-​term absolute benefit of
lowering blood pressure in hypertensive patients according to the JNC VI risk stratification.
Hypertension, 35(2), 539–​543. doi:10.1161/​01.HYP.35.2.539
Sadatsafavi, M., Bansback, N., Zafari, Z., Najafzadeh, M., & Marra, C. (2013). Need for
speed: An efficient algorithm for calculation of single-​parameter expected value of partial
perfect information. Value in Health, 16(2), 438–​448. doi:10.1016/​j.jval.2012.10.018
 APPENDIX 7

Decision model

HepB_CEA-Tree_PartA

HepB_CEA-Tree_PartB
396 Appendix 7

HepB_CEA-Tree_PartC

HepB_CEA-Tree_PartD

HepB_CEA-Tree_PartE
Index

For the benefit of digital users, indexed terms that span two pages (e.g., 52–​53) may, on occasion,
appear on only one of those pages.
Tables, figures, and boxes are indicated by t, f, and b following the page number

absorbing Markov chain 267 allocation key 109–10

accelerated transition 41t ambiguity aversion 55–56
Gavi 358 annualization, data analysis 116–17
access, immunization program annualization factor 117, 118b
financing 336 calculation 117
Access to COVID Tools Accelerator annualized cost calculation 117
(ACT) 17 annual vaccination rate
accountability, immunization program calculation 184–85
financing 336 definition 174–75
activity-based costing 101 anthrax, vector control 160
acute infections 274 antibiotics, reduced use 51–52
added value of vaccines 137–38 anti-vaxxers (vaccination refusal) 54–55,
adjuvants vii 56–57
administrative records, primary costing antiviral medications, COVID-19
study 106–7 pandemic 313–14
adoption question, economic ASEAN region, pooled procurement 39–40
evaluations 156–58 Asia, immunization program financing 333,
adults, immunization programs 9 342–43
Advance Market Commitment (AMC) 353 asset useful life 117
Gavi contributions 354t assumptions, documentation 122f, 122
Pneumococcal Advance Market attitude roots literature 56
Commitment (Pneumococcal
AMC) 42 Bacillus–Calmette–Guérin (BCG)
advance market commitments vaccine 20t, 361
immunization program lung cancer effects 51–52
financing 360–61 Bangladesh, immunization program
vaccine procurement 42–43 financing 342
Advisory Committee on Immunization base-case analysis, COVID-19
Practices (ACIP) 9–10 pandemic 313–14, 320–22
Africa, immunization program Bayesian multivariate probabilistic sensitivity
financing 342–43 analysis 319
aged adults see elderly benchmark development, costing
age groups, COVID-19 Markov model 324 studies 86–87
agent-based models 286 benefit-cost ratio (BCR) 242
disease transmission modeling 287 calculation 149
aggregate outcomes 58 beta distribution 194
Airtel 363 calculation 194
alcohol taxes 345–46 BIA see budget impact analysis (BIA)
398 Index

bias central limit theorem, probabilistic

cost-effectiveness analysis 209 sensitivity analysis 300
definition 170–71 Centre for Global Development 36–37
bilateral donors, Gavi 354–55 CEPI see Coalition for Epidemic
Bill & Melinda Gates Foundation Preparedness Innovation (CEPI)
(BMGF) 43 Cervical Cancer Prevention and Control
Advance Market Commitment 361 Costing Tool (WHO) 89–90, 90t
Gavi 353–54, 354t CET see cost-effectiveness thresholds (CET)
Healthy Markets Framework 35 CFA (cost-finding analysis) 147
push funding 43–44 chance nodes, definition 246
biotechnology firms, vaccine supply 34 CHEERS (Consolidated Health Economic
Bolivia, immunization program Evaluation Reporting Standards) 233,
financing 342–43 234
budget impact analysis (BIA) 144, 146t, 147, child health service financing 344
150–51, 239–41 Children’s Vaccine Initiative 351–52
calculation 239 China, Developing Countries Vaccine
decision tree modeling 256–57 Manufacturing Network 36
payment definitions 239 cholera vaccine 6
see also cost-effectiveness analysis (CEA) oral 43–44
budgeting CholTool, International Vaccine
budget impact analysis 240 Institute 89–90
immunization programs 85–86, 336–37 Choosing Interventions that are Cost-
Building Blocks Framework (WHO) 18–19 Effectiveness (CHOICE) program
buildings, economic vs. financial costs 99t (WHO) 86–87
burden of illness analysis see cost of illness chronic hepatitis B infection 175
(COI) chronic illness, COVID-19 Markov
model 323–24
calibration estimators 119 chronic infections 274
Cambodia, immunization disparities 69–70 CIA see cost-identification analysis (CIA)
Canada CIs see confidence intervals (CIs)
Advance Market Commitment 361 citation bias 170–71
Gavi contributions 354t Clinical Institute for Clinical and Economic
capital investments, new and underutilized Review (US) 231
vaccine introduction 129 Clostridium difficile infection 51–52
case-control studies 182 CMA (cost-minimization analysis) 144, 145,
odds ratio 177 146t, 147–48
case series 182 cMYP see Comprehensive Multi-Year Plan
causal study design, vaccine (cMYP)
effectiveness 203 Coalition for Epidemic Preparedness
CCA (cost-consequence analysis) 144, 146t, Innovation (CEPI) 17
147, 148 COVAX AMV 361
CEA see cost-effectiveness analysis (CEA) co-financing policy, Gavi 357–58
Center for Global Development 360–61 cohort life tables 187–88
Centers for Disease Control and Prevention cohort studies 177, 182
(US) 314–19 cohort transition, Markov models 271–76
National Vital Statistics System 187 COI (cost of illness) 144, 146t, 147, 148
13-valentpneumococcal conjugate cold chain viii, 5
vaccine 69–70 costs 97, 100
vaccine schedules 11 economic vs. financial costs 99t
central externalities 54 expansion of 89, 128
 Index 399

Gavi 355, 363 Markov models 266

immunization costs 71–73 uncertainty 121–22
maintenance costs 100t cost-benefit analysis (CBA) 47–48, 85, 87,
maintenance of 12 141–42, 144, 146t, 147, 148–49
management 127, 131t cost-effectiveness analysis vs. 87
new and underutilized vaccine cost-utility analysis vs. 60–61, 62
introduction 126t, 127, 128, 131t cost-comparison analysis 144, 146t
quality effects 70 cost-consequence analysis (CCA) 144, 146t,
vaccine loss 283 147, 148
cold chain equipment cost-effectiveness 140–55
economic vs. financial costs 99t COVID-19 Markov model 320–22
new and underutilized vaccine probabilistic sensitivity analysis 298–300
introduction 129 cost-effectiveness acceptability curves
Collaborative Procedure for Accelerated COVID-19 Markov model 323f
Registration (WHO) 38 probabilistic sensitivity analysis 303, 304f
combination vaccines 126–27, 126t, 129 cost-effectiveness analysis (CEA) 85, 87,
Comic Relief, Gavi 353–54 89, 141–42, 144, 145, 146t, 149–50,
commercialization, vaccine production 156–66
costs 29 comparators in 161
Commission on Macroeconomics and cost-benefit analysis vs. 87
Health 217–18 COVID-19 Markov model 321t
common diseases 7 decision tree modeling 251–54
community-level health workers 126t funding for 233
Community of Practice of immunization impact inventory 233–34
economics 88 Markov models 271–76
comparative statistics 181 process of 216–17
comparators, decision tree modeling 248 reference case 233–34
competition, vaccine supply 35 reimbursement decisions 143
competitive pricing 32–33 reporting 213
Comprehensive Multi-Year Plan (cMYP) 90t return on investment analysis vs. 241
data on immunization expenditure 338 scatter plots 297f, 302–3
new and underutilized vaccine structural uncertainty 292
introduction 130 wide use of 147
confidence intervals (CIs) 122 see also budget impact analysis (BIA)
calculations 178–79 cost-effectiveness analysis (CEA)
parameter estimation 194 plane 213–16
parametric uncertainty 293–94 alternative approaches 219
probabilistic sensitivity analysis 302 definition 213
confirmational bias 56 quadrants 214–15
conflicts of interest 233 results table 214f, 214t
Consensus Statement 90t, 91–92 see also incremental cost-effectiveness
Consolidated Health Economic ratio (ICER)
Evaluation Reporting Standards cost-effectiveness thresholds (CET) 152,
(CHEERS) 233, 234 216–20
consumables 29 definition 217
consumption level 49–50 equity issues 231
contacts 285 estimates 218
contract matrices 284–85 factor represented 217–18
cost analysis identification of 218–19, 220t
data aggregation 120 inequality and 221b, 223b
400 Index

cost-effectiveness thresholds (CET) (cont.) start-up costs see start-up costs

quantitative approaches 219 structures 30–32
vaccines and 219–20 total costs 253
cost-finding analysis (CFA) 147 treatment cost 322–23
cost-identification analysis (CIA) 147 unit cost parameters 191–92
quality-adjusted life years 302 unit cost per dose 96
costing studies 85–94 unit costs 190
available resources 88–92 vaccine production 29–30
benchmark development 86–87 value vs. 144
economic evaluations 87–88 variable costs 30
societal perspective 97–98 variation in vaccine delivery 86
cost-minimization analysis (CMA) 144, 145, wastage costs 12, 12t
146t, 147–48 cost-utility analysis (CUA) 47–48, 141–42,
cost modeling exercises, primary costing 144, 145, 146t, 150, 288
studies vs. 83 cost-benefit analysis vs. 60–61, 62
cost of illness (COI) 144, 146t, 147, 148 COSTVAC
cost of input calculation 118 Pan American Health organization 89
cost profile 118 Pan American Health Organization 90t
costs WHO 89–90
annualized calculation 117 cosy comparison analysis 145
decision tree modeling 250 country-level studies, vaccination
direct labour costs 29 policies 346
disease treatment 51 country-specific regulatory approval 31–32
economic see economic costs COVAX AMV 361
estimated costs 320 coverage, vaccine delivery 71
estimation 86 COVID-19 Markov model 310–27, 313f, 315t
evaluation in data analysis 118 cost-effectiveness 320–22, 321t
financial see financial costs costs 314
fiscal costs see fiscal costs estimated costs 320
fixed see fixed costs health utilities 319
immunization see immunization costs limitations 324
incremental costs of new and mitigation strategies 313–14
underutilized vaccine 125 model structure 312–13
indirect see indirect costs probabilities 314–19
input cost 68 sensitivity analysis 319, 322f, 322–23
interest costs 29 study design 312
investment costs 96–97 treatment/vaccine effects 320
labor costs 98 see also Susceptible–Exposed–Infected–
logistic costs 128 Recovered (SEIR) model
marginal costs 32 COVID-19 pandemic ix, 50–51, 157, 160
Markov models 271 base-case analysis 313–14, 320–22
measurement, value equations 144 Gavi 360
mixed costs 30 relative risk of symptoms 179t, 180
ongoing costs 127–28, 129–30 vaccination status and 179–80, 179t
opportunity costs 153 COVID-19 Vaccine Global Access (COVAX)
outcome of Markov models 268 Facility 17, 42–43
production costs 31 COVID-19 vaccines/vaccination vii, 3, 314
program costs 239–40 costs & development 16
retrospective costs 91–92 COVID-19 Markov model 320
semi-variable costs 30 international effort 8
 Index 401

Johnson & Johnson 13 decision node definition 246

nucleic aid vaccines 6–7 decision tree modeling 181, 246–57
priority grouping 17 budget impact analysis 256
programs 323–24, 325 budget impact interpretation 256–57
value assessment 169 case study 247–49, 248f
whole pathogen vaccines 6 conceptual framework 247–48
Cox proportional hazard curves 188 cost-effectiveness analysis 251–54
cross-country variability, inequality data extraction 250–51
and 221b–22 decision model 248–49
CUA see cost-utility analysis (CUA) incremental cost-effectiveness
cumulative incidence ratio 254–56
definition 175–76 parameter estimation 249
literature data 183 see also Susceptible–Infected–Recovered
cycle length, Markov models 267 (SIR) model
demand curve 22–23, 23f
DALYs see disability-adjusted life years (DALYs) demand-side, cost-effectiveness
Darrow–Lind principle 298 analysis 217–18
data Demographic and Health Surveys,
aggregation in cost analysis 120 Ethiopia 208–9
applicability limitations 166 dengue vaccine 6
appropriateness, parameter estimation 182 Department for International Development
immunization financing 337–38 (UK) 354–55
management costs 100t Department of Finance (DOF) 59
modeling role 165–66 Department of Health (DOH)
parameter estimation 169–74 health payer perspective as 61
primary costing study 106–9 investments and 59
quality, parameter estimation 182 Department of Health and Human
validation requirements 287–88 Services 230–31
data analysis 115–23 derivation tools, parameter
annualization 116–17 estimation 167–81
cost evaluation 118 desired outputs, input cost vs. 68
definition 115 deterministic analysis 295–96, 297f
discounting 116 deterministic models 281–83, 282t
parameter estimation 183–89 see also Susceptible–Infected–Recovered
sample weights 120 (SIR) model
data collection Deutsche Post DHL Group 362
decision tree modeling 250–51 Developing Countries Vaccine Manufacturing
primary costing studies 106–10 Network (DCVMN) 34, 36
DataVerse 90t, 92 Developing Countries Vaccine
DCVMN (Developing Countries Vaccine Manufacturing Network (India) 36
Manufacturing Network) 34, 36 diaries 109
death rates 187–88 diphtheria, eradication target 351
decision-analytic models 140–55, 165, 190, diphtheria–tetanus–pertussis (DTwP)
288, 290 vaccine 332
probabilistic sensitivity analysis 299–300 Malaysia 71, 72b
decision making, expected value of perfect new and underutilized vaccine
information 305 introduction 127
decision models prices 20t
decision tree modeling 248–49 diphtheria vaccine 7
probabilistic sensitivity analysis 294 subunit vaccines 6
402 Index

direct labour costs 29 evaluation 224, 225t

direct observation, study data 108–9 incremental cost-effectiveness
disability-adjusted life years (DALYs) 60, 213 ratio 224–25
averting of 217–18 donor funding, Gavi 353–54
cost analysis 87 DTwP see diphtheria–tetanus–pertussis
decision tree modeling 249, 255 (DTwP) vaccine
disability weights 204 Dunning–Kruger effect 56
economic evaluation 204, 291 duplication bias 170–71
health outcome measure 145, 198 dynamic modeling 279–89
incremental costs 217 appropriateness 279–80
low- and middle-income countries 204–5 decision flow chart 281f
theoretical assumptions 208b definition 280
vaccine benefits 140, 205, 206f static models vs. 280
value judgements 205–8 stochastic models 286
disability weights 204 studies and 288
discounting vaccination models 288
calculation 116
data analysis 116 Ebola vaccine 6
discrete-event simulation models 286–87 ecological effects 141
disease eradication 158–59 economic costs 96
disease incidence 287 COVID-19 infection 311
disease surveillance financial costs vs. 99t
new and underutilized vaccine new and underutilized vaccine
introduction 126t introduction 126
new and underutilized vaccines 130 economic evaluations 137–39, 156–62
disease transmission modeling 280–87 adoption question 156–58
agent-based models 287 advanced methods 261–63
classification of 282t costing studies 87–88
contract matrices 284–85 definition 140
deterministic models 281–83 importance of 141–43
discrete event simulation 286–87 investment decision 158
dynamic stochastic models 286 limitations 230–32
SIRV models see Susceptible–Infected– quality of analysis reporting 232–34
Recovered (SIR) model; Susceptible– reporting/interpretation 213–38, see also
Infected–Recovered–Vaccination cost-effectiveness analysis (CEA) plane
(SIRV) model types of 146t
stochastic models 285–86 uncertainty in 291–94
distribution effectiveness parameters 190, 192–93
cost-effectiveness analysis 231 efficiency
equity of 57 immunization program financing 337
rates and 176 vaccine delivery 68
resources needed 143–44 elderly
disutility 193 COVID-19 Markov model 323–24
DOF (Department of Finance) 59 influenza vaccines 53–54
DOH see Department of Health (DOH) routine immunizations 7
domestic resources, inequality and 222b–23 eligibility, Gavi 356–57
domestic suppliers emerging markets 34
high-income counties 38–39 empirical data, parameter estimation 169–70
middle-income counties 38–39 endpoints, decision tree modeling 248
dominance entry barriers 32, 35
 Index 403

EPI (Expanded Program on expected value of perfect parameter

Immunization) 7, 336, 351 information (EVPPI) 304
epidemiological modeling 285 EVPI vs. 307–8
see also dynamic modeling; static modeling expected value of sample information
EQ-5D see EuroQol five dimensional (EQ- (EVSI) 304
5D) questionnaire extended cost-effectiveness analysis
EQ-5DY (EuroQol five dimensional methods 209
questionnaire for children) 201–2 extended dominance, incremental cost-
equipment effectiveness ratio 224–25
cold chain see cold chain equipment extended partners 353
repair 98 external financing, immunization
vaccine production costs 29 programs 347
equity 141 externalities 54
definition 69–70 external validity 182
health outcome measurement 208–10
promotion by vaccines 209 facilities
vaccine delivery 69–70 cost evaluation 119–20
Escherichia coli infection 51–52 sampling 112–13
estimated costs, COVID-19 Markov scale/right sizing 37
model 320 vaccine production costs 29
Ethiopia, Demographic and Health failure classification 190–91
Surveys 208–9 FDA (Food and Drug Administration) 21
ethnic minorities, COVID-19 Markov finances
model 323–24 records 130
European Center for Disease Prevention and security 141
Control 11 vaccine development 19–21
EuroQol five dimensional (EQ-5D) financial costs 96, 97
questionnaire 192–93, 201–2, 203b, economic costs vs. 99t
203–4, 274–75 financial sustainability
COVID-19 Markov model 319 Gavi 356
EuroQol Group 201–2, 203b vaccine delivery 68–69
EuroQol five dimensional questionnaire for fiscal costs 96, 97
children (EQ-5DY) 201–2 new and underutilized vaccine
evaluation, Markov models 268 introduction 125
EVPI see expected value of perfect fixed costs 31
information (EVPI) allocation of 32
EVPPI see expected value of perfect vaccine research and development
parameter information (EVPPI) costs 30
EVSI (expected value of sample fixed-effect meta-analysis 171
information) 304 follow-up life tables 187–88
Expanded Program on Immunization Food and Drug Administration (FDA) 21
(EPI) 7, 336, 351 for-profit investments 62–63
expected value of perfect information free cost (zero cost of vaccines) 22–24
(EVPI) 304, 306f freeze-dried vaccines 129
definition 305 fuel, economic vs. financial costs 99t
estimation of 306 full economic evaluation 146t
expected information value 306f, 307 fully self-financing 41t
expected value of perfect parameter
information vs. 307–8 gamma distribution 191–92
research funding 307 gap measures 204
404 Index

Gavi 9–10, 351–55 Global Health Expenditure Database

bilateral donors 354–55 (GHED) 337
classification & support eligibility 41t GNI (gross national income), Gavi and 334,
co-financing policy 357f, 357–58 356–57
country support 356t gonorrhea treatment 159
COVAX AMV 361 government investments 57–62
COVID-19 pandemic 356, 360 immunization program financing 333f,
COVID-19 vaccine allocation 17 333, 341–46
donor funding 353–54, 354t governments 58
eligibility & transition policy 356–57 gross domestic product (GDP) 137–38
financial sustainability 356 Commission on Macroeconomics and
foundation of 352 Health and 217–18
government of 352 gross national income (GNI), Gavi and 334,
Healthy Markets Framework 35 356–57
immunization program financing 333,
334, 337–38, 347, 360–61 Haemophilus influenzae infection
INFUSE platform 363 contraction risk 168
LMIC vaccination spending 334 transmissibility 280
middle-income countries 360 Haemophilus influenzae type b vaccine
name changes 352 benefits of 149
new and underutilized vaccine cost-effectiveness 168
introduction 124, 130 DPT Hib lyophilized vaccine 20t
origins and structure 351–53 DTaP/IPV/Hib (hexavalent acellular)
post-transition engagement 359–60 vaccine 20t
procurement 38 preference-based measures 200b
support types 355–56 private clinic delivery 74
UPS Foundation 362 subunit vaccines 6
vaccine procurement 41 half-cycle corrections, Markov models 276
Gavi transition 358–60 HALY metrics 203
challenges 359 hantavirus vector control 160
planning 358 hazard ratio definition 181t
GDP see gross domestic product (GDP) health capital 48–51
geographical areas investment in 49
sampling 112 healthcare sector
unit cost per dose 96 costing studies 104
Germany, Gavi contributions 354t COVID-19 Markov model 319
Ghana immunization program impact 163
mixed delivery 75 pooling and allocation 341
National Health Insurance Scheme 68–69, health economic models, input
345–46 parameters 181
Zipline drone network 362–63 health outcome measurement 145, 198–212
GHED (Global Health Expenditure equity and 208–10
Database) 337 Markov models 268
GlaxoSmithKline, pricing strategies 33 preferences 198–202, 200b, 201b
Global Alliance for Vaccines and health payer perspective 61
Immunization see Gavi health-related quality of life (HRQoL) 192–
global donors 36 93, 198, 274–75
Global Health Costing Consortium 88–89, health attribute focus 199
105 Markov models 264
resource guides 90t patient preferences 199
 Index 405

health states HSIS (health system and immunization

Markov models 265–66 strengthening) support 355
probabilities calculation 251 human capital 47–66
health system and immunization health as 48–51
strengthening (HSIS) support 355 vaccinations as 48–51
health taxes, immunization human papillomavirus (HPV)
programs 345–46 vaccine 359–60
Health Utilities Index 201–2 costs 142
health utility/survival, decision tree Merck & Co. Inc. 34
modeling 250, 254 pooled procurement 40
Healthy Markets Framework 35 price negotiations 143
heat-stable vaccines 351–52 prioritization of 142
hepatitis B 271, 272t procurement 41
hepatitis B vaccine start-up costs 126t
budget impact analysis 240 subunit vaccines 6
decision tree modeling see decision tree
modeling ICER see incremental cost-effectiveness ratio
high-quality randomized controlled (ICER)
trials 182 IDCC (Immunization Delivery Cost
return on investment analysis 243 Compendium) 90t, 92
subunit vaccines 6 ideal length of life 204
whole pathogen vaccines 6 IFFIm see International Finance Facility for
herd immunity 22, 25, 54, 158–59, 341 Immunization (IFFIm)
benefits of 22, 311 immune response 18–21
definition vii, 156 triggering of 5
hepatitis B 247, 251 Immunization Agenda 2030 – A Global
immunization programs 5 Strategy to leave No One Behind 335
individual infection as 50 immunization costs 83–84, 95–102
SIR framework 249 categories of 96–98
social benefits 4 definitions 95–98
static immunization models 279–80 studies 110–13
heterogeneity Immunization Delivery Cost Compendium
definition 226–27 (IDCC) 90t, 92
uncertainty vs. 226–27 immunization program financing 332–39,
HICs see high-income countries (HICs) 340–50
high-income countries (HICs) advance market commitments 360–61
domestic suppliers 38–39 budgetary space improvements 347–48
needs of 35–36 costs 98–101, 99t
quality-adjusted life years 204–5 data 337–38
high-quality randomized controlled trials 182 data comparison 338
historical prices 98 donor architecture 351–64
HIV economic benefit analysis 241–42
infection treatment 159 external financing 347
vaccine 19–20 private financing 346–47
hospitalization, pneumonia 51–52 program planning and budgeting 336–37
household security 141 resource tracking 335–36
HPV see human papillomavirus (HPV) return on investment analysis 241
vaccine scale-up costs 335
HRQoL see health-related quality of life sources 332–33, 333f, 340–47
(HRQoL) spending patterns 333f, 334f
406 Index

immunization programs 5, 7–12, 157 individual patient data 170

coverage vs. disease incidence 337 Indonesia
delivery vehicles 11–12 local immunization programs 343
finances see immunization program mixed delivery 75–76
financing infants, immunization programs 9
impact perspective 162–64 infected by disease (I) 283
national, deviation from 74–75 see also Susceptible–Infected–Recovered
schedules 10–11 (SIR) model; Susceptible–Infected–
strategic planning and budgeting 85–86 Recovered–Vaccination (SIRV) model
target populations 8–9 infection risk, unvaccinated population 185
time horizon 164–65 influence/state transition diagram, Markov
vaccine selection 9–10 models 268
value assessment 265 influenza 277
see also national immunization program influenza vaccines
(NIP); National Immunization aged adults 53–54
Strategy (NIS) inequality and 221b–22
Immunization Schedules for Africa 11 older adults 9
impact inventory probabilistic sensitivity analysis
cost-effectiveness analysis 233–34 model 295, 296f, 297f
immunization program financing 337 treatment 159
impact perspective, immunization whole pathogen vaccines 6
programs 162–64 information analysis value see value of
inactivated polio vaccine (IPV) 20t information analysis
incidence information role 21
definition 174–75 INFUSE platform 363
literature data 183 initial self-financing 41t
incremental cost-effectiveness ratio injection supplies, economic vs. financial
(ICER) 149–50, 151–52 costs 99t
calculation of 223–25, 274 input cost 68
COVID-19 Markov model 319 institute for Health Metrics and Evaluation,
decision tree modeling 254–56 COVID-19 Markov model 314–19
definition 213, 230 instrumental value 47, 50
dominance 224–25 insurance claim data 107
extended dominance 224–25 interest costs 29
interpretation 228–30 intergenerational benefits 53
most effective strategy 228–29 internal validity 182
strategy comparisons 223 International Decision Support Initiative 121
strategy ranking 229–30 reference cases 233
vaccine programs and 223 International Federation of Pharmaceutical
value as 151 Wholesalers 353–54
willingness-to-pay threshold 152 International Finance Facility for
incremental costs, new and underutilized Immunization (IFFIm) 353
vaccines 125 Gavi contributions 354t
incremental intervention value 160 vaccine bonds 361–62
India, Developing Countries Vaccine International Society for Health Economics
Manufacturing Network 36 and Outcomes Research (ISPOR) 226
indirect costs 209 International Vaccine Institute
allocation of 32 CholTool 89–90
indirect treatment comparisons 173–74 partnerships 43–44
individual decision-making 53–57 push funding 43–44
 Index 407

interquartile ranges, parameter limitations, cost-effectiveness analysis 209

estimation 194 limited effectiveness, vaccines 50
intrinsic value 47, 50 literature data, incidence rates/cumulative
investment costs 96–97 incidence 183
investment decisions, economic literature data sources, parameter
evaluations 158 estimation 174–81
IPV (inactivated polio vaccine) 20t LMICs see low- and middle-income
isolation, disease eradication 160 countries (LMICs)
ISPOR (International Society for local government health officers 343
Health Economics and Outcomes location bias 170–71
Research) 226 location-specific case-count 287
Italy, Advance Market Commitment 361 logistic costs 128
log-normal distributions 178–79
Japan long-lasting benefits 165–66
human papillomavirus vaccine lotteries, immunization program
introduction 10 financing 347
International Cooperation low- and middle-income countries
Agency 354–55 (LMICs)
Japanese encephalitis 280 cost-effectiveness analysis study
vaccine 6 reviews 219
Jenner, Edward vii cost-effectiveness thresholds 219, 220
Johns Hopkins Coronavirus Resource delivery fund allocation 68–69
Center 314–19 Developing Countries Vaccine
Johnson & Johnson, COVID-19 vaccines 13 Manufacturing Network 36
Joint Commission on Vaccination and disability-adjusted life years 204–5
Immunization (UK) 220 distributional cost-effectiveness
Joint Committee on Vaccination and analysis 231
Immunization (JCVT) 9–10 dominance evaluation 225
Joint Reporting Form (JRF) 337 EQ-5D questionnaire 201–2
global agencies 71–73
Kaplan–Meier curves 188 immunization benefits 76–77
Klebsiella infection 51–52 mixed financing and health insurance 76
PEP providers 74–75
labor private not-for-profit providers 73–74
costs of 98 procurement 38
quality of costing studies 107 vaccination economics 21, 333–35
time allocation 108, 121 vaccination policies 346
language bias 170–71 vaccine delivery 67–68
Latin America, immunization program vaccine-preventable diseases 208–9
financing 342–43 lung cancer, reduction 51–52
Learning Network for Countries in
Transition 358 M14A vaccine purchase database
legislature, investments and 59 (WHO) 34, 39
licensing main outcome measures, decision tree
requirement by country 38 modeling 248
vaccine production costs 29 malaria vaccine 19–20
life tables 187–88 Malawi, vaccination policies 346
lifetime horizon 271 Malaysia
lifetime utility 49, 50 diphtheria, tetanus and pertussis
life year calculations 271–73 vaccine 71, 72b
408 Index

Malaysia (cont.) Medical Expenditure Panel Survey (MEPS)

mixed delivery 75–76 costing studies 107
rotavirus vaccine 220 COVID-19 Markov model 319
marginal costs 32 Medicare, costing studies 107
Market Information for Access to Vaccine Medieval Black Plague 160
(MI4A) (WHO) 33 mega-analysis 172
markets MenAfriVac 21, 36
failure 47 Meningitis Vaccine Project (MVP) 21, 36
investments and 58 meningococcal meningitis vaccine
vaccine research and prices 20t
development 35–37 subunit vaccines 6
Markov models 168, 181, 183, 264–78 MEPS see Medical Expenditure Panel Survey
cohort transition 271–76 (MEPS)
cost-effectiveness analysis 271–76 Merck & Co. Inc., human papillomavirus
costs/benefits 271 vaccine 34
COVID-19 pandemic see COVID-19 messenger RNA vaccines vii
Markov model meta-analyses 182
definition 265–67 parameter estimation 171–73
development of 267–68 meta-regression 172
discounting 275 micro-costing, COVID-19 Markov
economic evaluations 262 model 314, 315t
half-cycle corrections 276 microeconomic benefits 141
hepatitis B vaccination example 268–69 microplanning, new and underutilized
influence/state transition diagram 268 vaccines 126t
input parameters and variables 272t middle-income countries
memoryless property 277 domestic suppliers 38–39
non-linearity 299–300 Gavi 360
parametrization 269–71, 270t mitigation strategies, COVID-19 Markov
trees as 276–77 model 313–14
see also Susceptible–Infected–Recovered mixed costs 30
(SIR) model mixed treatment comparison, parameter
Mastercard, Gavi 353–54 estimation 173–74
maternal health service financing 344 mixed vaccine delivery 75–76
Mathematica 286 MNCs (multinational corporations) 34
Matlab 286 modeling
means agent-based see agent-based models
continuous distributions 189 cost-effectiveness analysis 233
definition 181t cost modeling exercises 83
parameter estimation 193–94 COVID-19 Markov modeling see
measles, mumps, and rubella COVID-19 Markov model
vaccine 53–54 data inputs 165–66
eradication target 351 decision-analytic models see decision-
vaccine prices 20t analytic models
measles, mumps, rubella, and varicella decision models see decision models
vaccine 147–48 decision tree modeling see decision tree
measles vaccine 7 modeling
general health improvement 51–52 decision trees see decision tree modeling
whole pathogen vaccines 6 deterministic models 281–83, 282t
measurement definition 199 discrete-event simulation
Medicaid, costing studies 107 models 286–87
 Index 409

disease transmission see disease National Health Insurance Scheme (NHIS)

transmission modeling (Ghana) 68–69
dynamic see dynamic modeling National Health Mission of India 11
economic evaluations 262 national immunization program
epidemiological modeling 285 (NIP) 53–54, 60
health economic models 181 vaccine delivery 67–68
Markov models see Markov models National Immunization Strategy (NIS) 348
multivariate meta-analysis model 173 new and underutilized vaccine
probabilistic models 300–2 introduction 130
random effect models 171 National Immunization Technical Advisory
static see static modeling Groups (NITAGs) 9–10
stochastic models see stochastic models National Institute for Health and Care
Susceptible–Exposed–Infected–Recovered Excellence (NICE) 220, 220t
(SEIR) model see Susceptible–Exposed– reference cases 233
Infected–Recovered (SEIR) model willingness-to-pay threshold 231
Susceptible–Infected–Recovered national level financing, immunization
model 246, 279–80, 284f programs 341–43
Susceptible–Infected–Recovered– National Program on Immunization 7
Vaccination model 283–84, 284f National Vital Statistics System (CDC) 187
transition models see Markov models neonates/newborns, immunization
monovalent new and underutilized programs 9, 341
vaccines 126t net health benefit (NHB) 229
Monte Carlo analysis 122, 294–95, 296 calculation 152
Markov model 286 vaccine programs and 223
moral foundation theory 56 Netherlands, Gavi contributions 354t
morbidity, COVID-19 infection 311 net monetary benefit (NMB) 229, 302–3
mortality calculation 152
COVID-19 infection 311 cost-benefit analysis 148–49
probability 187 not vaccinating 306
vaccine impact 164 vaccine programs and 223
multi-course study 68 network meta-analysis 173–74
Multi-Dose Vial Policy (WHO) 11 new and underutilized vaccine introduction
multigovernmental purchasing 17 (NUVI) 124–34
multinational corporations (MNCs) 34 cost estimates 124, 126t, 127–30
multiparameter evidence costs 142
synthesis 173–74 data sources 130
multiple study data, parameter Gavi 355
estimation 170–74 inequality and 223b
multivariate meta-analysis model, parameter Nexleaf sensor technology 363
estimation 173 NGOs (nongovernmental
multivariate sensitivity analysis 243 organizations) 158
mumps vaccine 6 NHA (National Health Accounts) 337
MVP (Meningitis Vaccine Project) 21, 36 NHB see net health benefit (NHB)
NHI see national health insurance (NHI)
narrow benefits 141 NICE see National Institute for Health and
National Health Accounts (NHA) 337 Care Excellence (NICE)
national health insurance (NHI) NIP see national immunization program (NIP)
immunization program NIS see National Immunization Strategy (NIS)
financing 343–44 NITAGs (National Immunization Technical
immunization programs 343–45 Advisory groups) 9–10
410 Index

NMB see net monetary benefit (NMB) out-of-pocket (OOP) direct non-medical
non-comparative statistics 181 costs 209
non-excludability, public good 54–55 outreach immunization delivery 100t
nongovernmental organizations (NGOs) 158 overheads, vaccine production 29
non-market time 49–50
non-medical interventions 161–62 PAHO see Pan American Health
non-parametric methods 188 Organization (PAHO)
non-sampling errors 293 paid labor 99t
normal distribution, probabilistic sensitivity Pakistan, local immunization programs 343
analysis 300–1, 301f Pan American Health Organization
North East (NE) quadrant, cost-effectiveness (PAHO)
analysis plane 214–15, 229 COSTVAC 90t
North West (NW) quadrant, cost-effectiveness procurement 38
analysis plane 214–15, 229 Revolving Fund 39–40
Norway Pan American Health organization,
Advance Market Commitment 361 COSTVAC 89
Gavi contributions 354t Panel on Cost-effectiveness in Health and
nosocomial infections 51–52 Medicine (US) 230–31
nucleic aid vaccines 6–7 parameter estimation 167–97
null hypothesis 297 confidence intervals 194
NUVI see new and underutilized vaccine data appropriateness 182
introduction (NUVI) data conversion 183–89
data quality 182
observed data, modeling 164–65 data sources 169–74
odds decision tree modeling 249
definition 176–77, 181t derivation tools 167–81
exposed individual calculation 177 distribution choice 190–93
translation to probabilities 185 interquartile ranges 194
odds ratio literature data sources 174–81
calculation 177, 178, 178t mean and standard errors 193–94
definition 177, 181t parameter types 168–69
probabilities, translation to 186–87 point estimates 189–94
relative risk vs. 180 sensitivity analyses 194–95
OECD (Organization for Economic parameter uncertainty, distributions 189–94
Co-operation and Development) 361 parametric uncertainty 290
older adults see elderly economic evaluations 292–94
omission biases 55–56 parametrization, Markov models 269–71
one-way sensitivity analysis 227, 228f, 243 partial economic evaluation 146t
ongoing costs 127–28, 129–30 patents 20
opportunity costs 153 PATH (Program for Appropriate Technology
optimal public spending, vaccine in Health) 36
development 59–60 Patient-Centred Outcomes Research
optimism biases 55–56 Institute 230–31
oral cholera vaccine 43–44 patient-level data 173
oral polio bivalent vaccine (bOPV) 20t payers, costing studies 104
Organization for Economic Co-operation payroll records 106–7
and Development (OECD) 361 PDPs (product development
out-of-pocket (OOP) direct medical partnerships) 36, 43–44
costs 209 per diems
immunization program financing 333f economic vs. financial costs 99t
 Index 411

new and underutilized vaccine preparatory transition 41t

introduction 129 prevalence definition 181t
personnel, new and underutilized vaccine price negotiation tools 143
introduction 129–30 primary costing studies 103–14
perspective, decision tree modeling 247 cost modeling exercises vs. 83
PERT distribution 194 data collection 106–10
pertussis (whooping cough) vaccine 7 design considerations 103–6
subunit vaccines 6 discounting 116
whole pathogen vaccines 6 health system levels 110
Pfizer, pneumococcal conjugate vaccine 34 perspective 104–5
PFM (public financial management) 347–48 scope 103–4
pharmaceutical companies, R&D shared costs 109–10
investment 63 time horizon 105–6
Philippines, national health primary healthcare spending, immunization
insurance 345–46 programs 341
Pneumococcal Advance Market printing
Commitment (Pneumococcal costs 99t
AMC) 42 reprinting 126–27
pneumococcal conjugate vaccine prioritization, immunization program
(PCV) 69–70 financing 342
Pfizer 34 private clinics, vaccine delivery 74–75
pooled procurement 40 private financing
procurement 41 immunization program financing 333f,
tiered pricing 33 346–47
pneumococcal disease vaccine, subunit vaccine development financing 17
vaccines 6 private gain, public good vs. 159
pneumococcal pneumonia treatment 159 private goods, vaccines as 22
pneumococcal rotavirus vaccine 359–60 private not-for-profit providers (PNFP) 67–
Advance Market Commitment 361 68, 73–74
conjugate vaccines 142 private sector partnerships 362–63
pneumonia, hospitalization 51–52 probabilistic models 300–2
PNFP (private not-for-profit probabilistic sensitivity analysis
providers) 67–68, 73–74 (PSA) 294–303
point estimates, parameter confidence intervals 302
estimation 189–94 cost-effectiveness acceptability
Poisson distributions 191 curves 303, 304f
polio cost-effectiveness scatter plots 302–3
eradication target 351 definition 290
transmissibility 280 expected cost-effectiveness 298–300
polio vaccine 20t influenza vaccine model 295, 296f, 297f
disease management dominance 224 interpretation 302–3
whole pathogen vaccines 6 normal distribution 300–1, 301f
POLYMOD matrices 285 null hypothesis 297
pooled procurement 39–40 outcomes 302
pooled purchasing 23–24 probabilistic models 295f, 300–2
preference-based measures 199, 200b purpose of 297–98
preferences reporting 302
health outcome measurement 198–202, probabilities 190
200b, 201b decay calculation 184
utility rates 192–93 decision tree modeling 250, 251–52
412 Index

probabilities (cont.) quality-adjusted life years (QALYs) 60, 213

definition 174, 181t associated interventions 274–75
means of continuous distribution, calculation 204
deviation to 189 cost-effectiveness analysis and 145
survival data, estimation from 187–89 cost-identification analysis 302
time frame changes 184–85 COVID-19 Markov model 312, 319,
translation from odds 185 320–22
translation from odds ratio 186–87 dominance evaluation 225
translation from rates 184 economic evaluation 203–4, 291
translation from relative risk/risk ratio 185 effectiveness in studies 230
translation from risk difference 186 ethical concerns 231
probability distribution 193–94 expected value of perfect information 306
probabilistic sensitivity analysis 294–95 health outcome measure 145, 198
probability parameters 190–91 high-income countries 204–5
product complexity 31 incremental costs 217
product development partnerships Markov models 268
(PDPs) 36, 43–44 theoretical assumptions 208b
production costs 31 vaccine benefits/effectiveness 140, 205, 206f
product packaging 31 value judgements 205–8
product presentation 38 Quality of Health Economic Studies
program costs, budget impact (QHES) 233
analysis 239–40 quality-of-life measurement 199
Program for Appropriate Technology in Quality of Well-Being Scale 201–2
Health (PATH) 36 questionnaires, labor time 107–8
program management, costs 100t
program planning, immunization R see re-infection (R)
financing 336–37 rabies transmissibility 280
proportion definition 174 rabies vaccine
PSA see probabilistic sensitivity analysis (PSA) vector control 160
publication bias 170–71 whole pathogen vaccines 6
public financial management (PFM) 347–48 racial minorities, COVID-19 Markov
public good model 323–24
non-excludability 54–55 random-effect meta-analysis 171
private gain vs. 159 random effect models 171
vaccines as viii, 22 random events, stochastic models 285
public health randomized controlled trials (RCTs) 167–68,
costing studies 104–5 182
vaccination value 62–63 economic evaluations 261–62
Public Health Service Panel (US) 233 expected value of perfect parameter
public sanitation 160–61 information 307
purchase orders, costing studies 106–7 high-quality 182
push funding 43–44 mixed treatment comparison 173
parametric uncertainty 293
QHES assessment instrument 234 random sampling 115
quality facility sampling 113
analysis, reporting of 232–34 immunization costing studies 111
assessment 38 random variables
data sets 287–88 parametric uncertainty 292–93
definition 70 uncertainty 301–2
vaccine delivery 70 rate of decay calculation 184
 Index 413

rates Revolving Fund (RF), Pan American Health

definition 174–75, 181t Organization 39–40
distribution and 176 rewards/payoffs, Markov models 266
translation to probabilities 184 risk difference
rational choice theory 48–49 definition 181t
RCTs see randomized controlled trials translation to probabilities 186
(RCTs) risk preferences 199–200, 201b
record keeping costs 100t risk ratios
reference case, cost-effectiveness definition 181t
analysis 233–34 translation to probabilities 185
reference pricing 32–33 Rockefeller Foundation, Children’s Vaccine
regional-specific vaccine efficiency rates 288 Initiative 351–52
regression, cost evaluation 120 ROI see return on investment (ROI)
regulatory requirements 31–32 analysis
re-infection (R) 283, see also Susceptible– rotavirus vaccine 220
Infected–Recovered (SIR) model; pooled procurement 40
Susceptible–Infected–Recovered– procurement 41
Vaccination (SIRV) model subunit vaccines 6
relative risk 190 whole pathogen vaccines 6
calculation 177, 178t routine immunizations 7–8
definition 181t facility-based delivery 100t
odds ratio vs. 180 government spending 333
parameters 192 rubella elimination 142
ratio calculation 178 rubella vaccine 6
translation to probabilities 185 Russian Federation, Advance Market
reprinting 126–27 Commitment 361
reproduction rate definition vii
research and development see vaccine Sabin polio vaccine vii
research and development safety assessment 38
resource guides 90t SAGE (Strategic Advisory Group of experts
resource items parameters 190, 191 on Immunization) 9–10
resources Salk polio vaccine vii
allocation 152–53 sample size, immunization costing
costing studies 88–92 studies 111
cost-minimization analysis 147–48 sample stratification, immunization costing
definition of 83 studies 111–12
guides 88–90, 90t sample weights, data analysis 120
inequality of 221b sampling errors 293
resource tracking SARS-CoV2 (severe acute respiratory
definition 336 syndrome coronavirus 2) 310–11
immunization financing 335–36 see also COVID-19 pandemic
retrospective costs 91–92 scenario analysis 122
retrospective self-reporting 107–8 schedules
return on investment (ROI) analysis 88, 144, immunization programs 10–11
145, 146t, 151, 241–44 vaccine schedule 8, 9–11, 157
calculation 151, 242 WHO 11
calculations 87 school attendance 52
cost-effectiveness analysis vs. 241 SDG (Sustainable Development Goals) 344
immunization program financing 342 Seasonal Influenza Immunization Costing
public health interventions 244t Tool (WHO) 89–90
414 Index

Second Panel on Cost Effectiveness in Health societal perspective

and Medicine 89, 105, 121, 164, 292, 312 costing studies 97–98, 104, 105
resource guides 90t immunization program impact 163
SEIR see Susceptible–Exposed–Infected– socioeconomic values 52
Recovered (SEIR) model vaccination value 62–63
selective outcome reporting bias 170–71 software packages
self-procurement 38–39 model parameter calculation 194
semi-variable costs 30 static/dynamic models 279–80
sensitivity analysis, COVID-19 Markov South Africa, cost-effectiveness
model 319, 322–23 thresholds 219
seroprevalence data 287 South East (SE) quadrant, cost-effectiveness
Serum Institute of India, Meningitis Vaccine analysis plane 214–15, 229
Project 36 South West (SW) quadrant, cost-effectiveness
SEs see standard errors (SEs) analysis plane 214–16, 229
severe acute respiratory syndrome Spain, cost-effectiveness thresholds 219
coronavirus 2 (SARSCoV2) 310 Spanish Flu (1918) 160
see also COVID-19 pandemic spillover benefits 156, 157
sexually transmitted diseases 284–85 staff, new and underutilized vaccine
SG see standard gamble (SG) introduction 129–30
shingles vaccine 6 standard errors (SEs) 189–90
SIAs (supplementary immunization estimation 120, 191
activities) 3, 8 parameter estimation 193–94
single study data, parameter standard gamble (SG) 199–200, 202b
estimation 169–70 time horizon 200–1
single (three-arm) trial 169 time trade-off vs. 199–200
SIR (Susceptible–Infected–Recovered) visual analog scale vs. 199–200
model 246, 279–80, 284f standard normal distribution 294, 295f
see also decision tree modeling; Markov Staphylococcus aureus infection 51–52
models start-up costs
SIRV (Susceptible–Infected–Recovered– definition 128
Vaccination) model 283–84, 284f new and underutilized vaccine
sizing, facilities 37 introduction 128–29
smallpox eradication 351 static modeling 279–89
smallpox vaccine appropriateness 279–80
discover vii decision flow chart 281f
whole pathogen vaccines 6 dynamic modeling vs. 280
SMART Vaccines software 157–58 vaccination models 288
social benefits of vaccination 53, 54 static populations, COVID-19 Markov
social insurance, immunization model 324
programs 343–44 statistical life year approach 242
social interventions 160 step-down accounting 101
social mobilization stochastic models
costs 100t calculations 286
new and underutilized vaccine disease transmission modeling 282t, 285–86
introduction 126t, 130 Strategic Advisory Group of experts on
social return on investment analysis 149 Immunization (SAGE) 9–10
societal costs 54 strategic planning, immunization
COVID-19 Markov model 314 programs 85–86
social distancing COVID-19 Markov strategy ranking, incremental cost-
model 322–23 effectiveness ratio 229–30
 Index 415

stratified sampling 113 sustainability

structural uncertainty 290 immunization program financing 336
economic evaluations 292 immunization programs 340–41
model type 292 Sustainable Development Goals
structure, decision tree modeling 247 (SDGs) 344
studies sustainable financing, immunization
comparability 230–31 programs 332
decision tree modeling 247 Sweden, Gavi contributions 354t
subjective assumptions 121 switching, vaccine procurement 41–42
subnational level financing 343 syringes 126–27, 129
subnational units, sampling 112 systematic reviews 182
sub-Saharan Africa, immunization
programs 333 target populations 8–9
subunit vaccines 6 TCVCVT (Typhoid Conjugate Vaccine
success classification 190–91 Costing Tool) 89–90
sugar taxes 345–46 Teaching Vaccine Economics Everywhere
supervisors (TVEE) 90t
costs 100t terminal nodes 246
new and underutilized vaccine tetanus
introduction 126t eradication target 351
supplementary immunization activities transmissibility 280
(SIAs) 3, 8 tetanus vaccine 7
suppliers, vaccine supply landscape 34 subunit vaccines 6
supply chain viii, 40, 127, 360, 362, 363 time horizon 105
supply curves 23–24, 24f Thailand
Supply Division (SD) (UNICEF) 39–40, cost-effectiveness thresholds
41–42 identification 220t
supply meeting demand 31 price negotiations 143
supply-side, cost-effectiveness therapeutic paradigms 51
analysis 217–18 third-party contributions, vaccine
surveillance, costs 100t production costs 30
survival curves 188 three-arm (single) trial 169
definition 181t threshold analysis 122
survival data 188 threshold estimate, cost-effectiveness
probability estimation 187–89 analysis 218
susceptibility to disease (S) 283 tiered pricing 33
see also Susceptible–Infected–Recovered TikTok, Gavi 353–54
(SIR) model; Susceptible–Infected– time-dependency, COVID-19 pandemic 324
Recovered–Vaccination (SIRV) model time horizon
Susceptible–Exposed–Infected–Recovered decision tree modeling 248
(SEIR) model 310 immunization programs 164–65
simulation of 312–13 Markov models 267
see also COVID-19 Markov model time trade-off & standard gamble 200–1
Susceptible–Infected–Recovered (SIR) time-lag bias 170–71
model 246, 279–80, 284f time-motion studies 108–9
see also decision tree modeling; Markov time preferences 199–200, 201b
models timescale
Susceptible–Infected–Recovered–Vaccination Markov models 267
(SIRV) model 283–84, 284f probability changes 184–85
Susceptible (S) State 248–49 time-to-event data 170
416 Index

time trade-off (TTO) 199–200 subunit vaccines 6

standard gamble vs. 199–200 Vi-conjugate vaccine 171
time horizon 200–1 whole pathogen vaccines 6
visual analog scale vs. 199–200
tobacco taxes 345–46 Uganda, immunization program
tornado diagrams financing 342
cost uncertainty 122f uncertain markets 43
COVID-19 Markov model 319 uncertainty 225–28
one-way sensitivity analysis 228f analysis 227
total annual life years 274 cost analysis 121–22
total costs, decision tree modeling 253 data sources 228
tracing factor 109–10 definition 226
trade-offs, risks and 199 distributions in parameter
training estimation 189–94
costs 100t economic evaluations 291–94
new and underutilized vaccine heterogeneity vs. 226–27
introduction 126–27, 126t random variables 301–2
transient Markov chain 267 risks and 199
transition matrices, Markov models 270 tornado diagram 122f
transition models see Markov models variability vs. 226–27
transition policy, Gavi 356–57 underutilized vaccines see new and
transition probability, Markov models 266, underutilized vaccine introduction
267 (NUVI)
transmission rate (β) 284–85 UNICEF 353
traveller’s vaccines 9 Children’s Vaccine Initiative 351–52
travel to vaccination costs 98 costing resource guides 89–90
treatment of disease 159 COVID-19 vaccine allocation 17
cost in COVID-19 Markov model 322–23 data on immunization
COVID-19 Markov model 320 expenditure 337–38
efficacy in COVID-19 Markov Healthy Markets Framework 35
model 322–23 supply division 359
value propositions 311 Supply Division 39–40, 41–42
tropical countries, routine immunizations 7 Universal Childhood Immunization 351
trust funds, immunization program unit costs 190, 191–92
financing 346–47 per dose 96
TTO see time trade-off (TTO) United Kingdom (UK)
tuberculosis Advance Market Commitment 361
eradication target 351 cost-effectiveness thresholds 219, 220t
treatment 159 Gavi contributions 354t
tuberculosis vaccine United Nations (UN), Children’s Vaccine
financial investment 19–20 Initiative 351–52
whole pathogen vaccines 6 United States (US)
Tufts Global Health Cost Effectiveness cost-effectiveness thresholds
Registry 204–5 identification 220t
TVEE (Teaching Vaccine Economics Gavi contributions 354t
Everywhere) 90t Universal Childhood Immunization
two-way sensitivity analysis 227 (UNICEF) 351
Typhoid Conjugate Vaccine Costing Tool universal health coverage 344f, 344
(TCVCVT) 89–90 costs of 344
typhoid vaccine unvaccinated population, infection risk 185
 Index 417

UPS Foundation, Gavi 362 quality 70

USAID, immunization financing 354–55 service delivery 126t
utilities Vaccine for Children (VFC) Act (1993) 76
economic vs. financial costs 99t vaccine hesitancy 4, 8, 13, 47, 56–57
preference rates 192–93 COVID-19 pandemic 17–18
utility rates 200–1 misinformation and 23
transitioning countries 359
vaccination WHO 47
hesitancy see vaccine hesitancy see also vaccination refusal (anti-vaxxers)
net benefits see vaccination net benefits Vaccine Impact Modeling Consortium
programs of see immunization programs (VIMC) 246
social benefits 53, 54 vaccine-preventable diseases (VPDs)
status in COVID-19 pandemic and 179– financial consequences 209
80, 179t low- and middle-income countries 208–9
systems of 19f vaccine procurement 37–44, 143–44
vaccination alternatives 158 economic effects 37–38
public sanitation 160–61 innovative methods 42–43
social interventions 160 pooled procurement 39–40
treatment 159 procurement modalities 38–40
vector control 160 self-procurement 38–39
see also disease eradication; herd switching 41–42
immunity; public sanitation; social trends in 40–42
interventions; treatment of disease; vaccine production viii, 3
vector control costs 29–30
vaccination net benefits 58–59 vaccine research and development 5, 28–29,
return on investment analysis 243–44 158
see also net health benefit (NHB) accounting for 30
vaccination rate cost-effectiveness analysis 162
annual see annual vaccination rate expected value of perfect information 307
COVID-19 Markov model 322–23 finances for 17
vaccination refusal (anti-vaxxers) viii, fixed cost as 28
54–55, 56–57 future investments 62–63
see also vaccine hesitancy investments in 47–48
Vaccine Alliance see Gavi market signals 35–37
vaccine allocation 16 pharmaceutical companies 63
inefficiencies of 221b process costs 28
vaccine coverage unsuccessful 29
inequality of 208–9 vaccines
shortfall 8 adoption in cost-effectiveness
vaccine delivery 3, 67–80 analysis 161–62
cost variation 86 allocation see vaccine allocation
coverage 71 competition 35
efficiency 68 costs 100t
equity 69–70 costs in type 100, 100t
financial sustainability 68–69 coverage see vaccine coverage
immunization programs 12t economic vs. financial costs 99t
increase in 71–76 effectiveness measurement 203–8
mixed vaccine delivery 75–76 efficacy in COVID-19 Markov
outreach immunization delivery 100t model 322–23
public delivery 71–73, 72b human capital as 48–51
418 Index

vaccines (cont.) whole pathogen vaccines 6

limited effectiveness 50 varicella zoster vaccine 6
new/underutilized see new and VAS see visual analog scale (VAS)
underutilized vaccine introduction vector control 160
(NUVI) vehicles, economic vs. financial costs 99t
new vaccine costs see new and VFC (Vaccine for Children) Act (1993) 76
underutilized vaccine introduction VIMC (Vaccine Impact Modeling
(NUVI) Consortium) 246
prices 20, 20t, 27–28, 32–33 visual analog scale (VAS) 199–200
procurement see vaccine procurement standard gamble vs. 199–200
production see vaccine production time trade-off vs. 199–200
research and development see vaccine volatile markets 43
research and development volume guarantees, vaccine procurement 43
scarcity during COVID-19 pandemic 18 volume weighted mean 119
selection 9–10 volunteer labor costs 99t
storage 5 VPDs see vaccine-preventable diseases (VPDs)
suppliers 34
transport of see cold chain; vaccine wastage costs 12, 12t
transport waste management
types of 6–7 economic vs. financial costs 99t
value propositions 311 new and underutilized vaccine
wastage rate 12 introduction 126t
Vaccines Global Access (COVAX) 124 whole pathogen vaccines 6
vaccine transport 5 whooping cough, eradication target 351
COVID-19 vaccines 17 willingness-to-pay threshold, incremental
economic vs. financial costs 99t cost-effectiveness ratio 152
new and underutilized vaccine worker productivity viii
introduction 126t, 129 workforce size 31
validation, data requirements 287–88 World Bank 353
value Children’s Vaccine Initiative 351–52
analysis methods 147–51 immunization financing 360–61
cost measurement equations 144 World Health Organization (WHO) 353
costs vs. 144 Building Blocks Framework 18–19
definition 48–49, 199 Cervical Cancer Prevention and Control
health terms 140 Costing Tool 89–90, 90t
investments of 153 Children’s Vaccine Initiative 351–52
measurement 143–45 Choosing Interventions that are Cost-
for money 144 Effectiveness program 86–87
preferences 199–200, 201b Collaborative Procedure for Accelerated
understanding 140–41 Registration 38
vaccination of 51–53 costing resource guides 89
value-based pricing 33 COSTVAC 89–90
value of information analysis 304–8 COVAX AMV 361
expected information 305–7 COVID-19 Markov model 314–19
perfect parameter information 307–8 COVID-19 vaccine allocation 17
sample information 307–8 Expanded Program on Immunization 351
variability, uncertainty vs. 226–27 M14A vaccine purchase database 34, 39
variable costs 30 Market Information for Access to
varicella (chickenpox) vaccine Vaccine 33
vaccine prices 20t Meningitis Vaccine Project 36
 Index 419

Multi-Dose Vial Policy 11 vaccine hesitancy 47, 56–57

National Immunization Strategy 348 vaccine schedules 11
new and underutilized vaccine vaccine volume calculators 128–29
introduction 126
resource guides 90t yellow fever vaccine 6
rubella elimination 142
Seasonal Influenza Immunization Costing zanamivir 313–14
Tool 89–90 Zenysis 363
Strategic Advisory Group of experts on zero cost of vaccines (free cost) 22–24
Immunization 9–10 Zipline drone network 362–63