SUSTAINABILITY OF THE HVAC

SYSTEMS IN THE CLINICAL

OPERATING ROOM DEPARTMENT

Bachelor thesis

M. de Groot (Matthijs)

University of Twente
Faculty of Behavioral, Management, and Social Sciences
Bachelor of Science in Industrial Engineering & Management
 July 2022

BSc Industrial Engineering & Management

Sustainability of the HVAC systems in the clinical operating room

department

Author: M. de Groot (Matthijs)

s2394243

This report was written as part of the bachelor thesis project assignment in the
Industrial Engineering & Management educational program.

University of Twente Amsterdam UMC

Drienerlolaan 5 Meibergdreef 9
7522 NB Enschede 1105 AZ Amsterdam

Supervisors Supervisors
Dr. A. Asadi (Amin) N. Sperna Weiland (Niek)
Dr. Ing. S. Faeghinezhad (Shiva) A. Timmermans (Anne)
Prof. Dr. Ir. E.W. Hans (Erwin)

Executive summary 2
Executive summary
Problem definition
The Amsterdam UMC is a hospital located in Amsterdam, Netherlands. The healthcare industry is one
of the largest polluters with an average share of 5.5% of the total national carbon footprint. In the
Netherlands, this share was higher than 5.5% according to Pichler et al. (2019). The healthcare industry
was responsible for 8.1% of the total national carbon footprint in the Netherlands (Pichler et al., 2019).
This is the highest share compared to a sample of 36 countries affiliated with the Organization for
Economic Co-operation and Development (OECD). Heating, Ventilation, and Air Conditioning (HVAC)
systems are responsible for approximately 30%-50% of the total energy use in hospitals, which means
HVAC systems emit large amounts of CO2. However, hospitals must obey guidelines concerning the
air treatment of operating rooms (ORs) to prevent surgical site infections. This makes it complicated
for hospitals to reduce the energy consumption of HVAC systems. Recent changes in the guideline for
air treatment in ORs stimulate hospitals to implement changes in their HVAC system usage, reducing
CO2 emissions. The Amsterdam UMC has not implemented changes since the revised air treatment
guideline was published.

This research consists of a problem-solving approach containing 7 steps designed to answer the
research question:
“How can the planning of the HVAC system usage and the scheduling of the surgeries in the ORs be
optimized so that a reduction in energy consumption is realized without compromising the efficiency
of the OR while obeying the guideline concerning air quality in the OR? “

Solely the clinical OR department of the Amsterdam UMC is within the scope of this study. The
department consists of 20 ORs of class 1 performance level 1. This is the OR class with the strictest
requirements. This means that all procedures can be executed at each OR. However, significant
amounts of energy are consumed due to the strict requirements of the class 1 performance level 1
ORs. Figure 1 depicts the performance requirements of class 1 performance level 1 ORs and class 1
performance level 2 ORs.

Figure 1: Performance requirements class 1 ORs “At rest”

The recently published revised guideline introduced a new classification of class 1 ORs. Class 1 ORs are
now distinguished by class 1 ORs and class 1+ ORs, replacing the performance level 1 and performance
level 2 classification. The performance level requirements of class 1+ and class 1 ORs are similar to the
requirements of performance level 1 and performance level 2 ORs, respectively. The revised guideline
abandons the recommendation of unidirectional flow HVAC systems because there is no evidence that
these systems prevent surgical site infections more effectively than turbulent HVAC systems. Major
joint replacements are recommended to be executed in class 1+ ORs, while class 1 ORs suffice for all
other surgeries. All surgeries can be executed in class 1+ ORs.

Executive summary 3
Since 2018, 15 of the ORs are in nighttime operation between 21:00 and 6:30 on Monday to Friday
and between 00:00 and 23:59 during weekends and holidays. The remaining five ORs are functional
24/7. This resulted in an estimated reduction in energy consumption of approximately 33%.

Research methods
A problem identification and motivation are performed to gain insights into the problem. An action-
and a core problem is selected during this stage, and the intended deliverables, key constructs and
variables, the theoretical perspective, and the research design are identified. Then, a literature review
is performed to solve the knowledge problems. By analyzing the current situation at the Amsterdam
UMC afterward, all knowledge problems must be solved. Improvement measures can be generated,
selected, and elaborated on as soon as all knowledge problems are solved. The research is concluded
after having determined the conclusion and recommendations. The deliverables consist of an effective
heuristic for efficient surgery scheduling, insights into the energy consumption before and after the
implementation of the proposed heuristic, and potentially other recommendations concerning the
sustainability of the hospital. A Monte Carlo simulation is used to assess the effectiveness of the
proposed heuristic. The goal of the heuristic is to effectively schedule surgeries for the two types of
ORs: Class 1 ORs and class 1+ ORs. Data concerning all executed surgeries in 2019 is used for this
research, as this is the most recent year unaffected by the COVID-19 pandemic.

Results
All executed distinct procedures in 2019 have been classified as either class 1 or class 1+ surgeries. We
conclude from the data of all executed surgeries in 2019 that two class 1+ ORs suffice. The remaining
18 ORs will be class 1 ORs. The data from 2019 suggested that five 24/7 ORs were rarely occupied
outside operating hours and if this was the case, the ORs were not used simultaneously. Decreasing
the number of 24/7 ORs to four 24/7 ORs is viable, and this should be implemented as this reduces
energy consumption. In 2019, only 19 surgeries classified as class 1+ surgeries started outside
operating hours (6:30 – 21:00). None of these surgeries occurred on the same day and not even in the
same week. We conclude that it would suffice if one class 1+ OR is appointed as a 24/7 OR. Two ORs
should also be appointed as emergency ORs. No surgeries are scheduled for these ORs, but they are
solely used for emergencies. The emergency ORs should be classified as class 1 ORs, as joint
replacements are generally not urgent.

Two heuristics proposed by Arnaout & Kulbashian (2008) for scheduling the surgeries while minimizing
the makespan are evaluated during this research: Longest Expected Processing Time (LEPT) and
Longest Expected Processing Time with setup times (LEPST). The LEPT heuristic first schedules the job
with the longest expected processing time, while the LEPST heuristic first schedules the job with the
highest value of the expected processing time plus the corresponding setup time multiplied by the
control parameter 𝛼. The control parameter 𝛼 determines the relative importance of the setup time
as opposed to the processing time. The optimal value of control parameter 𝛼 is researched using the
Monte Carlo simulation and it equals 𝛼 = 0.65. The proposed heuristics LEPT and LEPST are designed
for assigning jobs to identical machines. We want to distinguish between class 1 ORs and class 1+ ORs.
Therefore, some changes are made to the heuristic. We denote the adjusted LEPT and LEPST heuristics
as LEPT* and LEPST*, respectively.

The Monte Carlo simulation is used to assess the effectiveness of the LEPT* and LEPST* heuristics. The
objective of the heuristics is to minimize the maximum completion time, which is also referred to as
the makespan. A completion time is defined as the point in time at which a job is finished. Minimizing
the makespan allows the ORs to switch to nighttime operation as soon as possible, which decreases
the energy consumption of the clinical OR department. In the simulation, the same number as the

Executive summary 4
average number of executed elective surgeries per week in 2019 is generated according to the
empirical distributions of the processing times and the setup times of both class 1 and class 1+. These
generated surgeries are then scheduled according to the LEPT* and LEPST* heuristics. The average
maximum completion time of every day (𝐶𝑚𝑎𝑥 ̅ ) is used as the output parameter, as this reflects the
extent to which the two heuristics effectively achieve the objective of minimizing the maximum
completion time. Figure 2 depicts the output of the Monte Carlo simulation using 1000 iterations. This
means that a week is generated repeatedly 1000 times to acquire the most accurate estimate of 𝐶𝑚𝑎𝑥 ̅ .
We conclude from Figure 2 that the LEPST* heuristic outperforms the LEPT* heuristic as the value of
̅
𝐶𝑚𝑎𝑥 ̅
for the LEPST* heuristic is less than the value of 𝐶𝑚𝑎𝑥 for the LEPT* heuristic.

Figure 2: Monte Carlo simulation output

̅
Data from 2019 showed that the OR department attained a 𝐶𝑚𝑎𝑥 of 812 minutes. The LEPST* heuristic
̅
attains a 𝐶𝑚𝑎𝑥 of 456 minutes, which is a reduction of the makespan of approximately 44%. After
implementing the proposed HVAC system changes, the LEPST* heuristic, and the proposed OR
classification, the hospital is expected to save 791,820.73 kWh annually compared to the energy
consumption in 2019, corresponding to a reduction of approximately 48% relative to 2019. Using a
CO2e emission intensity of 0.390 kg CO2e per kWh, 791,820.73 ∗ 0.390 = 308,810.085 kg CO2e can
be saved compared to 2019. This conversion factor for the Netherlands was determined by the
European Environmental Agency (2019).

Executive summary 5
Acknowledgments
Dear reader,

You are about to read the thesis that concluded my Bachelor of Industrial Engineering and
Management. The research “Sustainability of the HVAC systems in the clinical operating room
department” was conducted at the Amsterdam UMC from April 2022 until August 2022. This thesis
aims to reduce the energy consumption of the Amsterdam UMC hospital while obeying the most
recent guidelines concerning air treatment of ORs.

I am grateful to all the people who contributed to this thesis. I would like to thank some in particular.
I would like to express my deepest gratitude to my first supervisor Amin Asadi for guiding me
throughout this research. I am also thankful to Shiva Faeghinezhad, who provided knowledge and
expertise. This endeavor would also not have been possible without Erwin Hans, who is my second
supervisor.

I would like to thank everyone who contributed to my thesis at the Amsterdam UMC. Special thanks
go to Niek Sperna Weiland and Anne Timmermans, my supervisors from the Amsterdam UMC
hospital. Both supervisors were supportive and consistently provided me with helpful feedback. The
atmosphere at the hospital was very pleasant, as everyone was willing to help.

Finally, I also wish to thank my family and friends for their love and support. They always provided
moral support when things got difficult.

Matthijs de Groot

Vianen, August 2022

Acknowledgments 6
Contents
Executive summary ........................................................................................................................... 3
Acknowledgments ............................................................................................................................ 6
List of acronyms ................................................................................................................................ 9
Reader’s guide ................................................................................................................................ 10
1 Introduction ............................................................................................................................ 11
1.1 Problem owner description .............................................................................................. 11
1.2 Problem identification...................................................................................................... 12
1.3 Selection of the action- and the core problem .................................................................. 12
1.4 Research questions .......................................................................................................... 13
2 Methodology .......................................................................................................................... 15
2.1 Problem-solving approach ................................................................................................ 15
2.2 Intended deliverables....................................................................................................... 17
2.3 Research design ............................................................................................................... 17
2.4 Validity and reliability of measurement ............................................................................ 18
3 Literature review..................................................................................................................... 19
3.1 Previous guidelines concerning the air treatment of operating rooms in Dutch hospitals . 19
3.2 Changes in the revised guidelines concerning the air treatment of operating rooms in
Dutch hospitals compared to the previous guidelines .................................................................. 26
3.3 Optimal job scheduling..................................................................................................... 27
4 Evaluation of the Amsterdam UMC ......................................................................................... 29
4.1 Scheduling ....................................................................................................................... 29
4.2 HVAC system.................................................................................................................... 33
4.3 Energy consumption of the clinical OR department .......................................................... 37
5 Conceptual design ................................................................................................................... 39
5.1 OR classification and classification of surgeries................................................................. 39
5.2 HVAC system .................................................................................................................... 40
5.3 Heuristics ......................................................................................................................... 41
6 Monte Carlo simulation .......................................................................................................... 45
6.1 Data preparation .............................................................................................................. 45
6.2 MC simulation .................................................................................................................. 46
7 Results .................................................................................................................................... 49
7.1 Control parameter α ........................................................................................................ 49
7.2 Heuristic selection ............................................................................................................ 49
7.3 Expected impact on sustainability .................................................................................... 50

Acknowledgments 7
Conclusions and recommendations ................................................................................................ 52
Conclusions ................................................................................................................................. 52
Recommendations....................................................................................................................... 54
Discussion ....................................................................................................................................... 55
References ...................................................................................................................................... 56
Appendices ..................................................................................................................................... 58
Appendix A: Map of the Amsterdam UMC ................................................................................... 58
Appendix B: Schematic overview of the OR department’s pressure levels .................................... 59
Appendix C: Semi-structured interviews ...................................................................................... 60
Appendix D: Surgery classification as proposed in the revised guideline....................................... 66
Appendix E: Surgery classification and allocation example ........................................................... 69
Appendix F: Example of daily planning ......................................................................................... 70
Appendix G: Requirements for critical parameters ....................................................................... 71
Appendix H: Frequently applied HVAC system configurations ...................................................... 72
Appendix I: Energy consumption tool assumptions ...................................................................... 74
Appendix J: Surgical procedures for class 1+ ORs ......................................................................... 75
Appendix K: Implemented Excel functions ................................................................................... 79
Appendix L: Original Excel file of historical data of 2019............................................................... 80
Appendix M: Data elective procedures 2019 ................................................................................ 81
Appendix N: Implemented Excel Visual Basic code ....................................................................... 82

Acknowledgments 8
List of acronyms
ACRONYM DEFINITION
AMC Academic Medical Center
Amsterdam UMC Amsterdam University Medical Center
BPMN Business Process Model Notation
CDF Cumulative Distribution Function
CF Carbon Footprint
CFU(s) Colony-Forming Unit(s)
EDD Earliest due date
FCFS First come, first served
GHG Greenhouse Gas
HVAC Heating, Ventilation, and Air Conditioning
IAQ Indoor Air Quality
KPI(s) Key performance indicator(s)
L(E)PT Longest (expected) processing time
LEPST Longest expected processing time with setup
times
MC Monte Carlo
MPSM Managerial Problem-Solving Method
NVMM Nederlandse Vereniging voor Medische
Microbiologie
NVVH Nederlandse Vereniging voor Heelkunde
OR(s) Operating room(s)
RH Relative Humidity
RQ(s) Research Question(s)
S(E)PT Shortest (expected) processing time
SLR Systematic Literature Review
SSI(s) Surgical Site Infection(s)
UDF Unidirectional Flow
VUmc VU University Medical Centers
WIP Werkgroep Infectie Preventie

List of acronyms 9
Reader’s guide
The reader’s guide briefly explains the general relevance and content of each chapter. It contains a
brief, general overview of the thesis.

1 Introduction
The first chapter introduces the research by describing the problem. This is done by performing a
problem identification, elaborating on the problem owner, selecting an action- and core problem, and
by dividing the research into research questions designed to be able to conclude the research.

2 Methodology
The “Methodology” chapter describes the methodology implemented in this research. The problem-
solving approach is described, and the intended deliverables, research design, and the validity and
reliability of measurement are stated.

3 Literature review
After introducing the research and describing the methodology, a literature review is performed.
Literature related to this research is studied in this chapter concerning the previous and revised
guideline concerning air quality in operating rooms and scheduling problems. The scheduling problem
of this research is formulated using optimal job scheduling. The literature review is therefore focused
on this topic specifically after having researched the guidelines.

4 Evaluation of the Amsterdam UMC

Relevant information concerning the Amsterdam UMC is described in this chapter. This chapter is
divided into the sections: “Scheduling”, “HVAC system”, and “Energy consumption of the clinical OR
department” as these are the relevant topics for this research.

5 Conceptual design
Chapter 5 describes a conceptual design that is proposed to the Amsterdam UMC. This conceptual
design proposes a (near-optimal) solution to the Amsterdam UMC. This solution obeys the guideline
and uses optimal job scheduling researched in the literature review. The conceptual design proposes
a classification concerning ORs and surgeries, the type of HVAC systems, and two heuristics that will
be compared using the Monte Carlo simulation.

6 Monte Carlo simulation

The effectiveness of the two proposed heuristics must be compared using the Monte Carlo simulation.
This chapter elaborates on the Monte Carlo simulation by elaborating on the data preparation and
simulation programming. Additionally, it depicts and explains the interface of the Monte Carlo
simulation. It does not elaborate on the most effective heuristic, as that will be done in Chapter 7:
“Results”.

7 Results
This chapter formulates the results of the research by describing the optimal value of control
parameter α for the LEPST* heuristic, a selection of the proposed heuristics, and the expected energy
consumption reduction while using the proposed optimal heuristic.

Reader’s guide 10
1 Introduction
Before describing the research methodology, we will start by giving elaboration on the problem owner
and by performing a problem identification in this chapter. The process of performing a problem
identification is described by Heerkens & Winden (2017) in Solving Managerial Problems
Systematically, where the seven steps of the Managerial Problem-Solving Method (MPSM) are
illustrated. It entails the selection of an action problem – defined as a discrepancy between the norm
and reality as defined by the problem owner - after which an inventory of potential causes of this
problem is made. Finally, a single core problem – defined as the problems which have no direct causes
themselves - is selected. Therefore, core problems can be found at the very beginning of a problem
cluster. Because core problems are at the beginning of a problem cluster, solving these problems has
the most impact when evaluating an action problem. This chapter also formulates the knowledge
problems – defined as knowledge gaps that need to be resolved to be able to solve the core problem
- which can be considered interchangeably with the term “research questions (RQs)” for the remainder
of this report, as the research questions currently cannot be answered because of a knowledge gap.
A knowledge gap can be defined as a discrepancy between the desired level of expertise and the
current expertise concerning a particular topic. RQs must be answered to solve the knowledge
problems. This research aims to solve the core problem after solving the knowledge problems.

1.1 Problem owner description

To get an understanding of the context of this research, a description concerning the problem owner
is given in this section. The problem owner is an academic hospital in Amsterdam called Amsterdam
University Medical Centers (Amsterdam UMC). “Together we discover the healthcare of tomorrow” is
what Amsterdam UMC stands for (Amsterdam UMC, 2022). Amsterdam UMC is one of the eight Dutch
academic hospitals. The hospital strives to provide the best patient treatment by continuously
conducting innovative research and educating future professionals in healthcare.

A merger in 2018 between the Academic Medical Center (AMC) and the VU UMC (VUmc) resulted in
the Amsterdam UMC. The merger is often referred to as an administrative merger, as the locations of
the original separate hospitals remain the same. The AMC and VUmc are nowadays referred to as
Amsterdam UMC location AMC and Amsterdam UMC location VUmc, respectively. Amsterdam UMC
consists of medical faculties from the University of Amsterdam and the Vrije Universiteit Amsterdam.
The two locations employ over 16.000 people in total (Amsterdam UMC, 2022). The scope of this
research is the Amsterdam UMC location AMC. The Amsterdam UMC location AMC is for simplicity
reasons referred to as Amsterdam UMC for the remainder of this research. This hospital consists of
two operating room (OR) departments: The day center and the clinical OR department. These OR
departments consist of five and 20 ORs, respectively. The OR departments consist of different OR
classes; The five ORs in the day center are classified as class 1 performance level 2 ORs, while the 20
ORs in the clinical OR department are class 1 performance level 1 ORs. Class 1 performance level 1 is
the OR class with the strictest requirements. The clinical OR department is located in departments D1
and E1, while the ORs of the day center can be found in department D0. See Appendix A for a map of
the hospital.

Niek Sperna Weiland and Anne Timmermans guided me during this research. They are anesthetist and
gynecologist at the Amsterdam UMC, respectively. Additionally, Niek Sperna Weiland founded the
Green Team OR, and they are both active members of the Green Team OR. This team consists of
ambitious medical specialists and healthcare professionals aiming to decrease CO2 emissions in the
OR. The Green team OR aspires to achieve a more sustainable OR by focusing on air treatment, waste
disposal, and anesthetic gases.

Introduction 11
1.2 Problem identification
In this problem identification, the effect and the significance of the problem are described. A brief
literature review is executed to research the problem’s significance. An action- and a core problem is
selected as a result of the problem identification. Section 1.3: “Selection of the action- and the core
problem” concludes the description given in this section. The goal of this section is to identify and
describe the underlying problem as perceived by the problem owner.

The healthcare industry is among the largest CO2 polluters in the industrialized world, as its share of
the total national carbon footprint (CF) was 5.5% on average back in 2014 and is therefore partially
responsible for climate change. Pichler et al. (2019) concluded that the Netherlands had the highest
share compared to a sample of 36 Organization for Economic Co-operation and Development (OECD)
countries with 8.1% of its total CF. The healthcare industry is responsible for a substantial amount of
greenhouse gas (GHG) emissions and to prevent climate change, these GHG emissions need to be
reduced. A gap between norm and reality can be identified where the norm entails a world without
climate changes due to GHG emissions. This gap between norm and reality cannot be resolved during
this research. However, a reduction in GHG emissions caused by Heating, ventilation, and air
conditioning (HVAC) systems is feasible. The large amounts of GHG emissions caused by hospitals are
partly because they operate continuously, which requires substantial amounts of energy. HVAC is
generally responsible for approximately 30%-50% of the total energy use in hospitals (Cubí Montanyà,
2014). Additionally, Guidelines for the internal air quality (IAQ) of ORs must be obeyed, which makes
it difficult for hospitals to decrease their HVAC system’s CF. However, recently published changes in
the Dutch guidelines for air treatment inside ORs enable hospitals to implement changes in their HVAC
usage to realize a decrease in their HVAC system’s CF. The Amsterdam UMC hospital did not
implement changes made possible by the latest air treatment guidelines thus far.

1.3 Selection of the action- and the core problem

Section 1.2: “Problem identification” elaborated on the problem as perceived by the problem owners
of the Amsterdam UMC. As a result of the problem identification, we can define the action problem
as follows:

- “Earth experiences climate change due to human activities.”

The core problem, found at the very start of the problem cluster, can be defined as:

- “The Amsterdam UMC hospital did not implement changes made possible by the latest air
treatment guidelines thus far.”

Figure 1-1: Problem cluster

Introduction 12
Figure 1-1 gives a clear overview of the causes and effects of the problems that can be identified
concerning the sustainability of the healthcare industry. It starts with the action problem that needs
to be tackled. This action problem concerns climate change due to human activities. This action
problem is caused by many polluters and the healthcare industry is not solely responsible for this
problem. However, they are responsible for large amounts of GHG emissions, and they are therefore
within the scope of this research. The causes of climate change within the healthcare industry are
visualized in Figure 1-1 afterward. This research also specifically focussed on the emissions regarding
the HVAC systems of the Amsterdam UMC because these systems are responsible for the largest part
of the emissions within ORs. This is caused by the fact that the ORs must obey guidelines concerning
their IAQ. Because of this, the Amsterdam UMC hospital retains all HVAC systems in the 20 ORs
switched on during the day to simplify the scheduling of surgeries. However, recent changes in the
guidelines make it possible to implement changes in the ORs potentially leading to a reduction in
energy consumption. These changes in the guidelines are currently not implemented in the scheduling
of surgeries at the Amsterdam UMC. This problem has no other causes, and it is the cause of all other
problems. It is therefore the core problem.

1.4 Research questions

This section elaborates on the knowledge problems related to the research. To solve the core problem
as defined in section 1.3: “Selection of the action- and the core problem”, research questions are
formulated that must be answered during this research. The knowledge gaps will be solved by
answering the research questions that are formulated in this section.

- RQ1: What were the previous guidelines concerning the air treatment of operating rooms
in Dutch hospitals?
To understand the impact of the previous guidelines on the current scheduling processes,
research should be conducted concerning the contents of the previous guidelines. Solving this
knowledge problem provides a better understanding of the current situation in the
Amsterdam UMC hospital concerning surgery planning and HVAC usage. This research
question is answered using a literature review, as documents of the Dutch organization
responsible for the guidelines provide information regarding the guidelines. Additionally, a
semi-structured interview with Ingrid Spijkerman is performed for additional information.
Ingrid Spijkerman is responsible for infection prevention and medical microbiology within the
Amsterdam UMC. Additionally, she was a member of the expert group “Luchtbehandeling op
de OK”, which can be translated as “Air treatment in the OR”. This expert group was involved
in writing the new guideline.

- RQ2: What are the most important changes in the latest guidelines concerning the air
treatment of operating rooms in Dutch hospitals compared to the previous guidelines?
The changes in the guidelines of 2022 compared to the previous guidelines should be
understood as these changes are why the problem owners believe there is room for
improvement. Solving this knowledge problem, therefore, simplifies idea generation. Hence,
it contributes to solving the core problem. The new guidelines might also form the basis of the
constraints when developing a new scheduling heuristic. This research question is solely
answered using a literature review, as documents of the Dutch organization responsible for
the guidelines provide all information regarding the guidelines. The semi-structured interview
with Ingrid Spijkerman is useful for this knowledge problem.

Introduction 13
- RQ3: What are the HVAC systems in ORs used for, which type of HVAC systems are used in
the Amsterdam UMC, and how are they operated?
The processes concerned with the HVAC systems need to be understood. To devise changes
in the processes, a clear understanding of the current processes is required. Clear knowledge
concerning the processes is crucial during this research. Obtaining that knowledge is the main
purpose while solving this knowledge problem. This RQ determines which specific guidelines
apply to the ORs in the Amsterdam UMC as the HVAC systems used in the hospital are
identified here. Observations and a literature study are used here as information can be
extracted during guides around the OR department for example, but data can also be found
in the air management plan Luchttechnischbeheersplan (Amsterdam UMC, 2018) of the
Amsterdam UMC hospital.

- RQ4: How are the planning of the HVAC system usage and the scheduling of the surgeries in
the ORs of the Amsterdam UMC currently executed?
The surgery scheduling processes and planning of the HVAC system usage in the hospital need
to be understood, as a change concerned with these processes is sought during this research.
To devise a change in the current processes, the current processes need to be understood.
Solving this knowledge problem simplifies the optimization of the OR as it provides a more
specific gap between norm and reality. Interviews, observations, and literature reviews are
used to answer this research question. Interviews concerning this knowledge problem are
held with Karin de Vlaming. Karin de Vlaming is the manager of the OR department. Observing
the current scheduling process is also effective when answering this research question. Finally,
the criteria for appointing a certain specialty to an OR will be researched as this is crucial when
developing the scheduling heuristic. This is researched by performing a literature review using
the air management plan of the Amsterdam UMC hospital.

- RQ5: What is the total energy consumption of the HVAC systems in the OR department of
the Amsterdam UMC hospital before and after implementation of the proposed heuristic?
This RQ makes the gap between norm and reality measurable. Data concerning this knowledge
problem is acquired by performing semi-structured interviews with Jelle Koeman, Anne
Brouwer, and Wilco van Nieuwenhuyzen. Jelle Koeman is responsible for housing and
technology within the Amsterdam UMC. Anne Brouwer and Wilco van Nieuwenhuyzen work
at the external companies Royal HaskoningDHV and Interflow, respectively. Both companies
have experience in the engineering of HVAC systems in hospitals. By predicting the energy
consumption after implementation using a Monte Carlo (MC) simulation, a difference can be
made quantifiable by comparing this to historical energy consumption. An MC simulation is
chosen as this is easy to implement. It can provide an approximately optimal solution to a
mathematical model. Additionally, stochastic data is included as the duration of surgeries for
example is probabilistic. An MC simulation can deal with uncertainty, and it has therefore
been chosen as the simulation method. This knowledge problem also underlines the relevance
of this research as it quantifies the problem. The distribution of the processing times of
surgeries is made as input for the MC simulation. This is done by using historical data
concerning surgery scheduling provided by “EVA stuurinformatie” from the AMC hospital.
Historical data concerning the year 2019 is selected as this year is pre-COVID-19. The
scheduling of the surgeries in this year was not affected by COVID-19 regulations.

Introduction 14
 - RQ6: How can the scheduling problem be formulated, and which heuristics should be used
for reducing carbon emissions while not hurting the efficiency of the OR?
This RQ supports RQ7 by researching which heuristics can be used when solving the scheduling
problem. The most effective heuristics are found using a literature review. Two heuristics will
be selected after having researched all models. “Loosely speaking, heuristic means to find or
to discover by trial and error (Yang, 2011). Examples of potential heuristics are “Earliest Due
Date” and “Shortest Processing Time”. A literature review is used when answering this
research question using databases such as Scopus, Google Scholar, and PubMed. The
heuristics will be applied to the OR scheduling problem of the Amsterdam UMC.

- RQ7: How can the planning of the HVAC system usage and the scheduling of the surgeries
in the ORs be optimally changed using a heuristic so that a reduction in energy consumption
is realized without compromising the efficiency of the OR while obeying the guideline
concerning air quality in the OR?
This is the main RQ, it concludes the research as solutions are generated. Afterward, possible
solutions are analyzed, and the best option is selected. The core problem is solved while
solving this knowledge problem, while all other knowledge problems support solving this
knowledge problem. After describing the scheduling problem using optimal job scheduling,
two heuristics will be used for the scheduling problem. These heuristics will be compared, and
the most effective heuristic will be selected. The solution will then be validated by comparing
it with the current situation.

2 Methodology
This chapter explains the research methodology by describing the contextual framework used in this
research. It consists of a problem-solving approach, a description of the intended deliverables, a
research design, and an assessment of the validity and reliability of the research.

2.1 Problem-solving approach

This section describes the problem-solving approach implemented during this research. The problem-
solving approach consists of 7 steps that are designed to answer the RQ. Each step is described and
explained in this section.

1. Problem identification and motivation

During this first stage, a clear overview of the problem is made. The most important aspect of this
phase is formulating the assignment. Knowing what the problem of the research is and why the
research should be conducted is crucial during research. A problem cluster is generated, and the action
problem, core problem, and knowledge problems are identified. The intended deliverables, key
constructs and variables, the theoretical perspective, and the research design are also identified.
Performing semi-structured interviews and making observations are the main activities at this stage.

2. Literature review
The second stage of the problem-solving approach concerns the execution of the literature review.
This stage is done according to the systematic literature review (SLR) protocol. The goal of the
literature review is to answer some knowledge problems. The knowledge problems can be found in
section 1.4: “Research questions”. RQ1 and RQ2 are answered entirely using a literature review and
these knowledge problems will therefore be solved during this stage. The knowledge problems RQ1

Methodology 15
and RQ2 are solved using documents including previous guidelines and current guidelines. RQ3, RQ4,
RQ5, and RQ6 will require some literature review and these knowledge problems will therefore also
be assessed. For RQ3, some general information regarding the purpose and functionalities of HVAC
systems is acquired using a literature review. RQ4 uses some information extracted from the literature
review regarding the allocation of surgeries to OR types. RQ5 includes calculations and some
assumptions might have to be made for this. A literature review is used for that. RQ6 requires the use
of documents where types of scheduling problems are explained. The OR scheduling problem of the
Amsterdam UMC has to be identified. Therefore, RQ6 cannot be answered completely using a
literature review.

This stage does not necessarily have to take place at the Amsterdam UMC because it relies on a
literature search. Databases such as Scopus, Google Scholar, and PubMed are used. However, internal
files from the Amsterdam UMC are also extracted and used for the literature review. The air
management plan is an example of an internal document from the hospital. Internal files are found in
the management software of the Amsterdam UMC called “Zenya”. Additionally, “EVA
stuurinformatie” should be able to provide data concerning for example all executed surgeries in 2019.

3. Analyze the current situation

During this stage, research is executed at the Amsterdam UMC hospital using observations and
interviews. For the remainder of RQ3, RQ5, and RQ6 both semi-structured interviews and observations
are executed at the hospital. RQ4 is answered during this stage entirely as the scheduling methods are
hospital-specific. Data is therefore acquired by both semi-structured interviews and observations as
well. The interviews are executed with the people introduced in section 1.4: “Research questions”.

4. Generate improvement measures

The remainder of RQ6 will be considered during this stage. Ideas are generated that potentially realize
a reduction in energy consumption while not hurting the efficiency of the clinical OR department. All
knowledge acquired using the previous RQs is considered during the generation of ideas in this stage.
Heuristics that can be used to solve the OR scheduling problem are selected using a literature review.

5. Analysis and selection of improvements

All the heuristics listed in the previous phase concerning RQ6 are analyzed. A literature review is
performed to analyze the performance of each of the proposed heuristics. After evaluating the
heuristics, the two best heuristics must be selected. The heuristics that are selected must apply to the
OR scheduling problem of the Amsterdam UMC.

6. Elaboration on improvement measure

This stage is also part of RQ6, and it concerns the elaboration of the selected heuristic in stage 5. A
detailed explanation of the selected solution is written during this stage.

7. Write conclusion and recommendations

A concise conclusion is written at this stage, where the most important findings resulting from the
research are stated. A summary of the core problem, the action problem, and each knowledge
problem are given. Recommendations for the Amsterdam UMC hospital are given concerning the
planning of the HVAC systems usage and the scheduling of surgeries. RQ7 is answered at this stage.

Methodology 16
2.2 Intended deliverables
The problem owner Amsterdam UMC aims to make the clinical OR department more sustainable by
decreasing the energy consumption of the HVAC systems in the ORs. To realize a reduction in energy
consumption, potential improvements regarding the scheduling of the surgeries are researched. This
research is intended to result in the following deliverables:

- A heuristic that can be implemented for surgery scheduling

- Insights into energy consumption before and after implementation of the heuristic using an
MC simulation
- Other recommendations regarding the organization of surgery scheduling and sustainability
of the clinical OR department

An effective heuristic is delivered to the Amsterdam UMC. This heuristic should be ready for
implementation in the clinical OR department of the hospital. Heuristics that cannot be applied to the
Amsterdam UMC hospital are therefore outside the scope of this research. Insights into the current
energy consumption of the HVAC systems in the clinical OR department and the expected impact of
the proposed heuristic on the energy consumption are delivered to the problem owner. The current
energy consumption and the impact on the energy consumption are estimated using an MC
simulation. Other recommendations regarding the sustainability of the clinical OR department and
surgery scheduling are also delivered, as well as potential further improvements to the heuristic. A
comprehensive thesis describes all the intended deliverables.

2.3 Research design

This research was started by executing exploratory research to get a full understanding of the topic.
The exploratory phase contained interviews, observations, and a literature review. The observations
were performed by receiving guides around the ORs and by spending time in the Amsterdam UMC
hospital. The interviews were semi-structured, which means that important elements that should be
elaborated on will be structured but the rest of the interview will not have a particular structure. After
having obtained a clear understanding of the processes at the Amsterdam UMC hospital, descriptive
research was performed as the current processes, guidelines, and emissions were analyzed and
described. In this phase, the topic was described in detail using academic theory. For the prediction of
the environmental impact of the ORs after implementation of the proposed heuristic, an MC
simulation is made. This MC simulation can deal with the uncertainty of implementing the
recommended solution. Interviews were also performed in the descriptive research phase.

2.3.1 Data collection

This research combined qualitative and quantitative research by answering both qualitative and
quantitative research questions. The method of combining these two data gathering and analysis
methods is called “triangulation”. Knowledge problems concerned with energy consumption are
quantitative as the consumption is quantified in these elements of the research. The rest of the
research is qualitative, as nonnumerical research is done there. Qualitative research was done by
performing interviews and literature reviews, while the quantitative elements are researched by
modeling the MC simulation. Data is therefore analyzed both qualitatively and quantitatively. Both
methods are necessary during this research as it is important to understand the processes and
guidelines to be able to recommend a solution. However, energy consumption also had to be
calculated to predict the reduction in energy consumption and compare this to the current situation.
That is, the solution should be verified by performing quantitative research.

Methodology 17
2.3.2 Limitations
Every research design, including this research design, contains some limitations. These limitations
must be identified to limit their impact on the study. This research design consists of both qualitative
and quantitative elements which might make the study too broad. It is therefore important to
formulate the scope of this research to narrow down the research. Additionally, the qualitative
element in this research requires numerical data. A potential limitation is that this data is unavailable
for the specific situation at the Amsterdam UMC hospital. This data should in that case be acquired by
observing and researching the ORs. Finally, this research has a period of ten weeks. This potentially
limits the scope of the research to finish in time. The scope of the research is narrowed down to still
provide added value. Data that is studied was data from a maximum of one year and solutions that
are not related to optimizing the scheduling of surgeries will not be assessed.

2.4 Validity and reliability of measurement

2.4.1 Validity
“Validity is the extent to which a test measures what we actually wish to measure (Cooper & Schindler,
2014). Additionally, validity can be divided into two forms: External- and internal validity. External
validity concerns whether data generalization is justifiable. This means that data that is acquired
during this research should apply to the research sufficiently. Internal validity assesses the extent to
which the research instruments are fit to measure what they are used for. In this section, the validity
is divided into three different forms: content validity, criterion-related validity, and construct validity.

Content validity
“The content validity of a measuring instrument is the extent to which it provides adequate coverage
of the investigative questions guiding the study” (Cooper & Schindler, 2014). This means that all
elements of the research questions should be covered, with sufficient elaboration concerning the
topic. Each knowledge problem should therefore be researched extensively, making sure all elements
are elaborated on. For each knowledge problem, elements that should be researched should be
determined before researching so that no elements are missed. To ensure content validity, this
research provides background information for every knowledge problem. After this, the research will
contain specific information regarding the research question.

Criterion-related validity
This type of validity “reflects the success of measures used for prediction or estimation” (Cooper &
Schindler, 2014). In this research a simulation will be made, predicting the energy consumption after
implementing a solution. This research has predictive validity if this simulation correctly predicts
energy consumption after implementation. Modeling should therefore be performed correctly and
considerate to ensure criterion-related validity.

Construct validity
Construct validity is the extent to which a test measures the construct it claims to measure. This means
that constructs should be clearly defined to prevent confusion. This will ensure construct validity for
this research. Valid calculations must be made during this research, and correct distribution must be
used as input for the MC simulation.

2.4.2 Reliability
“Reliability has to do with the accuracy and precision of a measurement procedure” (Cooper &
Schindler, 2014). It contributes to the validity of research; however, it does not ensure validity. During
this research, important experiments are performed multiple times to ensure stability.

Methodology 18
3 Literature review
A literature review is conducted using databases such as Scopus, Google Scholar, Pubmed, and Zenya.
Zenya is a database containing documents originating from the Amsterdam UMC. The literature
review elaborates on the previous guideline on air treatment of ORs, the most recent guideline on air
treatment of ORs, and optimal job scheduling.

3.1 Previous guidelines concerning the air treatment of operating rooms in

Dutch hospitals
We currently have a knowledge problem as the previous guidelines of 2014 are unknown and it should
therefore be researched. The previous guidelines should be compared to the most recent guidelines
to obtain insight into the processes concerned with OR planning at the Amsterdam UMC. The
guidelines concerned with internal air quality (IAQ) are to be identified for different OR types and
surgeries, and only Dutch guidelines are within the scope of this RQ. Two guidelines concerning air
treatment apply to the air management of the clinical OR department of the hospital, as described in
Luchttechnischbeheersplan OK (Amsterdam UMC, 2018). These guidelines are the Werkgroep Infecite
Preventie (WIP) guidelines Luchtbehandeling in operatiekamer en opdekruimte in operatieafdeling
klasse 1 (WIP, 2014) and Omstandigheden (kleine) chirurgische en invasieve ingrepen (WIP, 2011).

3.1.1 Introduction
An OR’s state is distinguished into three states. These three states are labeled: “As-built”, “At rest”,
and “Operational”. “As-built” is referred to as the OR as delivered including the fixed equipment but
excluding medical equipment and persons present in the room. “At rest” is the OR with a set-up ready
for use with all equipment present, but without people in the OR. Finally, “Operational” is the situation
where the OR is used, with the prescribed number of people present in the OR. Additionally, a
distinction is made between class 1 and 2 ORs. Next to the ORs of classes 1 and 2, independent
treatment rooms also exist. However, only class 1 ORs are within the scope of this research as the
scheduling of the clinical OR department will be optimized. Within class 1, a distinction is made
between performance levels 1 and 2. These performance levels do not strictly correspond to the two
types of HVAC systems to prevent potential obstacles when striving for innovation concerning the
HVAC systems. As SSIs are largely caused by contaminated particles carrying micro-organisms, SSI
prevention mainly contains the minimization of particles in the OR (Chauveaux, 2015). ISO levels
classify the number of particles per 𝑚  .

There are currently two types of HVAC systems: A unidirectional Flow (UDF) and a turbulent HVAC
system. A UDF HVAC system vertically ventilates the protected area, causing the micro-organisms to
flow horizontally across the floor towards the walls, where suction is applied. A turbulent HVAC system
does not have a protected area. Here, the IAQ is the same in the entire OR. For the performance of
UDF HVAC systems, the degree of protection within the protected area and the recovery time of the
protected area are performance measures. The degree of protection depends on the difference
between the number of particles per 𝑚  within the protected area and the number of particles per
𝑚  outside the protected area. For turbulent HVAC systems, the degree of protection is not applicable
as there is no protected area. The performance of turbulent HVAC systems is therefore singularly
assessed by using the recovery time of the entire OR. The recovery time is defined as: “The time it
takes after an increase in the concentration of particles with a size equal to or greater than 0.5 µm, to
lower the concentration by a factor of 100 compared to the 'at rest' situation (WIP, 2014). Additionally,
the degree of protection is calculated using equation 1 (Traversari et al., 2019):

Literature review 19
 𝑪𝒙
𝑫𝑷𝒙 = −𝒍𝒐𝒈(𝑪 ) (1)
𝒓𝒆𝒇

, where:

𝐷𝑃𝑥 = Degree of protection at place x

𝐶𝑥 = Concentration of particles at place x (within protected area) [particles/m3 ]
𝐶𝑟𝑒𝑓 = Reference concentration of particles (outside protected area) [particles/𝑚  ]

Additionally, both the ISO class and the air permeability are performance requirements for ORs. ISO 5
means that a maximum of 3,520 particles/𝑚  with a size equal to or greater than 0.5 µm may be found
'at rest' in the protected area. ISO 7 corresponds to a maximum of 352,000 particles/𝑚  with a size
equal to or greater than 0.5 µm may be found 'at rest' in the protected area (WIP, 2014). Air
permeability is defined as the “air leakage rate per envelope area at the test reference pressure
differential across the building envelope” (NEN, Stichting Koninklijk Nederlands Normalisatie Instituut,
2015).

3.1.2 Performance, construction, and other guidelines HVAC system of class 1 ORs
Performance requirements HVAC system
Table 3-1: Performance requirements class 1 ORs “At rest” summarizes the performance requirements
of ORs “At rest” (WIP, 2014). As can be seen, the degree of protection only applies to performance
level 1. In the table, UDF HVAC systems are not necessarily referred to as “performance level 1”, while
turbulent HVAC systems are also not strictly referred to as “performance level 2”.

Table 3-1: Performance requirements class 1 ORs “At rest”

ORs with performance level 1 must have a degree of protection of at least 3.0 at the center of the
protected area, while the extreme edges of the protected area must have a degree of protection of
2.0 or more. The recovery time should be under minutes in the center of the protected area, and the
OR should have ISO 5 or better. ORs classified as performance level 2 should have a recovery time of
maximum 20 minutes while maintaining an ISO class of 7 or better. Both ORs with performance levels
1 and 2 should have a maximum air permeability of 1.5 times the OR’s volume.

Construction guidelines
The OR should be provided with an HVAC system that should not depend on the hospital's ventilation
system to provide clean air. This minimizes the chances of system failure. The ORs department needs
to be separated from the rest of the hospital using airlocks. The airlocks consist of an area between
two doors. These two doors cannot be opened at the same time to make differences in air pressure

Literature review 20
possible. This is important as the airflow should be directed towards the zone with the lowest air
quality. Figure 3-1 provides a schematic overview of the clinical OR department (WIP, 2014). The
airflow direction is controlled by maintaining increasing pressure levels from zone D toward zone A.
Appendix B depicts an overview of the pressure levels in the clinical OR department of the Amsterdam
UMC (Amsterdam UMC, 2018). The walls around the OR department are also not allowed to have any
openings in them to prevent unwanted outside air from entering the clinical OR department. An
exception is made for emergency exits.

The air that is provided to zone A, B, and C of the OR department should have an air quality of at least
F9 filtered air. An F9 air filter is designed for HVAC usage and provides a high level of air filtration. The
air that is provided to the OR itself should be filtered using an H13 filter. Additionally, the air coming
from zone A may only be circulated within zone A. This is important as the circulation of air cause a
risk of surgical site infection (SSI) occurrence. By excluding recirculation of air between different ORs
in zone A, the air from different ORs cannot mix and the risk of SSI occurrences caused by the air via
the HVAC system is reduced. Additionally, the temperature of the airflow in zone A originating from
the HVAC systems preferably is controlled using an automized central system. The reason for this is
that changing the temperature of this airflow might temporarily affect the performance of the HVAC
system.

Figure 3-1: Schematic overview of the OR department with class 1 ORs

The floor, walls, and ceilings should be smooth, seamless, and closed off to prevent dirt accumulation
and to make effective cleaning possible. Zone A is only allowed to have openings that are connected
to zone B. Additionally, prevent particles (dust) from entering Zone A from the ceiling. This can
preferably be prevented by implementing a box-in-box construction for Zone A.

Other guidelines
Provide Zone A within a class 1 OR department with a dashboard that monitors and indicates whether
the HVAC system is functioning properly. Some key performance indicators (KPIs) such as room
temperature and relative humidity (RH) are shown here. This dashboard prevents the execution of
surgeries at times when the HVAC system does not function properly causing some KPIs to exceed
their limit values. Additionally, equipment should be used that does not unnecessarily interfere with
the airflow of the UDF HVAC system. Equipment used in the OR might produce heat causing upward
airflow. This upward airflow results in clean air mixing with the polluted air. Air quality in ORs using a
turbulent HVAC system is not affected heavily by equipment as these ORs’ air quality does not depend
on the direction of the airflow to the extent of UDF HVAC systems. Finally, the protected area in an

Literature review 21
OR using a UDF HVAC system should be marked on the floor. This area in the OR should be at least
large enough so that the operating table, the instrument tables, and the operating team can be placed
within the protected area during a procedure. The OR should also preferably have a minimum distance
of 1.5 meters between the walls of the OR and the extreme edges of the protected area. Dutch
hospitals should have an air management plan, describing the HVAC system, the quality management
system related to air quality at the ORs, the maintenance procedures, limit values of the KPIs, etc. A
document indicating which interventions may be performed under which conditions and indicating a
classification of surgeries are also mandatory (Inspectie voor de Gezondheidszorg, 2016).

HVAC system usage guidelines

The display in the OR should indicate whether important KPIs are within their limit values. These values
should be established using the documentation of the HVAC system. Continuous monitoring of the
KPIs should be executed automatically. The OR should only be used when all these values are within
their limit values. Clear procedures on how to act if the KPI values are outside their limit values should
also be established. Additionally, procedures should be established concerning the release of the OR
at the start of the working day. Here, the system should be checked for any errors and functionality.
However, procedures should also be made concerning OR usage outside regular working times.

The patient must be placed so that the operating area, surgical team, tables, and sterile instrument
tables are in the protected area of the OR. The operating lamp should not be placed horizontally below
the HVAC system, disrupting the airflow. Additionally, the air temperature of the air provided by the
UDF HVAC system should only be altered by strict necessity. Air temperature is a crucial KPI for the
effective functioning of the HVAC system. Changing the temperature has a temporary negative effect
on the air quality, as a new balance needs to be formed. Finally, the patient’s body temperature should
be between 36 degrees Celsius and 38 degrees Celsius, as this reduces the chances of SSI occurrence.
Heating blankets or mattresses can be used for the preservation of patients’ body temperature. These
products do not have a proven negative effect on air quality within the OR (WIP, 2014).

The number of people present in the OR should be limited. People radiate heat and micro-organisms,
and they negatively influence the airflows in the OR. Only OR clothing should be worn in the OR, as
clothing plays an essential role in the emissions of micro-organisms. During surgery, the door of the
OR should not be opened if this is not necessary. It is advised to preferably provide each OR with a
separate restroom with at least comparable performance requirements for the air treatment as for
the associated OR (WIP, 2014). It must be prevented that instruments that must be sterile become
contaminated, which could increase the risk of SSI occurrences.

HVAC type recommendation

Factors that influence the selection of a vertical UDF HVAC system or a turbulent HVAC system include
the number of SSIs; the number of bacteria in the air and above the wound while the OR is in use; the
degree of contamination of the wound; the cost, the patient's body temperature and the effect of
heating blankets on both systems (WIP, 2014).

Table 3-2 shows the recommendations as proposed by WIP (2014) concerning the performance level.
Arthroplasty surgery and procedures with implantation of (endo)prostheses are recommended to be
performed in an OR with performance level 1 and not in an OR with performance level 2. All other
surgeries do not have a pronounced recommendation concerning a performance level 1 or 2 (WIP,
2014). These recommendations result in the hypothesis that an OR equipped with a UDF HVAC system
generally has fewer bacteria in the air as opposed to an OR equipped with a turbulent HVAC system,
as UDF HVAC systems are more frequently used for performance level 1 ORs compared to turbulent
HVAC systems.

Literature review 22
The hypothesis that an OR equipped with a UDF HVAC system generally has fewer bacteria in the air
as opposed to an OR equipped with a turbulent HVAC system is confirmed by R.J. Knudsen et al. (2021).
Their research counted colony-forming units (CFUs) during live surgery in ORs with a UDF HVAC system
and ORs with a turbulent HVAC system. These units are used to estimate the number of bacteria in
the air.

Table 3-2: Performance level recommendations

They concluded that ORs equipped with a UDF HVAC system had lower CFU counts compared to ORs
equipped with a turbulent HVAC system. ORs with a UDF HVAC system were able to provide ultra-
clean air, while turbulent HVAC systems failed to provide ultra-clean air (Knudsen et al., 2021). This
fact is also confirmed by WIP (2014), who stated: “The scientific research shows that when a UDF
HVAC system is used in the OR, fewer micro-organisms are present in the air and there is less
contamination of instruments than with a turbulent HVAC system. This is not only “At rest” but also
when the OR is in operation.” However, this research limits itself to surgeries concerning knee and hip
prostheses. The researchers add: “We conclude that there is unambiguous support in the literature
for the fact that UDF leads to less contaminated air above the surgical site and less contamination of
instruments during procedures and that there is inconclusive evidence for a decrease in the number
of SSIs in the application of UDF.” By this, they mean that UDF HVAC systems lead to better air quality.
However, there is no clear evidence found that these systems also lead to fewer SSIs.

However, we conclude that a UDF HVAC system protects the protected area in an OR better against
micro-organisms than a turbulent HVAC system and is therefore superior. Using a UDF HVAC system
also makes it easy to indicate a clear protected area. Some potential disadvantages of a UDF HVAC
system compared to a turbulent HVAC system are listed below (WIP, 2014):

- the technical complexity of the design, installation, maintenance, and management;

- the impediment to the functionality of the HVAC system caused by objects that are placed in
the airflow such as the operating lamp and screens;
- effect of operating and user errors when the temperature in the OR changes – the airflows
have to stabilize again. If one starts operating immediately after the temperature change, the
air in the OR usually does not meet the requirements for the protected area;
- overestimation of the importance of the UDF, with underestimation of the other measures.
These potential theoretical disadvantages have been considered. However, they do not compensate
for the advantages of UDF HVAC systems as opposed to turbulent HVAC systems according to WIP
(2014). Dutch hospitals should therefore have at least one OR equipped with a UDF HVAC system, or
they should have advanced plans to implement it.

Literature review 23
3.1.3 Guidelines of class 2 ORs compared to class 1 ORs
During the remainder of this section, guidelines concerning class 2 ORs are briefly compared to the
guidelines concerning class 1 ORs. As the Amsterdam UMC does not have class 2 ORs, this comparison
will not be detailed. Table 3-3 gives an overview of some general differences between the guidelines
(WIP, 2011).

Table 3-3: Overview of general differences in the guidelines between class 1 ORs
and class 2 ORs1

The layout of class 2 ORs is different, as different zones and airlocks can be identified. An OR with class
2 requires only two separate zones. The connections between these zones are not as strict in class 2
ORs, as only personnel airlocks are required. Zone B does not need to have overpressure compared to
zone C for class 2 ORs. Zone B of the OR department with class 2 ORs integrates zone B and C of the
class 1 OR department (WIP, 2011). Check Figure 3-2 for a schematic overview of an OR department
with ORs of class 2.

Figure 3-2: Schematic overview of the OR department with class 2 ORs

Literature review 24
Commissie kwaliteitsdocumenten NVOG (2020) concludes that class 2 ORs emit significantly smaller
amounts of CO2 compared to class 1 ORs mainly because standard ventilation systems consume a
smaller amount of energy compared to UDF HVAC systems or turbulent HVAC systems. The energy
consumption of an OR can be reduced by 70% if the ventilation rate is reduced from 30 to six times
per hour. Additionally, six air changes per hour are sufficient to maintain ISO 7, even in class 2 ORs.
Figure 3-3 verifies this (Commissie kwaliteitsdocumenten NVOG, 2020).

Figure 3-3: The influence of the number of air changes on the number of particles
found. The blue dotted line corresponds to the ISO 7 standard

Literature review 25
3.2 Changes in the revised guidelines concerning the air treatment of
operating rooms in Dutch hospitals compared to the previous guidelines
Recently published changes in the guidelines have been published initiated by Nederlandse Vereniging
voor Medische Microbiologie (NVMM). The guideline generally remains the same. However, some
crucial changes have been made. This section elaborates on the most important changes between the
revised guidelines and the previous guidelines described in the previous section. The newest guideline
Richtlijn Luchtbehandeling in operatiekamers en behandelkamers (NVMM, 2022) aims to reduce the
discrepancy and confusion around the guideline (I. Spijkerman, personal communication, June 22,
2022) (See Appendix C.1: Interview with Ingrid Spijkerman (Amsterdam UMC) for the interview). To
achieve this, a new guideline is composed and for example, the recommendations regarding the HVAC
system selection have been altered.

3.2.1 Recommendations regarding HVAC system

As explained in the previous section, UDF HVAC systems are more effective at limiting the number of
micro-organisms in the OR as opposed to turbulent HVAC systems. However, no evidence is found
that these systems also lead to fewer SSI occurrences. The international guidelines World Health
Organization (2016), Centers for Disease Control and Prevention (2017), and National Institute for
Health and Care Excellence (2020) also state that research in this area has serious shortcomings and
that this low to very low level of evidence suggests a choice for equipping an OR with a UDF HVAC
system or a turbulent HVAC system is not supported by research (NVMM, 2022). More, but especially
qualitatively better research into the effect of different HVAC systems on SSIs is urgently needed. The
guidelines in for example the UK, the USA, Germany, and Belgium do not recommend a specific HVAC
system an OR should be equipped with (NVMM, 2022). The revised guideline in the Netherlands
abandons the recommendation of UDF HVAC systems because there is no evidence that these systems
prevent SSIs more effectively as opposed to turbulent HVAC systems.

3.2.2 Revised classification

The revised guideline proposes a new system to classify HVAC systems in ORs as class 1+, class 1, or
class 2. The classifications “performance level 1” and “performance level 2” have been removed. To
prevent SSIs as effectively as possible, ORs should be equipped with an HVAC system that (regardless
of the type of HVAC system) at least meet the minimum criteria stated in Table 3-4. Additionally, for
elective major joint replacement operations (knee, hip, shoulder), an OR with ultra-clean air is
preferred. Notice that this was already the case in the previous guideline, as we see in Table 3-2. That
is, an OR that meets the criteria of an OR class 1+ described in Table 3-4 is prescribed (NVMM, 2022).

Table 3-4: Minimum criteria for ORs in the revised guideline

NVMM (2022) also provides a classification of surgeries in accordance with every specialism in the
revised guideline. The table for classification of interventions was completed by the scientific
associations of the accompanying specialism. This extensive surgery classification can be seen in
Appendix D: Surgery classification as proposed in the revised guideline.

Literature review 26
3.3 Optimal job scheduling
To formulate the surgery scheduling problem, we use optimal job scheduling. We view the surgeries
as independent jobs that must be scheduled on one of the machines. In this case, ORs are referred to
as machines in optimal job scheduling. After having formulated the scheduling problem using optimal
job scheduling, we can search for effective heuristics to optimize surgery scheduling.

Optimal job scheduling is an optimization problem related to scheduling processes using one or
several machines. It refers to a set of tasks that must be performed under different configurations of
production resources, such as a single machine or multiple machines (Azimpoor & Taghipour, 2020).
A typical optimal job scheduling problem aims to find a schedule that optimizes an objective function
using a list of jobs and machines. A scheduling problem might be either a single-stage or a multi-stage
schedule problem. In a single-stage optimal job scheduling problem, each job requires only a singular
phase. Each job in a multi-stage optimal job scheduling problem requires multiple phases before
completion. These phases can be carried out either in sequence or in parallel. The types of job
scheduling problems will now be discussed (Graham et al., 1979).

3.3.1 Single-stage job scheduling problems

- 1: Single-machine scheduling
Single-machine scheduling is an optimization problem where 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different
processing durations that need to be scheduled on a single machine.

- P: Identical-machines scheduling
This optimization problem concerns where 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different processing
durations that need to be planned on 𝑚 identical machines. Job 𝑗 has a processing duration
denoted by 𝑝𝑗 on any of the machines.

- Q: Uniform-machines scheduling
The uniform-machines scheduling problem concerns 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different
processing times that need to be scheduled on 𝑚 machines. Each machine 𝑖 has a factor
representing its relative speed 𝑠𝑖 and every job 𝑗 has an individual processing duration on a
𝑝𝑗
specific machine 𝑖 of 𝑝𝑖,𝑗 . 𝑝𝑖,𝑗 is calculated by: 𝑝𝑖,𝑗 = 𝑠𝑖
.

- R: Unrelated-machines scheduling
𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 need to be scheduled on 𝑚 different machines. The time that machine 𝑖
needs for job 𝑗 is denoted 𝑝𝑖,𝑗 . There is no relation between the values of 𝑝𝑖,𝑗 for the different
jobs and machines.

3.3.2 Multi-stage job scheduling problems

- J: Job-shop scheduling
This optimization problem concerns 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different processing durations that
need to be planned on 𝑚 machines with varying processing speeds. Every job consists of
different operations 𝑂1 , 𝑂2 , … , 𝑂𝑛 . These operations need to be processed in a specific order
in a job-shop scheduling problem. Additionally, every operation can only be handled by a
dedicated machine, and no more than one operation in a job can be executed simultaneously.

Literature review 27
 - F: Flow-shop scheduling
In this case, there are 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different processing durations that need to be
planned on 𝑚 machines with varying processing speeds. Every job consists of different
operations 𝑂1 , 𝑂2 , … , 𝑂𝑛 . These operations also need to be processed in a specific order. Also,
only a single operation can be performed by each machine simultaneously and the next
operation can only be started before the previous operation is finished. There is not one
dedicated machine for a specific operation.

- O: Open-shop scheduling is a similar problem but also without the order constraint
The open-shop scheduling problem concerns 𝑛 jobs 𝐽1 , 𝐽2 , … , 𝐽𝑛 with different processing
durations that need to be planned on 𝑚 machines with varying processing speeds. Every job
consists of different operations 𝑂1 , 𝑂2 , … , 𝑂𝑛 . These operations can be scheduled in any order
and on any machine for this specific scheduling problem.

3.3.3 Job data

The following data concerning optimal job scheduling problems can be identified for each job 𝐽𝑗 :
- The completion time 𝐶𝑗 is the point in time at which a job 𝐽𝑗 is finished;
- The processing time 𝑝𝑗 or 𝑝𝑖𝑗 is the duration of completing job 𝐽𝑗 on machine 𝑚𝑖 , either
machine-independent or machine-dependent, respectively;
- The release date 𝑟𝑗 is the moment job 𝐽𝑗 becomes available for processing;
- The due date 𝑑𝑗 is the moment at which job 𝐽𝑗 must be finished;
- The setup time 𝑠𝑗 is the duration required for machine preparation before job 𝐽𝑗 can start.

3.3.4 Optimality criteria

When solving a scheduling problem, a particular objective must be achieved. Several objectives can
be discovered when performing a literature review. These objectives, or optimality criteria, can for
example be the following:
- 𝐶𝑚𝑎𝑥 : The maximum completion time, or makespan, can be minimized;
- 𝐶̅ : The average completion time can be minimized;
- 𝐿𝑚𝑎𝑥 : The maximum lateness can be minimized. 𝐿𝑗 = 𝐶𝑗 − 𝑑𝑗 ;
- ∑ 𝑇𝑗 : The tardiness represented by the sum of 𝑇𝑗 ∀ 𝑗 can be minimized. 𝑇𝑗 = max{0; 𝐶𝑗 − 𝑑𝑗 };
- ∑ 𝐹𝑗 : The total flow time represented by the sum of 𝐹𝑗 ∀ 𝑗 can be minimized. 𝐹𝑗 = 𝐶𝑗 − 𝑟𝑗 ;
- ∑ 𝑈𝑗 : The number of jobs that are completed before their deadline represented by the sum of
𝑈𝑗 ∀𝑗 can be maximized. This is called the throughput;
- 𝐸𝑚𝑖𝑛 : Maximize the minimum earliness. The earliness is calculated using 𝐸𝑗 = max{0; 𝑑𝑗 −
𝐶𝑗 }.

3.3.5 Notation for optimal job scheduling problems

This research uses the three-field problem classification notation as described by Graham et al (1979).
This notation describes a scheduling problem using a three-field notation denoted α|β|γ, where:

- α is used to describe the type of scheduling problem;

- β is used to describe the job characteristics;
- γ is used to describe the optimality criteria.

Literature review 28
For example, 1|𝑑𝑗 |𝐿𝑚𝑎𝑥 is a single-machine scheduling problem with given due dates for each job
where the aim is to minimize the maximum lateness.

4 Evaluation of the Amsterdam UMC

This chapter elaborates on the case study. The data collected in this chapter originates mainly from
observations and interviews but also using a literature review found in the “Zenya” database from the
Amsterdam UMC. Topics such as scheduling, HVAC systems, and energy consumption are discussed in
this chapter. As the Amsterdam UMC has not implemented the new guideline thus far, we will use the
old OR classification (e.g., class 1 performance level 1) in the evaluation of the Amsterdam UMC.

4.1 Scheduling
Each surgery is classified and scheduled in either a class 1 OR or in an independent treatment room.
Another classification is made regarding the urgency of surgeries. This requires systematic planning
strategies. This section elaborates on the surgery process and classifications of surgeries used by the
Amsterdam UMC. Additionally, their current planning is explained.

4.1.1 Surgery process description

The process of carrying out surgery on one patient is depicted in Figure 4-1 using a Business Process
Model Notation (BPMN) model. The BPMN model consists of four pools representing the
corresponding department. These departments are the nursing department, holding, OR, and the
recovery room. Every activity depicted in the BPMN model represents a stage to be registered in the
information system EPIC.

Figure 4-1: BPMN model of an individual surgery process

The process starts in the nursing department, where the patient is called. The patient is then taken to
the waiting room in the holding. After this, the patient is brought to the OR. The surgery then generally
consists of five stages: Start anesthesia, release by the anesthetist, incision, wound closed, and the
termination or end of the surgery. The surgeon can start preparing for the surgery as soon as the
patient is released by the anesthetist. This can occur before the anesthetist has finished the

Evaluation of the Amsterdam UMC 29

anesthesia. Preparing the patient consists of positioning the patient, applying iodine compounds, and
covering the patient. The incision is then started, and the wound is closed afterward. The surgery is
then finished, and the patient is brought to the recovery room. After having recovered, the patient is
taken to the nursing department or intensive care. These stages are not registered in EPIC and are
thus not depicted in the BPMN model. Two points in time where the patient arrives and departs the
OR must also be registered in EPIC. This sequence of steps is executed repeatedly in all 20 ORs.

We define the processing time as the duration between the points in time where the patient arrives
at the OR and leaves the OR. The time difference between the point in time where a patient leaves
the OR and where the next patient arrives at the OR is the setup time. However, in this definition, we
assume that the idle time is equal to zero. Figure 4-2 visualizes the processing time and setup time if
we assume the idle time to be equal to zero.

Figure 4-2: Processing time and setup time

4.1.2 Current OR classification of surgeries

The guidelines do not prescribe the type of OR that should be used for each type of surgery, because
it is not possible to give a strict classification of interventions according to the circumstances under
which they must be performed based on the literature. Only an example of surgery classification and
allocation is given in the guideline (Check Appendix E for the example that is provided) (WIP, 2011).
Note that this is solely an example and that it is not used in practice. However, an institution that has
several OR types as described in the guideline where interventions can be performed, must explicitly
describe which interventions may be performed at which OR(s). That is, each hospital should make a
classification of surgeries and which ORs can be used for the type of surgery individually. The following
criteria must be considered when deciding where to perform which interventions:

- the size of the incision;

- the depth of the incision;
- the duration of the procedure;
- implantation of a foreign material;
- opening sterile cavities, bones, or large joints;
- the consequences of an SSI for the patient.

As mentioned in 3.1.2: “Performance, construction, and other guidelines HVAC system of class 1 ORs”,
every hospital is obliged to compose an air management plan, a document indicating the requirements
for surgery allocation, and a classification of surgeries. Van der Steen – de Vlaming (2022) composed
a document for the Amsterdam UMC indicating a classification of surgeries and which interventions
may be performed under which conditions (treatment room, OR class 1 performance level 1, OR class
1 performance level 2). Notice that class 2 ORs are not included in the classification because there are
no class 2 ORs in the Amsterdam UMC hospital, as mentioned in 1.1: “Problem owner description”.
The following classification and allocation are described by Van der Steen – de Vlaming (2022):

Evaluation of the Amsterdam UMC 30

 - OR class 1 performance level 1
The following surgical procedures may take place in the ORs class 1 performance level 1:
 All surgical procedures may be performed in the clinical ORs, which are all class 1
performance level 1.

- OR class 1 performance level 2

The following surgical procedures may take place in the ORs class 1 performance level 2:
 Intraocular procedures.

The following surgical procedures may not take place in the ORs class 1 performance level 2:
 Total hip- and knee arthroplasty;
 Breast implants;
 Central vascular prostheses;
 Surgical procedures in which an endoprosthesis or implant is inserted, which in terms of
SSI risk and consequences are comparable to a joint replacement.

- Treatment room
 No surgical procedures may take place in a treatment room.

We conclude that the Amsterdam UMC hospital adheres to the performance level recommendations
in the guidelines composed by WIP (2014) (Table 3-2), as surgeries concerning joint replacement or
procedures with implantation of (endo)prostheses must be performed in clinical ORs classified as class
1 performance level 1. This was also prescribed in the guideline discussed in 3.1.2: “Performance,
construction, and other guidelines HVAC system of class 1 ORs”. However, all surgeries are currently
executed in class 1 performance level 1 ORs in the clinical OR department. This is not necessary. ORs
are also equipped with different equipment, which complicates the scheduling of surgeries.

4.1.3 Urgency classification of surgeries

Some surgeries are prioritized over other surgeries in the Amsterdam UMC. Acute and emergency
surgeries for example have more urgency than elective surgeries. Prioritizing a surgery over another
surgery means scheduling the prioritized surgery earlier. An urgency classification is made to
determine the priority of every surgery. The Amsterdam UMC uses urgency classes S1, S2, S3, and S4
(Steen - de Vlaming & Hamersveld, 2021). We define the urgency classes as:

- S1: Acute, patient must be on the operating table within one hour at the latest.
- S2: Urgent, patient must be on the operating table within six hours at the latest.
- S3: Semi-urgent, patient must be on the operating table within 24 hours at the latest.
- S4: (Semi-)elective, patient must be on the operating table within between 24 and 72 hours.
Surgery is not scheduled outside business hours.

There is no nationally used classification system, but local systems are used per hospital or region. The
Amsterdam UMC applies the surgery urgency classification as proposed by Werkgroep Nederlandse
Vereniging voor Heelkunde (NVVH) (2017). An extensive urgency classification over every surgery can
be found on their website. This classification is formulated and clustered by the following specialisms:

- Surgery - Otolaryngology surgery

- Pediatric surgery - Oral and maxillofacial surgery

Evaluation of the Amsterdam UMC 31

 - Neurosurgery - Thorax surgery
- Obstetrics and gynecology - Traumatology & orthopedic surgery
- Ophthalmology - Urology
- Plastic surgery - Vascular surgery
- Radiology

4.1.4 Currently implemented scheduling

The planning must be within the established (financial) budget of the OR department and the
specialisms. The OR planning and control at the Amsterdam UMC is done at a strategic level, tactical
level, and operational level (Steen - de Vlaming & Hamersveld, 2021). The management levels are
distinguished by different planning horizons, different frequencies, and differences in the detail level.
The strategic level is involved with monitoring and controlling the organization. Long-term decisions
are made which are often annually. The decisions made on a strategic level impact the decisions on a
tactical level. The tactical level consists of monthly or weekly decisions and this level must implement
the decisions on a strategic level. The operational level concerns controlling daily operations. The
management at the tactical level is informed about potential problems (Kaplan Financial Limited,
2020).

Strategic level
The annual planning is determined by the management at a strategic level of the Amsterdam UMC.
The annual planning includes the planning of the budgeted OR capacity for each specialism for that
year (K. de Vlaming, personal communication, June 30, 2022) (See Appendix C.3: Interview with Karin
de Vlaming (Amsterdam UMC). The annual plan is released no later than two months before the start
of the new year. Approximately 30% of this planning consists of stochastic OR-hours. These stochastic
hours are allocated two months in advance in the monthly planning. This allocation depends on
available personnel and unforeseen circumstances (Steen - de Vlaming & Hamersveld, 2021).

Tactical level
On the tactical level, the monthly planning and the weekly planning are formulated. The monthly
planning includes the final planning of the OR capacity concerning the corresponding month. The
stochastic OR-hours and the reserved OR-hours as described in the annual planning are definitively
assigned depending on available personnel and unforeseen circumstances. This planning must be
released at least two months in advance. The weekly planning for the next week is determined each
Thursday at 12:00. Each specialism is responsible for scheduling surgeries within the released OR
capacity in the weekly planning. Before the weekly planning on Thursday, all specialisms have a
multidisciplinary meeting. During this meeting, the past week is reflected on and the planning for the
next week is discussed. The reflection aims to identify potential medical and logistical bottlenecks. The
planning of surgeries is executed in the system Optime by EPIC, and the weekly planning is based on
historical data concerning the processing time of surgeries per operator, procedure, and anesthesia
type. The concept of the weekly planning of every specialism must be available in EPIC Optime on
Wednesday at 17:00 at the latest.

Operational level
The daily OR planning on the operational level is determined at 10:30 one working day before the day
of surgery. To make this possible, the planning per specialism must be uploaded to EPIC Optime at
10:00 at the latest (Steen - de Vlaming & Hamersveld, 2021). Appendix F depicts an example of the
planning of a random day in a Gantt chart. The appendix shows the planning, as executed on 4-3-2019.

Evaluation of the Amsterdam UMC 32

4.2 HVAC system
The guidelines concerning HVAC systems are explained in previous sections. However, it is not clear
what HVAC systems are generally used for thus far. Additionally, there remains a knowledge problem
as it is unknown how HVAC systems are implemented in the Amsterdam UMC hospital. Different types
of HVAC systems are available, so the type(s) of HVAC systems used, and the operation of the systems
should be identified. This section elaborates on the importance of HVAC systems and on which type(s)
of HVAC systems are implemented in the Amsterdam UMC.

4.2.1 Importance of HVAC systems

The HVAC system in ORs ensures comfortable climate conditions within the OR by heating, cooling,
and ventilating. The HVAC system ensures the RH can be controlled and an environment free of micro-
organisms is created. Microorganisms such as bacteria are released into the air via, among others,
human skin, dust, and hair. Therefore, people present in the OR are the main source of the spread of
micro-organisms into the air. The air treatment in ORs via the use of HVAC systems aims to minimize
micro-organisms to prevent SSIs. This minimizes the chances of SSI or any other type of contamination
occurrences.

4.2.2 Operation of the HVAC systems

The HVAC system is set to daytime operation via a clock program that is adjustable at the reception
of the hospital, but it can also be set to daytime operation via a software switch. The clock program
automizes the operation of the HVAC systems by automatically switching to the daytime operation or
nighttime operation. Currently, the daytime operation is switched on between 6:30 a.m. and 9:00
p.m., from Monday to Friday. When switching an HVAC system from nighttime operation to daytime
operation, approximately 30 minutes are required before the system optimally operates. Out of the
20 class 1 performance level 1 ORs, 15 ORs are in nighttime operation between 9:00 p.m. and 6:30
a.m., from Monday to Friday. These ORs are continuously in nighttime operation on weekends and
during holidays. OR 4, 5, 8, 12, and 14 are exceptions to the clock program operation, as they are
operational 24 hours a day, seven days a week. This means that the HVAC systems in these ORs are
always in daytime operation. These ORs are preferably used for emergency procedures. The HVAC
systems are 100% operational during daytime operation, while the airflow rate of the HVAC systems
is a maximum of 50% compared to daytime operation during nighttime operation.
Each OR is equipped with the signaling and operating panel depicted in Figure 4-4. This is referred to
as a “Bender panel” in the hospital, as it is produced by the company Bender. This is a display
visualizing some KPIs. Color coding is used for visualizing whether the KPIs are within their limit values,
and both visual and audible notifications can be given in case of an alarm. Additionally, the panel can
be used for the operation of the HVAC system. All values of the KPIs on the dashboard can be changed
by using the interface of the panel. The panel overrides the central operating system. Limit values of
all KPIs related to the OR are described by Amsterdam UMC (2018) in their air management plan.
Check Appendix G for the requirements for critical parameters in class 1 performance level 1 ORs of
the Amsterdam UMC. The plenum is divided into two zones T1 and T2. The air in the downflow zone
T1 is approximately 1.5 °C lower than the downflow zone (T2). These temperature differences are
necessary to ensure downflow stabilization. As can be seen in this table, all clinical ORs currently
satisfy ISO 5 standards.

Evaluation of the Amsterdam UMC 33

Figure 4-4: Bender panel in OR 11

If surgery is still in progress outside daytime operation hours, the temperature and RH cannot be kept
within the set limits if no action is taken. This can be done by interacting with the panel in the OR. This
panel visualizes the daytime operation time window and some other relevant parameters for that
specific OR. By interacting with the “Bedrijfstijd” tab of the “Bender panel” of Figure 4-4, the switch
to nighttime operation can be postponed (Check Figure 4-3) (van der Steen - de Vlaming, 2022). The
screen will be green when the HVAC system of the corresponding OR is in daytime operation. If the
switch to nighttime operation is planned in an hour or less or five minutes or less, the screen will be
orange or red, respectively.

Figure 4-3: Bender panel displaying an OR in daytime operation

Evaluation of the Amsterdam UMC 34

4.2.3 HVAC systems at the Amsterdam UMC
As described in section 1.1: “Problem owner description”, the clinical OR department consists of 20
clinical ORs. In accordance with the guideline, the air treatment in the ORs was measured. It appears
that the clinical ORs meet the requirements for OR class 1 performance level 1 (van der Steen- de
Vlaming, 2022). The 20 ORs in the OR department of the Amsterdam UMC are equipped with UDF
HVAC systems. UDF HVAC systems can both have horizontal or vertical airflow, both of which can meet
ISO 5 standards. Figure 4-5 depicts the airflow patterns of horizontal and vertical UDF HVAC systems.
This airflow is often referred to as laminar. Laminar airflow is defined as airflow taking place along
constant streamlines, without turbulence (Oxford languages, 2022). However, in practice airflow
provided by UDF HVAC systems is never without turbulence as fixed equipment, medical equipment,
or even the surgeon redirect airflow. This causes some degree of turbulence. Horizontal UDF HVAC
systems provide an airflow of clean, filtered air from a plenum in one of the walls of the OR
horizontally. The wall in the opposite direction will remove the polluted air. In building construction,
a plenum is defined as a separate space provided for air circulation for HVAC (C2G, 2021). Horizontal
UDF in an OR as built causes less turbulence as opposed to vertical UDF, as turbulence is not the result
of airflow being redirected by a perpendicular surface. Because less turbulence is occurring with
horizontal UDF, better protection against contamination can be achieved. However, large materials
upstream can obstruct the operating area and contaminate the area.

Vertical UDF or downflow HVAC systems are implemented in every OR of the Amsterdam UMC
hospital. This type of HVAC system provides a protected area in an OR of filtered air originating from
a plenum in the ceiling. Polluted air is afterward removed by applying suction from every wall. A
vertical UDF system does not need as much floor space as a horizontal UDF. Additionally, the airflow
is not parallel to the operating area, which means that cross-contamination occurs less frequently
compared with a horizontal UDF HVAC system. The motivation to include only vertical UDF HVAC
systems stem from the fact that horizontal UDF HVAC systems are in practice found to be more
sensitive to airflow disruption caused by the positioning of personnel, material, and equipment. This
susceptibility to airflow disturbance was also observed under experimental conditions.

Figure 4-5: Airflow patterns of horizontal and vertical UDF HVAC systems

Evaluation of the Amsterdam UMC 35

4.2.4 HVAC system design
Different configurations of HVAC system elements are applied in hospitals. These configurations
consist of different applications of the clean air system, recirculation system, plenum, and swirl
diffusers. NVMM (2022) stated: “Of the two most commonly used HVAC systems, UDF and turbulent,
there seem to be advantages in terms of sustainability for using a turbulent HVAC system. Exact data
on this cannot be given because the environmental impact depends on more factors than just the type
of HVAC system. However, the use of a turbulent HVAC system generally leads to less energy
consumption because less air is transported”. We conclude that generally, vertical UDF HVAC systems
should be used for ORs with class 1 performance level 1 and turbulent HVAC systems for class 1
performance 2 ORs as turbulent HVAC systems consume less energy compared to UDF HVAC systems,
and class 1 performance 2 requirements can be met using turbulent HVAC systems. Figure 4-6 depicts
the configurations that are applied frequently by hospitals.

Figure 4-6: HVAC system configurations

UDF HVAC systems that are applied to class 1 performance level 1 generally have two potential
configurations, as the clean air system can be centralized or decentralized. Decentralized clean air
systems feature a clean air unit for every OR, while centralized clean air systems consist of central
clean air units providing clean air to the plenums of every OR. Recirculation systems are decentralized,
recirculating air for every corresponding OR individually. ORs classified as class 1 performance level 2
are often equipped with turbulent HVAC systems. These HVAC systems distribute air to the OR using
swirl diffusers instead of plenums. A decentralized recirculation system is optional in the turbulent
HVAC system. Clean air units are centralized.

Enlarged figures of the configurations are depicted in Appendix H. Figure H-2 visualizes the
configuration applied in the clinical ORs of the Amsterdam UMC hospital. The UDF HVAC systems
consist of a decentralized clean air system. This means that every OR is equipped with a clean air

Evaluation of the Amsterdam UMC 36

system that provides clean air, instead of a central clean air system that provides clean air to all ORs.
Decentralized recirculation systems are installed for every individual OR. These systems filter the
polluted air originating from the OR to provide clean air. The laminar airflow providing clean air is
distributed via a plenum installed in the structural ceiling of the OR. Polluted air is removed as depicted
in the vertical UDF HVAC system illustrated in Figure 4-5.

4.3 Energy consumption of the clinical OR department

The goal of this research is to decrease the CF in the clinical OR department of the Amsterdam UMC.
Insights into the current energy consumption are crucial to verify the effectiveness of the potential
recommendations.

4.3.1 Current energy consumption

As can be read in 4.2.3: “HVAC systems at the Amsterdam UMC”, the clinical OR department consists
of 20 class 1 performance level 1 ORs using UDF HVAC systems. Since 2018, 15 of these systems are in
nighttime operation between 21:00 and 6:30 on Monday to Friday and between 00:00 and 23:59
during weekends and holidays. This measure dramatically decreased the energy consumption of the
OR department. During the COVID-19 pandemic, the measure was temporarily not in effect. It was
introduced again in March 2021. Figure 4-7 depicts the energy consumption of the OR department
during a week in April 2019 (J. Koeman, personal communication, June 16, 2022). The blue graph
represents the energy consumption before the measure, while the red graph depicts the energy
consumption of the OR department after the implementation of the day- and nighttime operation
measure. When assessing the local extrema of the two graphs, we conclude that the local maxima of
the red graph are not notably decreased, while the local minima of the graph are significantly
decreased as opposed to the blue graph. The decreases in the local minima are caused by the ORs in
nighttime operation.

Figure 4-7: Energy consumption of the clinical OR department

The clinical OR department is in departments D and E of the Amsterdam UMC. These two departments
consumed a total of 9.2 GWh in 2018 (J. Koeman, personal communication, June 16, 2022). To
estimate the energy consumption of the ORs in the clinical OR department of the Amsterdam UMC, a
tool is used provided by W. Nieuwenhuyzen (personal communication, June 21, 2022). This tool
generates the annual energy consumption and energy costs as output using input parameters such as
the square meter of OR, operating hours, percentage of energy reduction due to airflow rate
reduction, and energy costs per kWh. All assumptions related to the input of this tool can be seen in
Appendix I. The tool is not depicted directly as it contains confidential information.

Evaluation of the Amsterdam UMC 37

Figure 4-8 depicts the output of the tool. It is an estimate of the total annual energy consumption
before and after implementation of the nighttime operation measure under the assumptions of
Appendix I. To acquire an estimate of the energy consumption before implementation, 20 ORs with a
total cleanroom surface of 18 ∗ 50 + 2 ∗ 70 = 1,040 𝑚 2 is assumed. Two separate calculations must
be made for an estimation of the current total energy consumption: The total annual energy
consumption of the 24/7 ORs and the total annual energy consumption of the elective ORs. We
assume a cleanroom surface of 4 ∗ 50 + 70 = 270 𝑚 2 for the calculation of the energy consumption
of the 24/7 ORs, as 24/7 OR 8 has a cleanroom surface of 70 𝑚 2, while the other four 24/7 ORs have
a cleanroom surface of 50 𝑚 2. For the calculation of the total annual energy consumption of all
elective ORs, 17 ∗ 50 + 70 = 770 𝑚 2 of cleanroom surface is assumed. The current total annual
energy consumption of the clinical OR department is then estimated to be approximately 1.65 GWh,
while the total annual energy consumption before implementation of the nighttime operation
measure is estimated to be approximately 2.46 GWh. The measure resulted in a reduction of the total
1,654,175
energy consumption of 33%. Currently, one OR is estimated to consume 20
≈ 82,709 kWh on
average annually. This total assumes 8,760 ∗ 5 + 3,654 ∗ 15 = 98,610 operating hours in total. This
98,610 82,709
implies an average of 20
= 4930.5 operating hours per OR, which results in 4930.5
≈ 16.77 kWh
per operating hour for one OR classified as class 1 performance level 1.

Figure 4-8: Energy consumption according to tool output

Evaluation of the Amsterdam UMC 38

5 Conceptual design
This chapter elaborates on the proposed conceptual design. The conceptual design describes the most
notable changes by elaborating on the proposed classifications, HVAC systems, and heuristics. As
opposed to Chapter 4: “Evaluation of the Amsterdam UMC”, we implement the classification of the
revised guideline (e.g., class 1+).

5.1 OR classification and classification of surgeries

The proposed classification of surgeries is described in this section. This classification is achieved by
classifying all surgeries concerning major joint replacements as class 1+. The remaining surgeries are
then classified as class 1 surgeries. This classification of surgeries adheres to the most recent guideline
concerning air treatment in ORs.

5.1.1 Proposed classifications of surgeries

The revised guideline proposed a new system to classify ORs as class 1+, class 1, or class 2. The current
classification of OR surgeries does not consider these new classifications. Additionally, only elective
major joint replacements are advised to be executed in a class 1+ OR (See Table D-1). Check Appendix
J for an elaborate list of all surgical procedures that are considered elective major joint replacements.
This comprehensive list is composed in cooperation with Niek Sperna Weiland, an anesthetist at the
Amsterdam UMC and organization supervisor of this project. A class 1 OR suffices for all other
surgeries. The current urgency classification does not require any changes. Assigning remaining
surgeries to ORs class 1, class 2, and individual treatment rooms can be executed in further research.

5.1.2 Proposed OR classification

According to historical data from 2019, 7.36% of all procedures in that year included a major joint
replacement. That is, 7.36% of all surgeries must be executed in an OR of class 1+. This results in
⌈20 ∗ 0.0736⌉ = 2 class 1+ ORs. 18 ORs will then be classified as class 1.

In 2019, there were some days (for example 3-2-2019) on which five ORs were occupied outside
business hours (6:30 – 21:00). However, none of the ORs were occupied simultaneously on this day.
From the data, we conclude that four 24/7 ORs instead of five 24/7 ORs would suffice. This saves
energy as a reduction of the operating hours will be realized. In 2019, only 19 surgeries classified as
class 1+ surgeries (Check Appendix J for these surgeries) have started outside business hours. None of
these surgeries occurred on the same day and not even in the same week. We conclude that it would
suffice if one class 1+ OR should be appointed as a 24/7 OR. The other class 1+ OR will be appointed
as an elective OR. Three class 1 ORs will then be appointed as 24/7 ORs and the remaining 15 class 1
ORs will be elective ORs. Check Figure 5-1 for a visual representation of the proposed OR classification.

Figure 5-1: Proposed OR classification

Conceptual design 39
5.2 HVAC system
This section elaborates on the proposed HVAC systems in the conceptual design. The expected energy
consumption is also determined in this section. Data that is collected for this section originates mainly
from the interview with A. Brouwer depicted in Appendix C.5.

5.2.1 Proposed HVAC systems

The proposed OR classes are class 1 and class 1+. This means that the class 1 ORs and class 1+ ORs in
the Amsterdam UMC will need to have a minimum air quality of ISO 7 and ISO 5 and a minimum
recovery time of 20 minutes and three minutes, respectively. The minimum number of air changes is
20 per hour for both classes (See Table 3-4). ISO 5 can be reached by turbulent HVAC systems, but it
takes a long time before this is reached. Therefore, turbulent HVAC systems cannot meet the
requirement for the recovery time of class 1+ ORs. UDF HVAC systems can meet this requirement and
should therefore be used in the class 1+ ORs, even though these systems consume more energy.
Turbulent HVAC systems can meet all requirements for class 1 ORs while consuming relatively small
amounts of energy compared to UDF HVAC systems.

The current ORs are equipped with UDF HVAC systems with decentralized clean air systems as
depicted in Figure H-2. The class 1+ ORs do not require changes as these ORs currently meet the
criteria for class 1+ ORs. However, energy might be saved by switching to a UDF HVAC system with a
centralized clean air system. Table 5-1 depicts the costs associated with each system according to
Braam (2022). The table uses general estimations of an average system. However, they do not
necessarily apply to the Amsterdam UMC.

Table 5-1: Costs of the HVAC system configurations

ORs “at-rest” with UDF HVAC systems with a laminar airflow will always maintain ISO 5 and a recovery
time of three minutes. However, this is not necessary for the class 1 ORs. The ORs classified as class 1
should be equipped with turbulent HVAC systems with decentralized recirculation systems, as can be
seen in Figure H-4. This configuration consumes the smallest amount of energy. The investment as
depicted in Table 5-1 will not be necessary, as the current UDF HVAC systems can be maintained.
Control technology can be built in with which the airflow rate can be reduced from high (laminar) to
low (turbulent) (A. Brouwer, personal communication, June 14, 2022) (Check Appendix C.5). If the
airflow rate is reduced, the plenum will function as a large diffuser of turbulent ventilation. It is
required that the fans in the HVAC system are adjustable and additional control technology must
therefore be installed. The “Bender panel” of the OR can then provide the interface for switching the
airflow rate. It needs to be validated that the OR meets the requirements of the intended OR class
after having switched from one class to another. Another aspect that needs to be validated is whether
the control system incorporates the switch to nighttime operation for the elective ORs.

Conceptual design 40
5.2.2 Energy consumption of proposed HVAC systems
As concluded in 4.3.1: “Current energy consumption”, ORs of class 1 performance level 1 consume
16,77 kWh per operating hour on average. We assume ORs of class 1+ and class 1 ORs to have similar
energy consumptions as ORs of class 1 performance level 1 and class 1 performance level 2,
respectively as the classes have similar characteristics. This means that we assume one class 1+ OR to
consume 16,77 kWh per operating hour.

Table 5-1 displays the costs and CO2 emissions of every HVAC configuration (Braam, 2022). Here, both
class 1 performance level 1 configurations use UDF HVAC systems, while both class 1 performance
level 2 ORs are equipped with turbulent HVAC systems. Check Figure 4-6 for a reminder of the
configurations. In Table 5-1, we see that the configuration that is currently implemented in the
Amsterdam UMC has the highest energy consumption. The color coding depicts this with the red color.
ORs with HVAC systems with a centralized clean air system as depicted in Figure H-1 can be used for
32,200−31,600
class 1+ ORs, and it could result in a reduction of 32,200
≈ 1.86% of energy costs. We assume
energy costs per kWh to be constant. Therefore, the energy consumption is reduced by approximately
1,86% as opposed to the current situation with decentralized clean air systems. Each class 1+ OR will
then consume 16.77 ∗ (1 − 0.0186) ≈ 16.46 kWh per operating hour. Class 1 ORs should be
equipped with turbulent HVAC systems with recirculation systems as visualized in Figure H-4, as this
HVAC system configuration has the smallest energy consumption while meeting class 1 standards
32,200−22,300
according to Braam (2022). The configuration for class 1 ORs consumes ≈ 30.75% less
32,300
energy as opposed to the current ORs. This corresponds to an energy consumption of 16.77 ∗
(1 − 0.3075) ≈ 11.61 kWh per operating hour of one class 1 OR.

The four 24/7 ORs will then consume a total of 8,760 ∗ (11.61 ∗ 3 + 16.46) = 449,300.40 kWh,
while the 16 elective ORs are expected to consume 3,654 ∗ (11.61 ∗ 15 + 16.46) = 696,488.94 kWh
annually if we assume the same number of operating hours as in 2019. This results in total energy
consumption of 449,300.40 + 696,488.94 = 1,145,789.34 kWh for the clinical OR department. In
total, 508,385.66 kWh is then saved annually. This corresponds to a reduction of the energy
consumption in the clinical OR department of approximately 31% as opposed to 2019.

5.3 Heuristics
This section describes the OR scheduling problem of the Amsterdam UMC, and it elaborates on
appropriate heuristics for assigning surgeries with the same urgency class to an OR. Data collection
for this section is executed using literature review and observations. The heuristics are designed for
management at a tactical level.

5.3.1 The OR scheduling problem

The Amsterdam UMC aims to reduce the maximum completion time of their surgeries to reduce the
energy consumption of the OR department. Minimizing the maximum completion time maximizes the
OR utilization and minimizes the duration that the machines, or ORs, must be operational. A reduction
of the maximum completion time enables the hospital to reduce the duration of the ORs in daytime
operation. Therefore, a reduction in the energy consumption of the HVAC systems can be realized.
Minimizing 𝐶𝑚𝑎𝑥 will be the objective of the heuristic. However, the throughput rate of the OR
department cannot suffer from the new planning heuristic. The maximum completion time 𝐶𝑚𝑎𝑥
defined as 𝑚𝑎𝑥{𝐶1 , 𝐶2 , … , 𝐶𝑛 }, is referred to as the makespan. This OR scheduling problem is a single-
stage schedule problem, as each job is finished after a singular execution phase, and each machine, or
OR, is identical in terms of processing speed as all the ORs have the same processing speed. However,
the scheduling heuristic will allocate surgeries to two types of machines which are class 1 and class 1+

Conceptual design 41
ORs. All surgeries can be performed in class 1+ ORs, while a subset of all surgeries can be performed
at the class 1 ORs. Therefore, this non-preemptive scheduling problem requires a variant of identical-
machine scheduling for the formulation of the scheduling problem. It is labeled non-preemptive as we
assume that surgeries cannot be interrupted and finished in a different OR. Additionally, sequence-
dependent setup times are required for each surgery. These setup times are, among other things,
caused by cleaning the OR, personnel preparation, and equipment preparation. We assume that the
setup time before a job depends on the previous job executed at a machine, regardless of the
machine. The problem is NP-hard. Here NP means “non-deterministic polynomial time”. An NP-hard
problem is considered the most difficult problem to find an optimal solution. We, therefore, use a
heuristic to find a near-optimal solution.

The clinical OR department consists of 20 ORs, which can be classified as either class 1 or class 1+ ORs.
We define 𝑉𝑝,𝑞 to be the scheduling variant applicable to the OR scheduling problem, where 𝑝 denotes
the number of class 1 ORs and 𝑞 denotes the number of class 1+ ORs at the clinical OR department.
We assume 𝑀 ∈ {𝑀1 , 𝑀2 , … , 𝑀𝑚 } to be the set of machines, while 𝐽 ∈ {𝐽1 , 𝐽2 , … , 𝐽𝑛 } denotes the set
of independent jobs. Using the three-field notation, this scheduling problem is then described as
𝑉18,2 |𝑠𝑘,𝑗 |𝐶𝑚𝑎𝑥 if the clinical OR department consists of 18 class 1 ORs and two class 1+ ORs as
explained in 5.1.2: “Proposed OR classification”.

5.3.2 Proposed heuristics

“Loosely speaking, heuristic means to find or to discover by trial and error (Yang, 2011). It consists of
various consequent steps to perform to approach an optimal solution. Well-known heuristics that are
implemented in many scheduling problems are First Come, First Served (FCFS), Earliest Due Date (EDD)
first, Shortest Processing Time (SPT) first, and Longest Processing Time (LPT) first. The EDD rule
generally minimizes the maximum job lateness for scheduling problems with a single machine
including job due dates, while the SPT rule generally minimizes the flow time of a job. These heuristics
solve relatively easy problems that are solvable in polynomial time. It can be shown that the OR
scheduling problem of the Amsterdam UMC is a complex problem that cannot be solved in polynomial
time. It is NP-hard (Pinedo & Hadavi, 1995). We are interested in minimizing the makespan.

Arnaout & Kulbashian (2008) evaluated five heuristics intending to minimize the makespan of an OR
scheduling problem in a Lebanese hospital. The following heuristics were considered:
- Longest Processing Time (LPT) and Shortest Processing Time (SPT)
- Longest Expected Processing Time (LEPT) and Shortest Expected Processing Time (SEPT)
- Longest Expected Processing Time with Setup Time (LEPST)

LEPT and SEPT are extensions of LPT and SPT, respectively. In contradiction to LPT and SPT, LEPT and
SEPT consider the average processing times of jobs, which is the only aspect in which the heuristics
differ. This enables the heuristic to approach an optimal solution for stochastic problems. LEPT and
SEPT selects the job with the maximum or minimum expected processing time 𝐸(𝑝𝑗 ), respectively.
The selected job is then assigned to the machine with the minimum completion time.

LPT generally minimizes the makespan relatively effectively. However, several heuristics outperform
the LPT heuristic. Arnaout & Kulbashian (2008) showed that Longest Expected Processing Time (LEPT)
is a heuristic that outperforms LPT. Additionally, the authors claimed Longest Expected Processing
Time with Setup Time (LEPST) to be the best heuristic to maximize utilization and minimize the
makespan. The study by Arnaout & Kulbashian (2008) showed that LEPST outperforms LPT, SPT, LEPT,
and SEPT. LEPT performed the second best out of all the heuristics. Therefore, we will elaborate on
the proposed heuristics LEPST and LEPT in the remainder of this research.

Conceptual design 42
LEPT
LPT selects the job 𝑗 with the longest processing time 𝑝𝑗 first and allocates this job to machine 𝑖. LEPT
is an extension of LPT, which select the job 𝑗 with the longest expected processing time 𝐸(𝑝𝑗 ), and
allocates this job to machine 𝑖. Priority is therefore given to jobs with the maximum value of expected
processing time.

Let 𝑆 be the set consisting of all unscheduled jobs. We assume 𝑗 to be an element of the set of
unscheduled jobs 𝑆: 𝑗 ∈ 𝑆. Additionally, we define 𝐶𝑘,𝑖 as the completion time of the previous job 𝑘
on machine 𝑖. A sequence of three steps can be formulated to define the LEPT heuristic (Arnaout &
Kulbashian, 2008):

1. Find and select the job 𝑗 and machine 𝑖 in the set 𝑆 with the maximum value of 2:

[𝑚𝑖𝑛{𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 )] (2)

2. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆.
3. If 𝑆 = ∅, then STOP. Else, return to step 1.

To implement the LEPT heuristic to the OR scheduling problem of the Amsterdam UMC, some
adjustments have to be made to the heuristic because we want to distinguish between class 1 and
class 1+ ORs. For this purpose, let 𝑆 remain the set of unscheduled jobs. A subset 𝑆𝐵 of 𝑆 contains all
unscheduled jobs of type B: 𝑆𝐵 ⊆ 𝑆. To specify the type of machine, we define 𝑀𝐵 to be a subset of
𝑀, which is the set containing all machines: 𝑀𝐵 ⊆ 𝑀. Let 𝑀𝐵 denote the set of machines of type B.
For the remainder of this chapter, we will define type A jobs as surgeries that can be executed in ORs
of class 1 and class 1+, excluding jobs that can only be executed in class 1+ ORs. A job of type B is
defined as a job that must be executed in ORs class 1+. This includes surgeries concerning elective
major joint replacements. ORs class 1 are machines of type A, while ORs class 1+ are machines of type
B. Jobs of type A can be executed using any machine, while jobs of type B can only be executed at
dedicated machines. These dedicated machines are ORs of class 1+. The following sequence of steps
can then be implemented by the Amsterdam UMC denoted by LEPT*:

1. Find and select the job 𝑗 in the subset 𝑆𝐵 and machine 𝑖 in the subset 𝑀𝐵 with the maximum
value of 2:
[𝑚𝑖𝑛{𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 )] (2)

2. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆 and subset
𝑆𝐵 .
3. If 𝑆𝐵 = ∅, then proceed to step 4. Else, return to step 1.
4. Find and select the job 𝑗 in the set 𝑆 and machine 𝑖 in the set 𝑀 with the maximum value of
2:
[𝑚𝑖𝑛{𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 )] (2)

5. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆.
6. If 𝑆 = ∅, then STOP. Else, return to step 4.

Conceptual design 43
LEPST
LEPST is an extension of both LPT and LEPT, considering setup times in contradiction to LPT and LEPT.
It is introduced by Arnaout & Kulbashian (2008), and it is used for strongly NP-hard problems with
sequence-dependent setup times (Pinedo & Hadavi, 1995). The idea of LEPST is to give priority to jobs
with relatively long setup times, minimizing the makespan.

Let 𝑆 be the set consisting of all unscheduled jobs and let 𝑠𝑖,𝑘,𝑗 denote the sequence-dependent setup
time between the previous job 𝑘 and the next job 𝑗 on machine 𝑖. We assume 𝑗 to be an element of
the set of unscheduled jobs 𝑆: 𝑗 ∈ 𝑆. Additionally, we define 𝐶𝑘,𝑖 as the completion time of the
previous job 𝑘 on machine 𝑖. The control parameter 𝛼 is a variable with a value between 0 and 1,
which indicates the importance of processing times when selecting a job. A value of 0 means that only
the setup times are considered while ignoring the processing times. A control parameter 𝛼 with a
value of one is used to assign equal weights to setup times and processing times when selecting a job.
The LEPST heuristic can then be described as a sequence of three steps (Arnaout & Kulbashian, 2008):

1. Find and select the job 𝑗 and machine 𝑖 in the set 𝑆 with the maximum value of 3:

[𝑚𝑖𝑛 {𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 ) ∗ 𝛼 + 𝑠𝑖,𝑘,𝑗 ] (3)

1. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆.
2. If 𝑆 = ∅, then STOP. Else, return to step 1.

To implement the LEPST heuristic to the OR scheduling problem of the Amsterdam UMC, some
adjustments must be made to the heuristic because we want to distinguish between class 1 and class
1+ ORs again. This is done in the same way as for the LEPT* heuristic. For this purpose, let 𝑆 remain
the set of unscheduled jobs. A subset 𝑆𝐵 of 𝑆 contains all unscheduled jobs of type B: 𝑆𝐵 ⊆ 𝑆. To
specify the type of machine, we define 𝑀𝐵 to be a subset of 𝑀, which is the set containing all
machines: 𝑀𝐵 ⊆ 𝑀. Let 𝑀𝐵 denote the set of machines of type B. The following sequence of steps can
then be implemented by the Amsterdam UMC denoted by LEPST*:

1. Find and select the job 𝑗 in the subset 𝑆𝐵 and machine 𝑖 in the subset 𝑀𝐵 with the maximum
value of 3:

[𝑚𝑖𝑛 {𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 ) ∗ 𝛼 + 𝑠𝑖,𝑘,𝑗 ] (3)

2. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆 and subset
𝑆𝐵 .
3. If 𝑆𝐵 = ∅, then proceed to step 4. Else, return to step 1.
4. Find and select the job 𝑗 in the set 𝑆 and machine 𝑖 in the set 𝑀 with the maximum value of
3:
[𝑚𝑖𝑛 {𝐶𝑘,𝑖 } + 𝐸(𝑝𝑗 ) ∗ 𝛼 + 𝑠𝑖,𝑘,𝑗 ] (3)

5. Assign the selected job 𝑗 to the selected machine 𝑖. Remove job 𝑗 from the set 𝑆.
6. If 𝑆 = ∅, then STOP. Else, return to step 4.

Conceptual design 44
6 Monte Carlo simulation
An MC simulation is made to verify the effectiveness of the LEPT* heuristic and the LEPST* heuristic.
This chapter describes the MC simulation by first explaining how the data of all surgeries from 2019 is
prepared for the MC simulation. After this, this chapter describes how the MC simulation is made and
how it should be used.

6.1 Data preparation

Historical data of 2019 is used in this research, as this is the most recent year that was not affected by
COVID-19. The data depicted all surgeries as executed in that year in an Excel file. Appendix L depicts
a fraction of the original Excel file as delivered by the Amsterdam UMC. The OR, date, start time and
end time, duration, procedure, procedure number, and log number of every surgery was included. No
indication of surgery urgency was provided. Therefore, the assumption was made that all surgeries
that started between 7:30 and 17:00 on Monday to Friday were elective, while all surgeries executed
at other times were emergency surgeries. All elective surgeries were extracted from the data in one
new table using a filter in the IN_OR and day column. The column with the day of the week was added
by using the Excel function depicted in Appendix K.1. Then, the procedure classification as depicted in
Appendix J was compiled. Every distinct elective surgery in the table was classified as either “class 1”
or “class 1+” by adding a column including the Excel function of Appendix K.2 in every row of this
column. Appendix M depicts a fraction of the resulting table of the data of elective procedures.

6.1.1 Setup times

The processing times were already included in the data. However, setup times were not. The data of
the elective procedures were used to determine the setup times of surgeries. Appendix K.3 depicts
the Excel function that was used to determine the setup times. The function checks if the day and the
OR are the same and subtracts the end time of the previous surgery from the start time of a new
surgery if this is true. If an error occurs, the function returns a blank cell. The column with setup times
is then filtered, as we assume the setup time to be greater than 0 minutes and less than or equal to
90 minutes. Check Figure 4-2 for a schematic overview of setup times.

6.1.2 Data analysis

The distribution of both the setup times and the processing times was analyzed. This was executed
separately for class 1 and class 1+ procedures. Descriptive statistics were made for the class 1
processing times, class 1 setup times, class 1+ processing times, and class 1+ processing times using
the Excel Data Analysis Tool. Four histograms were then made by implementing the rule of thumb of
assigning √𝑛 bins, where 𝑛 represents the number of observations. Bins with less than five
observations were merged with other bins. Figure 6-1 depicts the resulting histograms.

Figure 6-1: Histograms of 2019 data

Monte Carlo simulation 45

The Chi-square goodness of fit test was applied to each dataset to test whether the datasets fit a
particular theoretical distribution. This was done by computing the Cumulative Distribution Function
(CDF) for every bin using the Excel functions of the corresponding theoretical distributions that were
tested. The following theoretical distributions were considered: Gamma, Normal, Lognormal,
Exponential, and Weibull. Check Appendix K.4: CDF for the Excel functions that were implemented to
calculate the values of the CDF. The CDF values were then multiplied by the total number of
observations. Now, we can compute the expected number of observations per bin by subtracting the
individual values. Then, the test statistic was determined by adding the errors for each bin using the
following formula:
(𝑶𝒊 −𝑬𝒊)𝟐
𝝌𝟐 = ∑𝒌𝒊=𝟏 𝑬𝒊
(4)

, where:

- 𝑂𝑖 = The number of observed observations (𝑖 = 1, 2, … , 𝑘)

- 𝐸𝑖 = The number of expected observations (𝑖 = 1, 2, … , 𝑘)

The test statistic was computed by adding the errors, and the critical value 𝑐 was determined by the
right-tailed probability of the Chi-squared distribution using the Excel function in Appendix K.5. This
function requires the significance level and the degrees of freedom as input parameters. The degrees
of freedom were equal to the number of bins subtracted by 1, with a significance level 𝛼 = 0.05. We
can conclude whether a theoretical distribution fits the dataset by comparing the values of 𝜒2 and 𝑐.
We rejected the two null hypotheses that any of the evaluated theoretical distributions fit one of the
datasets, as 𝜒2 > 𝑐 for every Chi-square goodness of fit test.
As we could not show convincingly that any of the evaluated theoretical distributions fit the datasets,
we conclude that all four datasets must be evaluated using empirical distributions. Empirical
distributions are unlike the Gamma distribution, Weibull distribution, Normal distribution, Lognormal
distribution, and Exponential distribution, not theoretical distributions. However, the distribution is
based on observation as opposed to logic or mathematical functions.

6.2 MC simulation
The processing times and the setup times are input parameters for the MC simulation. Therefore, data
must be generated according to their empirical distributions. For every input parameter (Class 1
processing times, class 1 setup times, class 1+ processing times, and class 1+ setup times) the
probability that the input parameter is in a bin was computed for every bin by dividing the observed
number of observations by the total number of observations. Then, the cumulative probabilities were
calculated. This resulted in cumulative probabilities between 0 and 1. By generating a random variable
𝑋 ~ 𝑈[0, 1], the bin for which holds that the random variable 𝑋 is greater than the lower bound of the
cumulative probability of the bin but less than the upper bound of the cumulative probability of the
bin is selected according to the empirical distribution. Then, a random number within the selected bin
is selected. This is done repeatedly to generate data according to the corresponding empirical
distributions of every input parameter. Appendix N.1 depicts the code implemented in Visual Basic for
the generation of processing times of class 1+. Similar codes have been used for the generation of the
other input parameters. The green letters represent pseudo codes: These codes are comments to
clarify the written codes. The number of observations that are generated is equal to the number of
observed observations in the data concerning elective procedures of 2019. The generated data is
printed on a separate sheet. This means that this data will solely be used as input for the MC
simulation, but it is not directly visible in the MC simulation.

Monte Carlo simulation 46

6.2.1 Generating jobs
The MC simulation will simulate one working week. Therefore, 171 jobs must be generated as this was
the average number of elective surgeries in one working week in 2019 if we assume 261 working days.
92.32% of these surgeries were class 1 surgeries, while the rest of the surgeries are classified as class
1+ surgeries. This was determined by implementing the Excel functions depicted in Appendix K.6. The
MC simulation will randomly select a processing time and setup time from the input parameters which
have empirical distributions. The processing time and setup time will be selected from either the class
1 or the class 1+ input parameters with the corresponding probability. Appendix N.2 depicts the code
implemented in Visual Basic for the generation of jobs. Emergency surgeries are not generated as
these surgeries are not simulated. However, at least two ORs must always remain available for
emergency surgeries. Therefore, 18 of the 20 ORs can be used for elective surgeries in the simulation.

6.2.2 Scheduling jobs

The proposed heuristics must be applied to the generated jobs to verify the effectiveness of the
individual heuristics. Therefore, the two heuristics must be programmed in Visual Basic. Check section
5.3: “Heuristics” for a reminder of the proposed heuristics. The LEPT* heuristic first schedules jobs
with the longest expected processing time. Therefore, the jobs are sorted automatically decreasingly
by processing time after generating jobs. The LEPST* first schedules jobs with the highest LEPST value.
The column with these values is therefore sorted decreasingly in the LEPST* code. The Visual Basic
code then schedules the class 1+ surgeries first, as these surgeries can only be assigned to dedicated
class 1+ ORs. Class 1 surgeries are then scheduled for all the ORs. A surgery is assigned to the OR with
the minimum completion time. Surgeries are only assigned to an OR if the surgery can start after 7:30
and before 17:00. As a working day generally starts at 7:30, surgeries can be started during the first
570 minutes. This means that the previous completion time of an OR plus the setup time of the
scheduled surgery must be less than 570 minutes. Surgeries that cannot be scheduled within these
570 minutes must be planned for the next day. Appendix N.3 depicts the code implemented in Visual
Basic for the application of the proposed scheduling heuristics. Pseudo-codes are provided for more
elaboration on the written code.

6.2.3 Iteration
MC simulations estimate the result of stochastic events by iteratively predicting outcomes. The
generation and scheduling of jobs must be performed iteratively to acquire more accurate results.
This is done by running the codes repeatedly using a for loop in Visual Basic. A large number of
iterations should be applied for accurate results. The MC simulation assesses the effectiveness of the
heuristics using the KPI “Average maximum completion time (𝐶𝑚𝑎𝑥 ̅ )”. This is the average of each
working day’s maximum completion time over one working week. This average is calculated
iteratively. Check Appendix N.4 for the implemented Visual Basic code for the iteration and calculation
of the output of the MC simulation.

6.2.4 Interface
Figure 6-2 depicts the interface of the MC simulation. The MC simulation consists of three tables. Note
that these tables are not depicted entirely in Figure 6-2, as the tables consist of many rows. The table
at the left depicts the generated jobs that must be planned. These jobs consist of a corresponding
class, expected processing time, and expected setup time. The LEPST value resulting from equation 3
is also included in this table. The two tables at the right visualize the schedule resulting from the LEPT*
heuristic and the LEPST* heuristic by showing the completion time in minutes for every OR during
each day. The MC simulation will run after clicking on the “Iterate button”.

Monte Carlo simulation 47

The input table in the middle of the interface allows the user to interact with the simulation by
changing the input variables. α is currently set to 0.65, but this value can be changed if needed. The
optimal value for α is researched in chapter 7: “Results”. 1000 iterations are used for the simulation
to ensure accurate output. More iterations will not result in a significant increase in stability, as we
already obtain stable results. The number of iterations can be changed to reduce the duration of the
MC simulation or to increase the accuracy of the MC simulation. Finally, the number of both class 1
and class 1+ elective ORs can be altered to research the effect on the completion times. 18 of the 20
ORs are to be used for elective surgeries. 5.1.2: “Proposed OR classification” explains that two class
1+ ORs suffice. Therefore, the MC simulation uses two class 1+ ORs and 16 class 1 ORs as input values.
The two remaining ORs are used for emergency procedures. Both ORs are class 1 ORs as orthopedic
procedures are generally classified as urgency class S3. This means that surgery is semi-urgent. The
patient then needs to be on the operating table within 24 hours.

The MC simulation also consists of an output table displaying the results of the previous simulation.
The output depicts the KPI “Average maximum completion time” as explained in 6.2.3: Iteration for
both the LEPT* heuristic and the LEPST* heuristic.

Figure 6-2: MC simulation interface

Monte Carlo simulation 48

7 Results
This chapter elaborates on the results of this research. The optimal value of control parameter 𝛼 is
researched. Additionally, the impact of the proposed OR classification and the proposed heuristic on
energy consumption is assessed. The output of the MC simulation is used for this chapter. A selection
is also made concerning the most effective heuristic.

7.1 Control parameter α

The effect of the value of α on 𝐶𝑚𝑎𝑥 in the scheduling problem 𝑉𝑖 |𝑠𝑝,𝑞 |𝐶𝑚𝑎𝑥 is researched by Arnaout
& Kulbashian (2008). The minimum makespan was achieved by implementing a value of 𝛼 equal to
0.63 (Arnaout & Kulbashian, 2008). The scheduling problem in that study is different because a single
type of machine is involved. Therefore, we do not assume the optimum value of α to be the same in
the OR scheduling problem of the Amsterdam UMC. The MC simulation is used to assess the impact
of the value of α, and to determine the optimum value of α. This is done by running the MC simulation
for different values of α using 100 iterations. The average maximum completion time 𝐶𝑚𝑎𝑥 ̅ using the
LEPST* heuristic is noted for every value of α. Figure 7-1 depicts the graph of the experiment. We
conclude that 0.65 is the optimum value for α with 100 iterations, as this value resulted in the lowest
̅
value of 𝐶𝑚𝑎𝑥 . The corresponding value of 𝐶𝑚𝑎𝑥 ̅ was 455 minutes in this experiment. For the
remainder of this research, the value of α is equal to 0.65 as this is optimal.

Figure 7-1: Control parameter α

7.2 Heuristic selection

̅
In section 7.1: “Control parameter α”, we observed a 𝐶𝑚𝑎𝑥 of 455 minutes. To research the stability
of the MC simulation, we will run the simulation several times with the same values for the input
parameters. We can then gain insights into the stability of the MC simulation, and we can provide a
̅
more accurate estimate of 𝐶𝑚𝑎𝑥 .

Figure 7-2 denotes the results of 6 runs. All of these runs have the same values for the input
parameters: 𝛼 = 0.65, 1000 iterations, 2 elective class 1+ ORs, and 16 elective class 1 ORs. The graph
verifies that the LEPST* heuristic consistently outperforms the LEPT* heuristic. The lowest value of
̅
𝐶𝑚𝑎𝑥 ̅
for the LEPST* heuristic is 455 minutes, while the highest value of 𝐶𝑚𝑎𝑥 attains a value of 458
minutes. This interval of 3 minutes is relatively small. Hence, we conclude that the output of the MC
̅
simulation is relatively stable. Over these experiments, the LEPST* heuristic attains an average 𝐶𝑚𝑎𝑥
̅
of 456 minutes and the LEPT* heuristic an average 𝐶𝑚𝑎𝑥 of 462 minutes.

Results 49
Figure 7-2: MC simulation output stability

The LEPST* heuristic consistently outperforms the LEPT* heuristic with 𝛼 = 0.65. The heuristic is
approximately 1.3% more effective compared to the LEPT* heuristic when using 1000 iterations and
̅
𝛼 = 0.65, as the LEPT* heuristic attains a 𝐶𝑚𝑎𝑥 ̅
of 462 minutes as opposed to a 𝐶𝑚𝑎𝑥 of 456 minutes
for the LEPST heuristic. This results from the MC simulation, check Figure 7-3 for the output of one
experiment in the simulation.

Figure 7-3: Output of the Monte Carlo simulation

7.3 Expected impact on sustainability

Currently, the elective ORs are operational between 6:30 and 21:00, which results in 14.5 operational
hours daily. However, the operational hours can be increased if surgeries exceed the end of the
scheduled operational hours in practice. The data showed that on average 14.53 operational hours
were required to complete the elective surgeries in practice in 2019. If we assume that the surgeries
̅
generally start at 7:30 like in the MC simulation, the 𝐶𝑚𝑎𝑥 of the OR department was 13.53 hours. This
is equal to 811,8 ≈ 812 minutes. The LEPST* heuristic attains a 𝐶𝑚𝑎𝑥 ̅ of 456 minutes. The heuristic
saves 355.8 minutes or approximately 5.93 hours on average daily. As the ORs are operational from
6:30 and the surgeries start at 7:30, the ORs will be operational for 456 + 60 = 516 minutes ≈ 8.6
hours per day. This results in approximately 2167 operational hours annually.

7.3.1 Validation of number of class 1+ ORs

Section 5.1.2: “Proposed OR classification” states that two elective class 1+ ORs suffice for the clinical
OR department of the Amsterdam UMC. This section validates this statement by researching the effect
of increasing and decreasing the number of elective class 1+ ORs. This is done by changing the input
parameters of the MC simulation and observing the output repeatedly. Figure 7-4 depicts a graph of
the results of this experiment. Decreasing the number of class 1+ ORs to one returns an error as the

Results 50
MC simulation simulates only one week. With one class 1+ OR, the surgeries for class 1+ ORs cannot
be finished within that week. It is, therefore, not viable to assign one elective class 1+ OR. Figure 7-4
depicts the graph resulting from this experiment. We conclude that 𝐶𝑚𝑎𝑥 ̅ decreases if the number of
elective class 1+ ORs increases from two to ten ORs. With ten elective class 1+ ORs, a 𝐶𝑚𝑎𝑥 ̅ of 444
minutes is achieved by the LEPST* heuristic. We conclude that with every class 1+ OR that is added,
approximately 1.5 minutes are saved. We assume that every class 1+ OR that is added switches to
1.5
nighttime operation at night and on weekends. Then, ∗ 5 ∗ 52 = 6.5 hours are saved annually with
60
every additional elective class 1+ OR. This results in energy consumption of (2167 − 6.5) ∗ 16.46 =
35,561.83 kWh per class 1+ OR annually, while every class 1 OR consumes 2167 ∗ 11.61 = 25,158.87
kWh annually. Adding more elective class 1+ ORs is not justified as this consumes more energy despite
̅
the smaller value of 𝐶𝑚𝑎𝑥 . Hence, sufficient ORs should be implemented to ensure enough capacity.
However, not more than necessary.

Figure 7-4: Validation of number of elective class 1+ operating rooms

7.3.2 Energy consumption

Section 5.2.2.: “Energy consumption of proposed HVAC systems” describes the energy consumption
of the proposed HVAC systems by calculating the energy consumption of the 24/7 ORs and the energy
consumption of the elective ORs separately. As a result of the proposed heuristic, the operational
hours of the elective ORs will be decreased. This realizes a decrease in energy consumption. The
energy consumption of the 24/7 ORs will not be affected by the proposed heuristic. Therefore, the
four 24/7 ORs are expected to consume 449,300.40 kWh after the implementation of the proposed
OR classification and heuristic. The elective ORs are affected by the proposed heuristic. The expected
energy consumption after implementation of the proposed OR classification and heuristic of the
elective ORs equals 2167 ∗ (11.61 ∗ 15 + 16.46) = 413,051.87 kWh. After implementing the
proposed changes, the clinical OR department is expected to consume 449,300.40 + 413,051.87 =
862,352.27 kWh. This results in a total expected reduction in energy consumption of 791,820.73
kWh compared to the energy consumption in 2019, corresponding to a reduction of approximately
48% relative to 2019.

7.3.3 Sustainability
The conversion factor in the Netherlands was 0.390 kg CO2e per kWh in 2019. We define CO2e as
carbon dioxide equivalents, which also takes other polluting gases besides CO2 into account. The
European Environment Agency (2019) based this conversion factor on the ratio of CO2 emissions from
public electricity production. We conclude using the CO2e emission intensity that 791,820.73 ∗
0.390 = 308,810.085 kg CO2e can be saved compared to 2019.

Results 51
Conclusions and recommendations
In this research, the objective is to decrease the energy consumption of the Amsterdam UMC hospital
by optimally changing the HVAC planning and surgery scheduling without compromising the efficiency
of the OR department while obeying the guideline concerning air quality in the OR. This chapter
concludes the research by summarizing the answers to the research questions and by answering the
research question of this research. Recommendations to the hospital are also given in this chapter.
The research question is formulated as:

“How can the planning of the HVAC system usage and the scheduling of the surgeries in the ORs be
optimized so that a reduction in energy consumption is realized without compromising the efficiency
of the OR while obeying the guideline concerning air quality in the OR? “

Conclusions
This section concludes the research by summarizing the answers to each research question. The main
research question is answered at the end of this section.

Guidelines
The previous guideline recommended that hospitals should be equipped with or have advanced plans
to implement UDF HVAC systems in their OR department. UDF HVAC systems outperform turbulent
HVAC systems when minimizing the number of micro-organisms in the OR. However, Nederlandse
Vereniging voor Medische Microbiologie (NVMM) has not found any evidence that UDF HVAC systems
also prevent SSIs more effectively relative to turbulent HVAC systems. Therefore, the revised guideline
abandons the recommendation of implementing UDF HVAC systems in ORs. ORs should now be
equipped with an HVAC system that (regardless of the type of HVAC system) at least meet the
minimum criteria stated in Table 3-4. ORs of class 1+ use UDF HVAC systems and they are solely
recommended for major joint replacements. Class 1 ORs with turbulent HVAC systems consume less
energy and these ORs suffice for all procedures except major joint replacements according to the
revised guideline.

HVAC systems
Besides ensuring comfortable climate conditions in the OR, the air treatment via the use of HVAC
systems minimizes micro-organisms to prevent SSIs. This minimizes the chances of SSI or any other
type of contamination occurrences. The ORs in the clinical OR department of the Amsterdam UMC
currently consists of vertical UDF HVAC systems with decentralized clean air systems and recirculation
systems. These HVAC systems use a clean laminar airflow to minimize micro-organisms in the
protected area of the OR. The HVAC systems use a clock program that switches the systems on at 6:30
and switches the systems into nighttime operation at 21:00. A maximum of 50% of the airflow rate is
applied by the HVAC systems when operating in nighttime operation. Exceptions to the clock program
are the systems in OR 4, 5, 8, 12, and 14, as these ORs operate 24/7 for potential emergencies. Each
OR is equipped with the “Bender” panel depicted in figure 4-4, displaying the operational hours of the
OR, the temperature, RH, and other KPIs. The operational hours can be extended here, procrastinating
the switch to nighttime operation. The current energy consumption of the clinical OR department was
estimated using a tool for calculating the energy consumption of HVAC systems. The output of this
tool can be found in Figure 4-8. 1,654,175 kWh are currently estimated to be consumed annually.

Conclusions and recommendations 52

Scheduling
All ORs have the same HVAC systems, which means that all surgeries can be executed at each OR.
However, ORs are equipped with different equipment, which complicates the scheduling of surgeries.
The scheduling of the surgeries occurs at a strategic level, tactical level, and operational level. On a
strategic level, the annual planning is determined, while the monthly and weekly planning are
determined on a tactical level. The daily planning is determined on the operational level by every
specialism. Urgency classifications are assigned to surgeries to determine the maximum duration
before a patient must be on the operating table.

2 class 1+ suffice when assessing the data from 2019, as 7.37% of all surgeries are classified as major
joint replacements. These surgeries are generally not classified as urgent procedures, so two ORs
suffice. 18 ORs will then be appointed as class 1 ORs. The data from 2019 suggested that five 24/7 ORs
were rarely occupied outside operating hours and if this was the case, the ORs were not used
simultaneously. Four 24/7 ORs instead of five 24/7 ORs also suffice, decreasing energy consumption.
In total, 19 class 1+ surgeries have been executed in 2019 outside operating hours. None of these
occurred simultaneously, not even on the same day. Appointing one class 1+ OR as 24/7 OR is viable.
The three other 24/7 ORs are then classified as class 1 ORs. Additionally, two ORs must be appointed
as emergency ORs for emergency procedures. These emergency ORs should be class 1 ORs, as class
1+ surgeries are generally not urgent.

When implementing two different types of machines (class 1 and class 1+), the single-stage job
scheduling problem of the Amsterdam UMC can be described as 𝑉18,2 |𝑠𝑘,𝑗|𝐶𝑚𝑎𝑥 if we define 𝑉𝑝,𝑞 to
be the scheduling variant applicable to the OR scheduling problem. 𝑝 denotes the number of class 1
ORs and 𝑞 denotes the number of class 1+ ORs at the clinical OR department. Two heuristics applicable
to the 𝑉18,2 |𝑠𝑘,𝑗 |𝐶𝑚𝑎𝑥 job scheduling problem are proposed by Arnaout & Kulbashian (2008): LEPT
and LEPST. The LEPT and LEPST heuristics cannot be applied directly to the 𝑉18,2 |𝑠𝑘,𝑗 |𝐶𝑚𝑎𝑥 job
scheduling problem as the heuristics are designed for identical-machines scheduling. The 𝑉18,2
|𝑠𝑘,𝑗|𝐶𝑚𝑎𝑥 job scheduling problem is similar, but it assigns jobs to two different types of machines
(class 1 and class 1+ ORs). Therefore, the heuristics are changed so that they can be applied to the
scheduling problem. This results in the LEPT* and LEPST* heuristics. The MC simulation is used to
compare the effectiveness of the two heuristics, and the LEPST* heuristic consistently outperforms
the LEPT* heuristic when assessing the average maximum completion time. The LEPST* heuristic has
an average maximum completion time of 455 minutes using 1000 iterations and 𝛼 = 0.65. This
optimal value of control parameter α was determined by running the MC simulation for different
values of α using 100 iterations.

Sustainability
Using the LEPST* heuristic, a reduction from the average maximum completion time 𝐶𝑚𝑎𝑥 ̅ of 812
minutes to 455 minutes can be realized by implementing the proposed solutions. This significant
reduction of 44% of the makespan is partially caused by the fact that all ORs are considered to be
equipped with identical equipment and resources. In 2019, 1,654,175 kWh are consumed by the
clinical OR department annually according to the implemented tool. After the implementation of the
proposed changes, an estimated 862,352.27 kWh is consumed annually. This corresponds to a
reduction of 48% as opposed to 2019. 791,820.73 kWh is estimated to be saved per year.

Using a CO2e emission intensity of 0.390 kg CO2e per kWh, we conclude that 791,820.73 ∗ 0.390 =
308,810.085 kg CO2e can be saved compared to 2019 (European Environment Agency, 2019).

Conclusions and recommendations 53

Conclusion
The LEPST* heuristic effectively schedules the surgeries, and it outperforms the current scheduling
and the LEPT* heuristic. This implies that the switch to nighttime operation can occur earlier due to
the decrease in the average maximum completion time. Less operational hours are therefore
required. 18 ORs can be changed to class 1 ORs while not hurting the efficiency of the clinical OR
department, and four instead of five 24/7 ORs are required. The classification of the ORs is depicted
in Figure 5-1. Two ORs should be appointed as emergency ORs. This research answers the research
question, as proposed changes are recommended that should realize a reduction of energy
consumption while not hurting the efficiency of the OR department. An increase in the efficiency of
the OR department can be realized.

Recommendations
Specific recommendations to the Amsterdam UMC are stated in this section. The section first
mentions the practical implications for the hospital. After this, recommendations for future research
are given.

Implications
The clinical OR department currently consists of solely class 1+ ORs. This is unnecessary as class 1+
ORs are only recommended for major joint replacements as depicted in Appendix D. Major joint
replacements make up for 7.36% of all executed procedures in 2019. ⌈20 ∗ 0.0736⌉ = 2 class 1+ ORs
suffice, which implies 18 class 1 ORs. The OR classification depicted in Figure 5-1 is recommended.
Four instead of five 24/7 ORs suffice. One of the 24/7 ORs is classified as class 1+, while the remaining
three 24/7 ORs are class 1 ORs. The elective ORs which switch to nighttime operation consists of one
class 1+ OR, while the other 15 elective ORs are class 1 ORs. Two class 1 ORs should not be used for
elective surgeries, as these ORs are destined for emergency surgeries only.
The class 1+ ORs must have a minimum air quality of ISO 7, the number of air changes must be at least
20 per hour, and the recovery time should not exceed three minutes. These requirements are already
met in the Amsterdam UMC hospital using the installed UDF HVAC systems. These systems are
equipped with a decentralized clean air system. It is recommended to install a centralized clean air
system for the class 1+ ORs, as this saves energy. The class 1 ORs must meet at least ISO 5
requirements, with a minimum number of air changes of 20 per hour and a recovery time of maximum
20 minutes. These requirements can be met using turbulent HVAC systems. These systems are
recommended as they consume less energy relative to UDF HVAC systems. A recirculation system is
advised as this minimizes the energy consumption of turbulent HVAC systems.

It is advised to equip all ORs with the same advanced equipment for all surgeries. This simplifies the
scheduling of surgeries significantly. On a tactical level, the LEPST* heuristic should be applied to the
clinical OR department of the Amsterdam UMC to minimize the average maximum completion time.
The LEPST* heuristic outperforms the LEPT* heuristic. It is recommended to automate this heuristic
as depicted in Appendix M.3. This code was implemented in the MC simulation. Applying the heuristic
by hand takes significant time, which is inefficient. Investing in automating the heuristic saves a
significant quantity of time. A programming specialist is required to automate the LEPST* heuristic.

Conclusions and recommendations 54

Further research
- The LEPST* heuristic does not include urgency classifications. The heuristic can be altered to
include the urgency classifications so that urgent procedures are given high priority.
- It needs to be verified that the current UDF HVAC systems can be converted to turbulent HVAC
systems while meeting class 1 requirements without major investments. Engineering
specialists are required to verify this. Research also needs to investigate cost savings.
- This research assumes that sufficient resources (personnel, equipment) are available for more
efficient scheduling. It needs to be verified that the more efficient scheduling is viable with
the current resources. If this is not viable, increasing the hospital’s resources might be
considered.
- The scope of this research is the clinical OR department. Researching improvements for the
day center might reduce more energy.
- The possibility to divide the ORs into class 1+ ORs, class 1 ORs, class 2 ORs, and even individual
treatment rooms should be researched to realize more energy consumption reductions.
- Research the setup times more extensively, as one year does not include significant amounts
of data concerning setup times.
- No indication of the surgery type was included in the data of 2019. Assumptions had to be
made to determine the elective surgeries. The results of this research can be more accurate
when researching which were emergency surgeries and which were elective surgeries.

Discussion
The goal of this research is to reduce the energy consumption of the clinical operating room
department of the Amsterdam UMC by answering the following research question: “How can the
planning of the HVAC system usage and the scheduling of the surgeries in the ORs be optimized so that
a reduction in energy consumption is realized without compromising the efficiency of the OR while
obeying the guideline concerning air quality in the OR? “

This research resulted in an expected reduction of approximately 308 tonnes of CO2e annually
compared to 2019 by applying the LEPST* heuristic for surgery scheduling. This heuristic schedules
surgeries on two types of ORs: Two ORs of class 1+ and 16 ORs of class 1. Two remaining ORs are used
for emergency procedures. These are ORs of class 1. Four instead of five ORs will be 24/7 ORs, meaning
that four instead of five ORs will not switch to nighttime operation, saving energy.

This is a significant decrease in CO2e emissions, and this study can therefore have a substantial impact
on both the entire healthcare industry and the global environment as this research might inspire other
hospitals besides the Amsterdam UMC to reduce their CO2e emissions. Niek Sperna Weiland (Hospital
supervisor, anesthetist, and member of the Green Team OR) is satisfied with the clear conclusions and
results of this research, and he thinks that the Amsterdam UMC can elaborate on this research in the
future. He is very enthusiastic about the substantial expected energy savings.

A limitation of this study is that some assumptions had to be made concerning setup times. For
example, we assumed setup times to be between zero and 90 minutes. This might be different in
reality, affecting the criterion-related validity of the research. Additionally, a Monte Carlo simulation
is used. This simulation method does not necessarily result in optimum results as the simulation is
stochastic. The scheduling problem is NP-hard, and it uses a heuristic to find a near-optimal solution.
This research, therefore, results in approximations of optimal solutions, but it does not provide exact
solutions. However, the Monte Carlo simulation and the heuristic result in reliable approximations.

Discussion 55
References
Amsterdam UMC. (2018). A. Luchttechnischbeheersplan OK (Versie 1). Amsterdam. Retrieved May
19, 2022

Amsterdam UMC. (2022). About Amsterdam UMC . Retrieved from Amsterdam UMC:
https://www.amsterdamumc.org/en/about.htm

Arnaout, J.-P. M., & Kulbashian, S. (2008, December 7). Maximizing the utilization of operating
rooms with stochastic times using simulation. Retrieved May 31, 2022, from
https://doi.org/10.1109/WSC.2008.4736245

Azimpoor, S., & Taghipour, S. (2019, October 30). Optimal job scheduling and inspection of a
machine with delayed failure. International Journal of Production Research, 6453-6473.
Retrieved May 4, 2022

C2G. (2021). What is plenum? Retrieved from C2G - A brand of legrand website:
https://www.cablestogo.com/learning/library/connected-classroom/what-is-plenum

Chauveaux, D. (2015, February). Preventing surgical-site infections: Measures other than antibiotics.
Orthopaedics & Traumatology: Surgery & Research, 101(1), S77-S83.
doi:https://doi.org/10.1016/j.otsr.2014.07.028
Commissie kwaliteitsdocumenten NVOG. (2020). Memo luchtbehandeling op operatiekamers.
Retrieved May 2, 2022
Cooper, D., & Schindler, P. (2014). Business Research Methods. Chegg Books. Retrieved April 30,
2022

Cubí Montanyà, E. (2014). Energy efficient ventilation strategies for surgery rooms. Universitat
Politècnica de Catalunya. Retrieved May 2, 2022

European Environment Agency. (2019). Indicator assessment. Retrieved September 20, 2022, from
Greenhouse gas emission intensity of electricity generation in Europe:
https://www.eea.europa.eu/data-and-maps/indicators/overview-of-the-electricity-
production-3/assessment

Graham, R., Lawler, E., Lenstra, J., & Rinnooy Kan, A. (1979). Optimization and approximation in
deterministic sequencing and scheduling: A survey. Annals of Discrete Mathematics, 287-
326. doi:https://doi.org/10.1016/S0167-5060(08)70356-X

Heerkens, H. (2017). Solving Managerial Problems Systematically. Noordhoff. Retrieved April 30,
2022

Inspectie voor de Gezondheidszorg. (2016). Toetsingskader luchtbeheersing operatieafdeling.

Retrieved May 25, 2022, from Ministerie van Volksgezonsheid, Welzijn en Sport:
https://mediservices.nl/wp-content/uploads/2017/03/toetsingskader-luchtbeheersing-
2014-2016.pdf

Kaplan Financial Limited. (2020). Information for management. Retrieved from Kaplan Financial
Knowledge Bank: https://kfknowledgebank.kaplan.co.uk/risk-ethics-and-
governance/internal-control-systems/information-for-management
Knudsen, R., Knudsen, S., Nymark, T., Anstensrud, T., Jensen, E., La Mia Malekzadeh, M., &
Overgaard, S. (2021). Laminar airflow decreases microbial air contamination compared with

References 56
 turbulent ventilated operating theatres during live total joint arthroplasty: a nationwide
survey. Journal of Hospital Infection, 113, 65-70. Retrieved May 6, 2022

L, B. (2022). Groene OK - Plaatjes met kosten en CO2. Amersfoort.

Nederlandse Vereniging voor Heelkunde (NVvH). (2017). Spoed-OK. Retrieved from Spoedingrepen:
http://www.spoedingrepen.nl/#/spoedingrepen

Nederlandse Vereniging voor Medische Microbiologie. (2022). Richtlijn Luchtbehandeling in

operatiekamers en behandelkamers. Retrieved May 2, 2022

NEN, Stichting Koninklijk Nederlands Normalisatie Instituut. (2015, August). Thermal performance of
buildings - Determination of air permeability of buildings - Fan pressurization method. 26.
Retrieved May 4, 2022

Pichler, P.-P., Jaccard, I. S., Weisz, U., & Weisz, H. (2019, May 24). International comparison of health
care carbon footprints. Environmental Research Letters, 14. Retrieved May 4, 2022

Pinedo, M., & Hadavi, K. (1992, May 30). Scheduling: Theory, Algorithms and Systems Development.
Retrieved 2022, from Operations Research Proceedings 1991: https://doi.org/10.1007/978-
3-642-46773-8_5

Steen - de Vlaming, K., & Hamersveld, S. (2021). OK-reglement (Klinische OK en dagcentrum).

Amsterdam.

Traversari, A., van Heurman, S., van Tiem, F., Bottenheft, C., & Hinkema, M. (2019, September).
Design variables with significant effect on system performance of unidirectional
displacement airflow systems in hospitals. Journal of Hospital Infection, 103(1), 81-87.
doi:https://doi.org/10.1016/j.jhin.2019.03.009

Van der Steen - de Vlaming, K. (2022). Dag- en nachtschakeling luchtbehandeling. Amsterdam UMC,
Amsterdam. Retrieved May 24, 2022

Van der Steen - de Vlaming, K. (2022). Operatiekamer klassen, prestatieniveau's en ingrepen.

Amsterdam UMC, Amsterdam. Retrieved May 24, 2022

Werkgroep Infectie Preventie. (2011). Omstandigheden (kleine) chirurgische en invasieve ingrepen.

WIP. Retrieved May 2, 2022

Werkgroep Infectie Preventie. (2014). Luchtbehandeling in operatiekamer en opdekruimte in

operatieafdeling klasse 1. WIP. Retrieved May 2, 2022

Yang, X. (2011). Metaheuristic optimization. Scholarpedia, 6, 11472.

doi:http://dx.doi.org/10.4249/scholarpedia.11472

References 57
Appendices
Appendix A: Map of the Amsterdam UMC

Figure A-1: Amsterdam UMC location AMC map

Appendices 58
Appendix B: Schematic overview of the OR department’s pressure levels

Figure B-1: Schematic overview of the clinical OR department's pressure levels

Appendices 59
Appendix C: Semi-structured interviews
This appendix depicts the structured element of the interviews that have been executed during this
research. The interviews are semi-structured, so there was a structure in each interview. However,
there were unstructured elements in the interviews. These elements are not depicted in this appendix.
This appendix provides a general summary of the interviews, and it does not contain transcripts.

Appendix C.1: Interview with Ingrid Spijkerman (Amsterdam UMC)

Mijn doel is om de OK afdeling te verduurzamen. Hierbij richt ik mij op de luchtbehandelingssystemen.
Mijn idee is om een aantal OKs op klasse 1+ te laten en de rest terug te schalen naar klasse 1 door het
luchtdebiet te verminderen. Dit bespaart energie. Ook ben ik aan het kijken naar de mogelijkheden
om de planning van de operaties efficiënter te maken. Natuurlijk heb ik ook onderzoek gedaan naar
de vorige richtlijnen en de nieuwe richtlijnen. Hier zou ik een paar vragen over willen stellen.

- Waarom is de nieuwe richtlijn opgesteld?

Belangenverstrengeling in de vorige richtlijn door multidisciplinair team. Veel discrepantie en
verwarring rondom de richtlijn maakte een nieuwe richtlijn nodig. De Inspectie
Gezondheidszorg (IGZ) handhaafde de orthopedierichtlijn.

- Op welke manier was u betrokken bij het samenstellen van de nieuwe richtlijn?
Ik ben lid van de expertgroep Luchtbehandeling op de OK.

- Is de energieverbruik en duurzaamheid een belangrijke factor geweest tijdens het opstellen

van de nieuwe richtlijn?
De duurzaamheid komt naar voren in de operationele kosten. Operationele kosten zijn de
kosten om een luchtbehandelingsysteem operationeel te houden. Energiekosten en
personeelskosten voor controle en onderhoud zijn hierbij de grootste kostenposten. In de
WIP-richtlijn ‘Luchtbehandeling in operatiekamer en opdekruimte in operatieafdeling klasse
1’ uit 2014 concludeerde de expertgroep dat op basis van het beschikbare bewijs geen
uitspraak kan worden gedaan over de kosten van een mengend ten opzichte van een verticaal
UDF systeem. Ook de huidige werkgroep kan niet inschatten of er een verschil is in
operationele kosten tussen het gebruik van een UDF en een mengend system, omdat de
grootte van deze kostenposten afhangt van veel verschillende lokale factoren. De richtlijn
noemt ook dat, in de regel, mengende systemen minder energie verbruiken. Hier is alleen niet
veel onderzoek naar gedaan door de expertgroep.

- Wat is in uw ogen verandert in de nieuwe richtlijn ten opzichte van de vorige richtlijn?
Een van de veranderingen is het loslaten van het advies een UDF systeem te hanteren. Eisen
omtrent de luchtkwaliteit mogen behaald worden met beide systemen.

- Wat is de reden van de nieuwe classificatie (Klasse 1+ en klasse 1)?

De expertgroep stelde klasse 1 samen en voor de orthopaedische operaties ontstond klasse
1+. Wij vonden dat er een andere classificatie nodig was omdat er veel discrepantie was
omtrent de vorige classificatie.

- Hoe komt het dat men overal klasse 1 prestatieniveau 1 OKs is gaan bouwen? De richtlijn van
2014 gaf al aan dat dit enkel nodig was voor gewrichtsvervangende operaties. Het AMC
bijvoorbeeld heeft bij de klinische OKs alleen maar OKs van klasse 1 prestatieniveau 1.

Appendices 60
 Uit gemakzucht. Het is natuurlijk makkelijker om geen onderscheid te maken in de OKs.
- In hoeverre verschillen OKs met klasse 1 prestatieniveau 2 van klasse 2 OKs en in hoeverre
mogen er meer operaties uitgevoerd worden in klasse 1 prestatieniveau 2 OKs?
Het ontwerp verschilt natuurlijk aanzienlijk, de klasse 2 OKs hebben geen
luchtbehandelingssysteem. In de nieuwe richtlijn staat een uitgebreide beschrijving van de
operaties die in klasse 2 uitgevoerd mogen worden.

Appendix C.2: Interview with Jelle Koeman (Amsterdam UMC)

Mijn doel is om de OK afdeling te verduurzamen. Hierbij richt ik mij op de luchtbehandelingssystemen.
Mijn idee is om een aantal OKs op klasse 1+ te laten en de rest terug te schalen naar klasse 1 door het
luchtdebiet te verminderen. Dit bespaart energie. Ook ben ik aan het kijken naar de mogelijkheden
om de planning van de operaties efficiënter te maken. Ik heb een aantal vragen met betrekking tot de
luchtbehandelingssystemen.

- Waarom gaan de luchtbehandelingssystemen terug naar 50% tijdens de nachtschakeling?

Kunnen ze niet uit uit of anders nog lager?
Ze kunnen niet uit doordat dit de spoed OKs die 24/7 aan staan nadelig kunnen beinvloeden
door luchtlekkages.

- Zou een situatie waarbij grote gewrichtvervangende operaties uitgevoerd worden in

specifieke Klasse 1+ OKs een mogelijkheid zijn? Hierbij zou de rest van de OKs Klasse 1 kunnen
hebben.
Nu is dat niet echt een mogelijkheid, alle OKs zijn ingericht voor klasse 1+ en kunnen niet
zomaar terug geschaald worden zonder verbouwingen.

- Waarom 100 luchtverversingen per uur? Volgens de richtlijnen moeten klasse 1 OKs minimaal
20 luchtverversingen
100 houdt de gerecirculaire en primaire luchtverversingen in. 20 is alleen primair.

- kWh verbruik van een enkel luchtbehandelingssysteem klasse 1 prestatieniveau 1 en 2?

Weet ik niet uit mijn hoofd, zal ik aanvragen bij het bedrijf dat metingen gedaan heeft en het
naar je doorsturen.

- Is het mogelijk om een keer de technische ruimte te zien samen met Anne Timmermans?
Prima, gaan we inplannen. Ik zal je dan een rondleiding geven.

Appendices 61
Appendix C.3: Interview with Karin de Vlaming (Amsterdam UMC)
Mijn doel is om de klinische OK afdeling te verduurzamen. Hierbij richt ik mij op de
luchtbehandelingssystemen. Mijn idee is om een aantal OKs op klasse 1+ te laten en de rest terug te
schalen naar klasse 1 door het luchtdebiet te verminderen. Dit bespaart energie. Ook ben ik aan het
kijken naar de mogelijkheden om de planning van de operaties efficiënter te maken, hier zou ik wat
vragen over willen stellen.

- Hoe wordt de dagelijkse planning van operaties nu gemaakt? Zit hier een bepaalde methode
achter?
De planning komt tot stand als gevolg van de jaarlijkse begroting van de OK-uren. Hierna wordt
een meer concrete wekelijkse planning gemaakt die voortkomt uit een wekelijkse bijeenkomst
tussen alle specialismen. Uiteindelijk wordt de dagelijkse planning vervolgens dagelijks
afgerond.

- Worden bepaalde OKs specifiek toegewezen aan bepaalde specialismen?

De OKs zijn in de basis allemaal hetzelfde, maar ze verschillen in apparatuur en uitrusting. Om
deze reden worden er bepaalde specialismen over het algemeen in dedicated OKs ingepland.

- Waarom gaan de luchtbehandelingssystemen terug naar 50% tijdens de nachtschakeling?

Kunnen ze niet uit uit of anders nog lager?
Deze vraag kun je beter stellen aan de technici van het AMC, zij weten precies hoe de
systemen er hier uitzien en waarom hiervoor gekozen is. Hier is volgens mij voor gekozen om
de andere OKs stabiel te houden en om te zorgen dat de OKs snel weer up en runing zijn. De
technici hebben voor de 50% gekozen en dit zal ongetwijfeld niet nog minder kunnen.

- Hoeveel wisseltijd is er nodig om na een operatie de OK voor te bereiden voor de volgende

operatie en verschilt deze tijd per operatie?
Verschillende operaties hebben natuurlijk ander apparatuur nodig en dit veroorzaakt
verschillen in de wisseltijd, maar de meeste wisseltijd komt van het transporteren van de
patiënten.

- In de data van 2019 zie ik veel tijd tussen opeenvolgende operaties in. Hoe komt dat?
Het transporteren van patiënten kost natuurlijk tijd. Daarnaast zijn er ook acute OKs waar de
operaties op dezelfde dag pas ingepland worden als dit nodig is. Dit zorgt ervoor dat er veel
tijd tussen de operaties zitten.

- In hoeverre denkt u dat het nodig is dat er 5 OKs 24/7 aan staan?
Om de patiënt gerust te stellen door genoeg OKs vrij te geven moeten er genoeg OKs
beschikbaar zijn. Ook duurt het in sommige gevallen te lang om een OK op te starten wanneer
hij nog uitstaat. Er is dus voor gekozen om 5 OKs ’s nachts vrij te geven om zo de zorgkwaliteit
te waarborgen. Wel denk ik dat het mogelijk is om er een minder in te zetten ’s nachts.

- Zou een situatie waarbij grote gewrichtvervangende operaties uitgevoerd worden in

specifieke Klasse 1+ OKs een mogelijkheid zijn? Hierbij zou de rest van de OKs Klasse 1 kunnen
hebben.

Appendices 62
 Als dit technisch mogelijk is zou dit zeker een mogelijkheid zijn. Dit moet gecheckt worden
met de technici van het AMC aangezien zij precies weten wat er kan in dit ziekenhuis. We
hebben namelijk de meest geavanceerde luchtbehandelingssystemen. Validatie is dus erg
belangrijk.

- En hoeveel Klasse 1+ OKs denkt u dat er dan nodig zullen zijn? Denkt u dat 2 voldoende zullen
zijn? (klasse 1+ is alleen voor grote gewrichtsvervangende operaties)
Ik denk dat 2 á 3 voldoende moet zijn. Houd hierbij wel rekening met onderhoud en potentiële
storingen. Dit mag niet voor problemen zorgen dus ik zou wel aan de veilige kant zitten.

Appendix C.4: Interview with Wilco van Nieuwenhuyzen (Interflow)

Mijn doel is om de OK afdeling te verduurzamen. Hierbij richt ik mij op de luchtbehandelingssystemen.
Mijn idee is om een aantal OKs op klasse 1+ te laten en de rest terug te schalen naar klasse 1 door het
luchtdebiet te verminderen. Dit bespaart energie. Ook ben ik aan het kijken naar de mogelijkheden
om de planning van de operaties efficiënter te maken. Ik heb een paar vragen met betrekking tot de
luchtbehandelingssystemen in het AMC.

- Heeft Interflow de OKs in het AMC opgeleverd en voert Interflow het onderhoud uit?
Nee, Interflow voert alleen de validatie van de OKs uit. Dit betekent dat wij alle waarden
meten en controleren of deze op gewenste waarden zitten.

- Weet u wat het huidige verbruik van de luchtbehandelingssystemen in operatiekamers met

klasse 1 prestatieniveau 1 in het Amsterdam UMC is?
Dat weet ik niet uit mijn hoofd. Ik heb hier geen exacte gegevens voor maar ik kan je wel een
Excel bestand doorsturen waarin het verbruik van de OKs in het AMC benaderd kan worden.
Door verschillende waardes als input in te vullen in het bestand kun je dit berekenen.

- Weet u wat het verbruik in het AMC zou zijn als het luchtdebiet en ventilatievouden
verminderd worden?
Om dit te benaderen kun je het Excel bestand gebruiken, waarbij je de input parameters
luchtdebiet en ventilatievouden verminderd.

- ’s Nachts wordt de ventilatie van 5 van de luchtbehandelingssystemen teruggeschakeld naar

50%. Weet u wat dan het energieverbruik is en is het niet mogelijk om de ventilatie verder
terug te schakelen?
Om dit te benaderen kun je ook het Excel bestand gebruiken.

- Hoe zien de luchtbehandelingssystemen er bij het Amsterdam UMC uit? Hebben deze
centrale verse luchtkasten of decentrale verse luchtkasten per OK?
Ik weet niet zeker hoe deze systemen eruit zien, wel denk ik dat er per OK een decentrale
verse luchtkast van beschikking is.

- Het aantal luchtwisselingen is momenteel ca. 100 keer per uur. De richtlijnen schrijven
minimaal 20 luchtwisselingen per uur voor. Hoe is dit verschil te verklaren? Kan dit minder
zonder dat het systeem zijn verdringende werking verliest?
Dit komt doordat de luchtbehandelingssystemen een verdringende werking hebben.
Hierdoor moet het luchtdebiet hoog genoeg zijn om de verdringende werking te behouden.

Appendices 63
 Met 20 luchtwisselingen per uur is het niet mogelijk om die verdringende werking te
verzekeren. Het luchtbehandelingssysteem verandert dan in een mengend
luchtbehandelingssysteem. Dat is de reden dat het ziekenhuis het aantal luchtwisselingen
momenteel op circa 100 per uur houdt.

Appendix C.5: Interview with Anne Brouwer (Royal HaskoningDHV)

Mijn doel is om de OK afdeling te verduurzamen. Hierbij richt ik mij op de luchtbehandelingssystemen.
Mijn idee is om een aantal OKs op klasse 1+ te laten en de rest terug te schalen naar klasse 1 door het
luchtdebiet te verminderen. Dit bespaart energie. Ook ben ik aan het kijken naar de mogelijkheden
om de planning van de operaties efficiënter te maken. Ik heb een paar vragen met betrekking tot het
omzetten van de verdringende luchtbehandelingssystemen naar mengende
luchtbehandelingssystemen in het AMC.

- Kan een verdringend systeem teruggeschakeld worden van ISO 5 naar ISO 7 zonder zijn
verdringende werking te verliezen (en minimaal 20 luchtwisselingen per uur en een
hersteltijd van maximaal 20 minuten te behouden)?

Als het systeem zijn verdringende werking behoudt, dan bereik ook altijd ISO 5 en zal de hersteltijd
ook gewoon 3 minuten blijven. Verdringend wil zeggen dat je echt de lucht vanuit het plenum naar
beneden wegduwt. Er is daarmee een constante schone luchtstroom over het operatiegebied. Bij
verlagen van de luchtstroom (luchtdebiet) verliest de luchtstroom haar verdringende werking en
wordt het meer een mengend systeem. De luchtwisselingen gaan dan terug van 70-80 per uur
(verdringend) naar de genoemde 20 luchtwisselingen. Dus het antwoord is nee, met dergelijke lage
ventilatievouden is er geen sprake van een verdringend systeem zoals bedoeld in de richtlijnen.

- Is het mogelijk om verdringende luchtbehandelingssystemen zonder een verbouwing om te

zetten naar een mengend systeem en hoe kan dit gerealiseerd worden? Bespaart dit
energieverbruik ten opzichte van een downflowsysteem bij minimaal 20 luchtwisselingen,
ISO 7, en een hersteltijd van maximaal 20 minuten? Hoeveel energie zou het systeem dan
verbruiken?

Ja, dit is heel goed mogelijk. Je kunt een regeling inbouw waarmee het luchtdebiet kan worden
verminderd van hoog (verdringend) naar laag (mengend). Bij verlagen van het debiet zal het
inblaasplenum functioneren als een groot inblaasrooster en mengende ventilatie gaan geven. Het is
hiervoor wel noodzakelijk dat de ventilatoren in de luchtbehandelingskast inderdaad regelbaar zijn
en je moet dus wat extra regeltechniek toevoegen. In het OK bedienpaneel kun je dan een
omschakelknop opnemen. Belangrijk om goed te valideren dat bij omschakelen de OK inblaas
inderdaad weer heen en terug gaat naar de juiste waarden en in de hoogste stand ook weer voldoet
aan ISO 5, hersteltijd 3 minuten. Belangrijk ook om de schakelstand onderdeel uit te laten maken van
de time-out procedure om schakelfouten te voorkomen. Ik zal een bijlage sturen met een indicatie
van de berekeningen.

- Omdat klasse 1+ (ISO5) nog maar noodzakelijk is voor een aantal ingrepren (uitsluitend
orthopaedie) denk ik aan een statisch systeem waarbij je een paar OKs inregelt als ISO 5 en
de rest als ISO 7.

Ja, het aantal ingrepen waarvoor Klasse 1+ of P1 nodig is, is heel beperkt. Het gaat alleen om
gewrichtsvervangende operaties of operaties met soortgelijk implantaten/endoprothesen. Kleine
implantaten hoeft zeker niet in klasse 1+. En alle andere operaties van andere discipline hoeven van
de verenigingen niet in Klasse 1+.

Appendices 64
Even voor de goede orde, dat was in de WIP 2014 ook al heel duidelijk aangegeven! Wat dat betreft
is er met de nieuwe luchtrichtlijn helemaal niets nieuws onder de zon. Alleen op een of andere manier
heeft men dat niet goed gelezen en is iedereen overal P1 OK’s gaan maken.

- Nog een laatste vraag: als je het ventilatievoud gaat verlagen richting ISO 7 voorwaarden,
maar je houdt ISO5 voorwaarden op, bijvoorbeeld, de OK ernaast. Is dan de drukhierarcie
nog wel goed te garanderen? Met andere woorden blaas je dan geen lucht vanuit een ISO5
OK een ISO 7 OK op?

Nee, dat zou geen probleem moeten zijn. Met de verse lucht wordt de druk hiërarchie geregeld. Dit
werkt over het gehele OK complex. Bovenop de verse lucht wordt per OK lucht gerecirculeerd (vaak
door een aparte recirculatie unit per OK). Die unit jaagt constant lucht over het plenum. Wat je als je
van Klasse 1+ naar Klasse 1 gaat (of van P1 naar P2) is dat je die recirculatie hoeveelheid gaat
reduceren. Maar de verse lucht moet gelijk blijven (vanwege luchtjes, anesthesiedampen, bezetting
personen en dus om de drukhierarchie te handhaven).

Met recirculatie in een kamer kan je geen drukverschillen creëren naar andere ruimten, dat kan
alleen via het verse luchtdeel door toe- en afvoer hoeveelheden verschillende in te stellen van elkaar
(maar toevoer dan afvoer is overdruk en andersom).

Let op: het kan zijn dat bij jullie de systemen net even anders in elkaar zitten, maar in basis is dit hoe
het zou moeten werken.

Je kunt ook nog in de nachturen het verse luchtdeel gaan aftoeren, maar dan loop je op een gegeven
moment wel tegen grens van het handhaven van de druk hierarchie aan, maar zeker ook een
interessante mogelijkheid.

En natuurlijk gewoon overdag en ’s avonds de Klasse 1+ op laag toeren zetten als ie niet gebruikt
wordt (kan evt. op een bewegingsmelder). Na 15 minuten is zo’n systeem weer up en running.

Overigens nog wel een opmerking over ISO 5 en 7. Deze waarden gelden in rust. En die ISO 5 haal je
uiteindelijk in een P2 OK wel als je maar lang genoeg wacht. Wat het name het verschil is is de
hersteltijd. Die is 3 minuten voor klasse 1+ en 20 minuten voor de andere. Om die 3 minuten heb je
een grotere luchthoeveelheid om deeltjes snel af te kunnen voeren.

Appendices 65
Appendix D: Surgery classification as proposed in the revised guideline
Appendices 67
Table D-1: Surgery classification as proposed in the revised guideline, completed by each specialisms' corresponding association

Appendices 68
Appendix E: Surgery classification and allocation example

Table E-1: Surgery classification and allocation example

Appendices 69
Appendix F: Example of daily planning

Figure F-1: Example of daily planning

Appendices 70
Appendix G: Requirements for critical parameters

Table G-1: Requirements for critical parameters in the clinical ORs of the Amsterdam UMC

Appendices 71
Appendix H: Frequently applied HVAC system configurations

Figure H-1: Class 1 performance level 1 – Centralized clean air system

Figure H-2: Class 1 performance level 1 – Decentralized clean air system

Appendices 72
Figure H-3: Class 1 performance level 2 – Without recirculation system

Figure H-4: Class 1 performance level 2 – With recirculation system

Appendices 73
Appendix I: Energy consumption tool assumptions
- 18 ORs of 50 𝑚 2 cleanroom, 2 ORs (OR 7 + OR 8) of 70 𝑚 2 cleanroom;
- ORs 4, 5, 8, 12, and 14 are 24/7 ORs;
- There are 52 weeks and 8,760 hours in a year. ORs shifting to nighttime operation operate in
daytime operation between 6:30 and 21:00 (14.5 hours) from Monday – Friday, excluding
official holidays (8 days in the Netherlands);
- Airflow rate reduction during nighttime operation is 50%. Reduction of energy consumption
due to the airflow rate reduction is also assumed to be 50%;
- €0.09 / kWh (J. Koeman, personal communication, June 16, 2022);
- Airflow flow rate (supplemented ventilator) is equal to 6,250 𝑚 3 /ℎ;
- Airflow flow rate (OR ventilator) is equal to 9,000 𝑚 3 /ℎ;
- Airflow flow rate (storage room ventilator) is equal to 3,000 𝑚 3 /ℎ;
- Airflow flow rate (exhaust ventilator) is equal to 2,500 𝑚 3 /ℎ;
- External pressure (supplemented ventilator) is equal to 150 𝑃𝑎;
- External pressure (OR ventilator) is equal to 350 𝑃𝑎;
- External pressure (storage room ventilator) is equal to 350 𝑃𝑎;
- External pressure (exhaust ventilator) is equal to 200 𝑃𝑎;
- Internal pressure (supplemented ventilator) is equal to 343 𝑃𝑎;
- Internal pressure (OR ventilator) is equal to 552 𝑃𝑎;
- Internal pressure (storage room ventilator) is equal to 600 𝑃𝑎;
- Internal pressure (exhaust ventilator) is equal to 100 𝑃𝑎;
- Static efficiency (all ventilators) is equal to 0.75;
- Motor efficiency (all ventilators) is equal to 0.85;
- Lights electric power is equal to 15 𝑊/𝑚 2;
- Heat load is equal to 25 𝑊/𝑚2;
- Coefficient of performance cooling machine is equal to 4.1;
- Other electric power (pumps, control system, etc.) is equal to 1.5 kW

Appendices 74
Appendix J: Surgical procedures for class 1+ ORs

Procedure Primary_procedure_id
BEKKEN - EXTERNE FIXATEUR NA GESLOTEN REPOSITIE 1071002320
BEKKEN - INCISIE EN SPOELDRAINAGE BEKKENGEWRICHT 1071002377
BEKKEN - KRAPPE EXCISIE AFWIJKING - EXCOCHLEATIE BOT 1071002288
BEKKEN - OSTEOSYNTHESE 1071002322
BEKKEN - OSTEOSYNTHESE ACETABULUM 1071002418
BEKKEN - OSTEOSYNTHESE ACETABULUM OVV PLAAT OS KLEIN 1071003978
BEKKEN - OSTEOSYNTHESE OVV GECANN. SCHROEVEN 1071003980
BEKKEN - OSTEOSYNTHESE OVV PLAAT OS GROOT 1071003979
BEKKEN - OSTEOSYNTHESE OVV PLAAT OS KLEIN 1071003977
BEKKEN - RADICALE EXCISIE AFWIJKING BOT 107100640
BEKKEN - SPONGIOSAPLASTIEK 1071002309
BEKKEN - VERWIJDEREN OSTEOSYNTHESEMATERIAAL 1071002715
BEKKEN - WINNEN BOT UIT CRISTA VOOR AUTOTRANSPLANT. 1071002281
BOT - BIOPSIE 1071002791
BOT - DEF.PERCUT.EPIFYSIOD.-WEGNEMEN-DEEL-EPIF 107100559
BOT - DRAINAGE ABCES ONTSTEKINGEN 1071002767
BOT - EXCOCHL.EN-OF SEKWESTROTOM.-MIDDELGR.BEEND. 1071004435
BOT - TRANSPLANTATIE BOT EN-OF KUNSTSTOF 1071004134
BOT - VERWIJD.EXOSTOSEN KLEINE BEEND.- VOOR DE 1E 107100699
BOT - VERWIJDEREN 1 OF MEER SCHROEVEN UIT EEN BOT 1071002731
BOT - VERWIJDEREN CENTRALE MERGPEN 1071002714
BOT - VERWIJDEREN EXTERNE FIXATEUR 107100686
BOT - VERWIJDEREN OSTEOSYNTHESEMATERIAAL 1071002742
BOT - VERWIJDEREN PLAAT EN SCHROEVEN UIT 1 BOT 1071002728
BOT - VERWIJDEREN PLATEN EN SCHROEVEN 1071002723
BOT - VERWIJDEREN SNAARTRACTIE OF EXT.FIXATEUR 1071002739
BOT-SPIEREN - BEHANDELING GROTE DIEPE ABCESSEN 1071003875
BOVENARM - AMPUTATIE 1071002067
BOVENBEEN - AMPUTATIE 1071002492
BOVENBEEN - KRAPPE EXCISIE BOT 1071002402
BOVENBEEN - KRAPPE EXCISIE BOT OVV FEMURKOP 1071003992
BOVENBEEN - REVISIE AMPUTATIESTOMP 1071002494
BOVENBEEN - VARISERENDE DEROTERENDE OSTEOTOMIE 1071002408
CLAVICULA - PLAATOSTEOSYNTHESE 1071002019
ELLEBOOG - PROTHESE IMPLANTATIE RADIUSKOP 1071004732
ENKEL - PLAATOSTEOSYNTH.BIMALLEOL. NA OPEN REPOS. 1071002637
ENKEL - PLAATOSTEOSYNTH.TRIMALLEOL.-OPEN REPOSITIE 1071002639
ENKEL - PLASTIEK LIGAMENT-CHRON.LUXATIE-ARTROTOMIE 1071002680
ENKEL - REVISIE GEWRICHTSPROTHESE 1071003885
ENKEL - SCHROEFOSTEOSYNTH.BIMALLEOL.NA OPEN REPOS. 1071002638
FEMUR - GESLOTEN REPOSITIE FRACTUUR-MERGPENFIXATIE 1071002431
FEMUR - PER-INTER-SUBTROCH.FRACT.HOEKPLAATOSTEOSYNT 1071002427
FEMUR - PER-INTER-SUBTROCH.FRACTUUR GAMMA-NAIL-FIX. 1071002426
FEMUR DIST. - HOEKPLAATOSTEOSYNTHESE 1071002442

Appendices 75
FEMUR DIST. - PLAATOSTEOSYNTHESE 1071002441
FEMUR DIST. - SCHROEFOSTEOSYNTHESE 1071002443
FEMUR PROX. - DYNAMISCHE HEUPSCHROEF - HALS 1071002421
FEMUR PROX. - HOEKPLAATOSTEOSYNTHESE - HALS 1071002420
FEMUR PROX. - INTRAMEDULLAIRE FIXATIE 1071002423
FEMUR PROX. - INTRAMEDULLAIRE FIXATIE OVV GAMMA NAIL LANG 1071004011
FEMUR PROX. - INTRAMEDULLAIRE FIXATIE OVV NAIL LFN 1071004007
FEMUR PROX. - INTRAMEDULLAIRE FIXATIE OVV TFN 107100463
FEMUR PROX. - OP.BEH.EPIFYSIOLYSIS CAPUT FEMORIS 1071002445
FEMUR PROX. - OSTEOSYNTHESE COLLUM FEMORIS 1071002419
FEMURSCHACHT- INTRAMEDULLAIRE FIX.NA GESL.REPOSITIE 1071002434
FEMURSCHACHT- INTRAMEDULLAIRE FIX.NA OPEN REPOSITIE 1071002433
FEMURSCHACHT- INWENDIGE FIXATIE ZONDER REPOSITIE 1071002448
FEMURSCHACHT- PLAATOSTEOSYNTHESE 1071002438
FIBULA - PLAATOSTEOSYNTHESE NA OPEN REPOS.SCHACHT 1071002535
FIBULA - SCHROEFOSTEOSYNTHESE NA OPEN REPOS.DISTAAL 1071002534
HEUP - INBRENGEN KOP-HALSPROTHESE - NNO 1071002470
HEUP - INBRENGEN KOP-HALSPROTHESE - NNO OVV MUELLER 1071004018
HEUP - REVISIE TOTAL HIP - BEIDE COMP.MET OPBOUW 1071002480
HEUP - REVISIE TOTAL HIP ACETABULUM MET OPBOUW 1071002478
HEUP - REVISIE TOTAL HIP FEMURCOMP. MET OPBOUW 1071002479
HEUP - TOTALE HEUPPROTHESE HYBRIDE 1071002476
HEUP - TOTALE HEUPPROTHESE MET HYDROXIE APATIET 1071002477
HEUP - VERVANGEN ONDERDEEL VAN HEUPPROTHESE 1071002473
HUMERUS - BEHANDELING PSEUDO-ARTROSE- FIX.EN PLASTIEK 1071002008
HUMERUS - INWENDIGE FIXATIE HUMERUS REPOSITIE 1071002012
HUMERUS - INWENDIGE FIXATIE HUMERUS REPOSITIE OVV GROOT LCP 1071004013
HUMERUS - MERGPENFIXATIE - SCHACHT 1071002031
HUMERUS - PLAATOSTEOSYNTHESE - SCHACHT 1071002027
HUMERUS - PLAATOSTEOSYNTHESE PROXIMAAL 1071002030
HUMERUS - PLAATOSTEOSYNTHESE SUPRA-TRANSCOND.FRACT. 1071002032
HUMERUS - PLAATOSTEOSYNTHESE SUPRA-TRANSCOND.FRACT. OVV GROOT LCP 1071004015
KNIE - PROTHESE HEMIARTROPLASTIEK MEDIAAL 1071003871
KNIE - REVISIE FEMURCOMPONENT KNIEPROTHESE 1071003879
KNIE - REVISIE KNIEPROTHESE 1071002800
KNIE - REVISIE KNIEPROTHESE ALLE COMPONENTEN 1071002798
KNIE - REVISIE TIBIACOMPONENT KNIEPROTHESE 1071002797
KNIE - VERVANGEN LAGER --SPACER -- KNIEPROTHESE 1071004716
KNIE - VERVANGEN ONDERDEEL VAN KNIEPROTHESE OVV INLAY WISSEL 107100471
KNIE - VERWIJDEREN PROTHESE GEWRICHT 1071002576
ONDERARM - OPER.BEH.FRACTUUR DISTALE RADIUS 1071004439
ONDERARM - OPERATIEVE BEHANDELING FRACTURA MONTEGGIA 1071004465
ONDERARM - PLAATOSTEOSYNTHESE ULNA EN RADIUS 1071002097
ONDERARM - PROTHESE IMPLANTATIE DISTALE ULNA 1071002125
ONDERBEEN - INTRAMEDULL.FIX.NA GESL.REPOS.CRURISFRACT. 1071002537
ONDERBEEN - PLAATOSTEOSYNTHESE NA OPEN REPOS.CRURIS 1071002536
POLS - PROTHESE IMPLANTATIE POLSGEWRICHT 1071002219
PRIMAIRE HEUP MET CEMENT, ROUTINE 107100790

Appendices 76
PRIMAIRE HEUP MET CEMENT, COMPLEX 107100792
PRIMAIRE HEUP MET CEMENT, INTERMEDIATE 107100791
PRIMAIRE HEUP ZONDER CEMENT COMPLEX 107100650
PRIMAIRE HEUP ZONDER CEMENT INTERMEDIATE 107100649
PRIMAIRE HEUP ZONDER CEMENT ROUTINE 107100648
RADIUS - PLAATOSTEOSYNTHESE DISTAAL 1071002100
RADIUS - PLAATOSTEOSYNTHESE INCL.GALEAZZI 1071002110
RADIUS - SCHROEFOSTEOSYNTHESE DISTAAL 1071002101
RADIUS - SCHROEFOSTEOSYNTHESE WGS RADIUSKOPFRACTUUR 1071002104
-RE-SPONDYLODESE 2 OF 3 OF 4 SEGMENTEN - CERVICAAL 1071004514
-RE-SPONDYLODESE 2 OF 3 OF 4 SEGMENTEN - THORACAAL 1071004515
-RE-SPONDYLODESE 2 OF 3 OF 4 SEGMENTEN- LUMBOSACRAAL 1071004517
-RE-SPONDYLODESE 2 OF 3 OF 4 SEGMENTEN-CRANIOCERVICAAL 1071004518
-RE-SPONDYLODESE 5 OF MEER SEGMENTEN CERVICOTHORAC 1071004506
-RE-SPONDYLODESE 5 OF MEER SEGMENTEN LUMBAAL 1071004505
-RE-SPONDYLODESE 5 OF MEER SEGMENTEN THORACAAL 1071004504
-RE-SPONDYLODESE 5 OF MEER SEGMENTEN THORACOLUMB 1071004507
SCHOUDER - PROTHESE IMPLANT.HUMERUSKOP-SCHOUDERKOM 1071002052
SCHOUDER - PROTHESE IMPLANTATIE HUMERUSKOP 1071002051
TIBIA - HOEKPLAATOSTEOSYNTHESE - PROX-PLATEAUFRACT. 1071002540
TIBIA - INTRAMEDULL.FIXATIE SCHACHT NA GESL.REPOS. 1071002530
TIBIA - INTRAMEDULL.FIXATIE SCHACHT NA GESL.REPOS. OVV ETN 10+11MM 1071004003
TIBIA - INTRAMEDULL.FIXATIE SCHACHT NA OPEN REPOS. 1071002531
TIBIA - INTRAMEDULL.FIXATIE SCHACHT NA OPEN REPOS. OVV ETN 10+11MM 1071004005
TIBIA - ONBLOEDIGE REPOSITIE-AANLEGGEN GIPS-SCHACHT 1071002525
TIBIA - OPER.BEH.INTRA-ARTICULAIRE PLATEAUFRACTUUR 1071002542
TIBIA - OPER.BEH.PSEUDO-ARTROSE MET BOTPLASTIEK 107100689
TIBIA - PLAAT-FIXATIE WEGENS PROX.TIBIAFRACTUUR 1071002544
TIBIA - PLAATOSTEOSYNTHESE SCHACHT 1071002528
TIBIA - SCHROEFOSTEOSYNTHESE - PROX-PLATEAUFRACTUUR 1071002541
TIBIA - SCHROEFOSTEOSYNTHESE TIBIA-PILON. 1071002647
ULNA - OPERATIEVE BEHANDELING SCHACHTFRACTUUR 1071002109
ULNA - PLAATOSTEOSYNTHESE INCL.MONTEGGIA 1071002107
ULNA - PLAATOSTEOSYNTHESE IVM OLECRANONFRACTUUR 1071002112
WERVELKOLOM - FRACT.BEH.LUMB.MET SPONDYL.EO.OSTEOSYN.POST 1071002319
WERVELKOLOM - INBRENGEN STACKABLE CAGE - CERVICAAL 1071002337
WERVELKOLOM - OPER.BEHAND.FRACTUUR MET INWENDIGE FIXATIE 1071002321
WERVELKOLOM - OPERAT.BIJSTELL.INTERN DISTRACTIESYST. 1071004219
WERVELKOLOM - OPERAT.PLAATSING INTERN DISTRACTIESYST. 1071004552
WERVELKOLOM - OPERATIEVE BEHANDELING WERVELFRACTUUR 1071002315
WERVELKOLOM - RESPONDYLODESE 1071002361
WERVELKOLOM - SPONDYLODESE CERV.ANT-KORT TRAJ-EO.INSTRUM. 1071002357
WERVELKOLOM - SPONDYLODESE CERV.POSTERIOR 1071002340
WERVELKOLOM - SPONDYLODESE CERV.POSTERIOR MET SCHROEVEN 1071002343
WERVELKOLOM - SPONDYLODESE CERV.POST-KORT TRAJ-EO.INSTRUM 1071002359
WERVELKOLOM - SPONDYLODESE CERVICAAL ANTERIOR 1071002344
WERVELKOLOM - SPONDYLODESE CERVICAAL ANTERIOR 1 SEGMENT 1071002347
WERVELKOLOM - SPONDYLODESE CRANIOCERVICAAL POSTERIOR 1071002341

Appendices 77
 WERVELKOLOM - SPONDYLODESE LUMB.ANT-KORT TRAJ-EO.INSTRUM. 1071004201
WERVELKOLOM - SPONDYLODESE LUMB.POST-KORT TRAJ-EO.INSTRUM 1071002355
WERVELKOLOM - SPONDYLODESE LUMBAAL POSTERIOR 1071002371
WERVELKOLOM - SPONDYLODESE LUMBAAL POSTERIOR OVV EXPEDIUM 1071004001
WERVELKOLOM - SPONDYLODESE THOR.POST-KORT TRAJ-EO.INSTRUM 1071002363
WERVELKOLOM - SPONDYLODESE THOR.POST-LANG TRAJ-EO.INSTRUM 1071002364
WERVELKOLOM - SPONDYLODESE THORACAAL POSTERIOR 1071002369
WERVELKOLOM - SPONDYLODESE THORACAAL POSTERIOR OVV EXPEDIUM 1071004002
WERVELKOLOM - SPONDYLODESE VAN DE DENS - ANTERIOR 1071002345
WERVELKOLOM - VERTEBRECTOMIE CERVICAAL MET SPONDYLODESE 1071002365
WERVELKOLOM - VERTEBRECTOMIE LUMBAAL MET SPONDYLODESE 1071002367
WERVELKOLOM - VERTEBRECTOMIE THORACAAL MET SPONDYLODESE 1071002366
WERVELKOLOM - VERTEBROPL.MET BALLON--KYPHOPLASTIEK-- LUMB 1071002312
Table J-1: Surgical procedures for class

Appendices 78
Appendix K: Implemented Excel functions
Appendix K.1: Day of the week
=TEXT($C2; "dddd")

Appendix K.2: Procedure classification

=VLOOKUP($G2; 'Procedure classification'!$A$2:$C$1709; 3; FALSE)

Appendix K.3: Setup times

=IFERROR(IF(AND(C4=C3; B4=B3); D4-E3; "")*1440; "")

Appendix K.4: CDF

=GAMMA.DIST($I3; $F$20; $F$21; TRUE)
=WEIBULL.DIST($I3; $F$20; $F$21; TRUE)
=NORM.DIST($I3; $F$4; $F$8; TRUE)
=LOGNORM.DIST($I3; $F$4; $F$8; TRUE)
=EXPON.DIST($I3; 1/$F$4; TRUE)

Appendix K.5: Critical value

=CHISQ.INV.RT(0,05; COUNT($I$3:$I$19)-1)

Appendix K.6: Surgery class probability

=COUNTIF($J$2:$J$9814; "Class 1+")
=COUNTIF($J$2:$J$9814; "Class 1")

Appendices 79
Appendix L: Original Excel file of historical data of 2019

Appendices 80
Appendix M: Data elective procedures 2019

Table L-1: Data elective procedures 2019

Appendices 81
Appendix N: Implemented Excel Visual Basic code
Appendix N.1: Generating processing times class 1+

Figure N-1: Excel VBA generating processing times class 1+

Appendices 82
Appendix N.2: Generating jobs MC simulation

Figure N-2: Excel VBA generating jobs

Appendices 83
Appendix N.3: Scheduling jobs MC simulation

Appendices 84
Figure N-3: Excel VBA scheduling jobs

Appendices 85
Appendix N.4: Iteration

Figure N-4: Excel VBA iteration

Appendices 86