Nurses and Midwives in the Digital Age 295

M. Honey et al. (Eds.)

© 2021 International Medical Informatics Association (IMIA) and IOS Press.
This article is published online with Open Access by IOS Press and distributed under the terms
of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0).
doi:10.3233/SHTI210726

AI in Healthcare
Eileen KOSKIa,1 and Judy MURPHYb
a
IBM TJ Watson Research Center, Yorktown Heights, New York, USA
b
IBM Global Healthcare

Abstract. The potential value of AI to healthcare, and nursing in particular, ranges

from improving quality and efficiency of care to delivering on the promise of
personalized and precision medicine. AI systems may become virtually
indispensable as ever more data is amassed about every aspect of health. AI can help
reduce variability in care, while improving precision, accelerating discovery and
reducing disparities. AI can empower patients and potentially allow healthcare
professionals to relate to their patients as healers supported by the combined wisdom
of the best medical research and analytic technology. There are, however, many
challenges to understanding the optimal uses of AI; addressing the technological,
systemic, regulatory and attitudinal roadblocks to successful implementation; and
integrating AI into the fabric of health care. This paper provides a grounding in the
origins and fundamental building blocks of AI, applications in healthcare and for
nursing, and the critical challenges facing implementation in healthcare.

Keywords. Artificial Intelligence (AI), Natural Language Processing (NLP),

artificial neural networks, classifiers, Machine Learning (ML), image analysis,
augmented intelligence, cognitive computing

1. Introduction

There is a great deal of interest, excitement and hype surrounding the potential for
applying artificial intelligence (AI) systems to some of the more complex challenges in
healthcare today. There are equally deep concerns about exactly what that means in
practice in terms of understanding how AI systems work, how best to incorporate them
into medical and nursing practice, how to balance potential benefits and risks, how to
address regulatory and accountability issues, and finally concerns about bias both within
the systems themselves, and in terms of their availability and who will benefit from them.
Our goal is to demystify some of these issues by providing a brief history of AI, define
some of the foundational components of most AI applications, describe what we see as
the value propositions for the use of AI in healthcare and highlight some of the ways it
is currently being used, and finally, consider both the opportunities and the challenges of
implementing AI for healthcare in the future.

1
Corresponding Author, Eileen Koski, M.Phil, FAMIA, IBM TJ Watson Research Center, 1101
Kitchawan Road, Yorktown Heights, New York 10598, USA; E-mail: [email protected].
296 E. Koski and J. Murphy / AI in Healthcare

2. A Brief History of AI

“Artificial Intelligence” refers to a computer system that can learn and make decisions
based on its own accumulated experience, as distinguished from expert, decision support,
or rules-based systems that benefit from the increased speed and precision of computers,
but are based entirely on expert human reasoning [1].
The earliest applications of expert systems in health care were designed to mimic
human reasoning processes, but they were necessarily bound by the limits of established
medical knowledge. Some of the more recent attempts to develop a learning computer in
health care use more cognitive computing strategies based on teaching the computer how
to digest and synthesis both medical literature and medical records in order to improve
the efficiency and accuracy of applying medical knowledge to diagnostic and treatment
challenges, but to go beyond that to a point where the computer can identify patterns and
suggest solutions that may not have been apparent before.
A concrete example of this changed approach, comes from the world of gaming. In
the 1990’s IBM’s Deep Blue computer was programmed with all known strategies for
playing chess, allowing the computer to identify the best response in any given situation,
and beat the world chess champion, Garry Kasparov [2]. In contrast, the recent
development of AlphaGo, was accomplished by teaching it the rules of Go, training it at
a low-level, and then allowing it to play against another computer and learn on its own.
As a result, the computer derived new winning strategies on its own and defeated human
grand masters using strategies that confounded established theories of the game [3,4].
The rules of chess and Go are well known, as are the complex and varied strategies
for playing both games. In applying AI strategies to healthcare, we must remember that
we have not yet uncovered all of the rules that govern how our bodies and minds
function – and malfunction. While we do want to apply what we have learned, we also
hope that AI systems will allow us to derive new insights and knowledge.

3. Foundational Concepts in AI

Core concepts and methodologies that form the basis of AI systems include:
x Natural Language Processing (NLP) - Extracting clinical concepts such as
symptoms, diagnoses and treatments from narrative text, such as clinical notes
x Classifiers - Mapping input data into categories or classes based on training data
in which the proper classification is known, i.e. labelled, so new data can be
correctly categorized [1]
x Artificial Neural Networks - Systems modeled on biological nervous systems
x Machine Learning (ML) - Systems able to process large volumes of data and
extract meaningful information from it to address practical problems [1]
x Augmented Intelligence - Technology intended to assist humans in utilizing or
extending their own capabilities [5]
x Image/Speech Analysis - Extracting meaningful information from images such
as MRIs or recorded speech, as opposed to numeric, categorical or text data
 E. Koski and J. Murphy / AI in Healthcare 297

4. Healthcare Value Proposition

Numerous challenges in healthcare can potentially benefit from AI, and some may render
AI virtually indispensable. While the science and practice of medicine have continually
advanced, the reality is that our health care systems do not function as well as is needed,
populations are not served equitably, the cost of healthcare is spiraling out of control and
yet, by many standards, the health of our population is not what it should be. AI cannot
address all the societal, political and environmental issues at play, however, there are key
ways that AI can contribute to increasing efficiency, raising standards of care, delivering
on the promise of precision medicine and supporting research.
x Information Synthesis. Due to the dramatic increase in the amount of medically
relevant data generated each year, there is simply too much information to be
handled without computational help.
o Patient Data - Test results, diagnoses, treatments and medical histories, as
well as lifestyle data on behaviors such as diet and exercise that can have
medical utility or implications.
o Data Complexity - Data-rich results, such as gene sequencing and MRIs,
represent both a qualitative and quantitative increase in the complexity of
the data. As populations have aged, the number of people with multiple co-
morbidities has also increased, yielding complex treatment regimens.
o Medical Literature - The U.S. National Library of Medicine [6] statistics
show that annual citations increased over 70% in the ten-year period from
2006 (688,444) to 2016 (1,178,360) and now exceed one million per year.
x Augmenting Human Performance. In clinical settings today, it is not possible
for even the most skilled clinicians to successfully digest all available
information, beyond which, not all clinicians have the same level of experience
to inform them in this process. In cases of rare diseases or unusual presentations
of common diseases in particular, a patient’s presentation may fall well outside
of most clinicians’ experience, often leading to delays or errors in diagnosis and
treatment. Treatment decisions are further complicated by the increasing
number of patients with multiple co-morbidities, increasing the likelihood of
interaction effects.

5. Selected Healthcare AI Applications

Selected applications spanning multiple disciplines and applications:

x Clinical Decision Support CDS) systems have been used since the 1970s-1980s
to reduce variation and improve adherence to guidelines.
x Precision/Personalized Medicine aims to use an individual’s genetic make-up
to determine the correct choice and dose of treatment [7].
x Image Analysis has been particularly important in reducing variability in the
interpretation of image data and has recently been shown to be on a par with
expert analysis in such areas as retinal imaging [8] and mammography [9].
x Internet of Things – IoT - the widespread adoption of smart devices, combined
with advances in sensor technologies, has yielded new opportunities to apply
AI not just in traditional medical settings, but wherever a patient may be.
298 E. Koski and J. Murphy / AI in Healthcare

6. Nursing AI Applications

Nurses will benefit from many types of AI applications. For example, clinical decision
support and analytics with AI can help nurses improve quality, safety and reduce costs
as they deliver care. The use of IoT can assist in the assessment and monitoring of
patients remotely, eliminating the need for some home care visits. But there are some
applications unique to nursing as well.
IoT connected to the EHR as well as speech recognition with NLP can ease the
burden of nursing documentation by automatically adding data from medical devices and
voice notes into the record. AI can assist in the organization and prioritization of the
nurses’ workload as their shift begins and adjusting interventions throughout the shift as
new orders and patient needs change. AI applications can serve as “nurse coaches” to
help patients manage a health condition or make behavior changes through the use of
pre-recorded video clips and training materials that are triggered by algorithms as each
patient uniquely works through the virtual session. AI can support nursing care
management applications as well as simulation training of nursing and other healthcare
professionals.
Nurses are uniquely positioned to gain value from the use of AI. However, for
systems to be configured correctly and work properly, nurses need to be involved and
engaged from the outset to ensure such systems are well-engineered and can be trusted.
The future will be informed by data and intelligent technologies that can recommend
action based on information, harnessing the power of AI so nurses can deliver care better,
faster, and more safely. Nursing’s charge will be to continue to integrate the human
aspects of care while automating some of the detection and reasoning processes [10].
First paragraph.

7. AI Implementation Challenges

One of the primary challenges to AI adoption is the availability and quality of data used
to train systems. There may also be limited generalizability of a model due to
characteristics of the population from which it was originally derived.
From a clinical acceptance standpoint, the primary issues are transparency,
explainability, validation, usability and liability. Many AI systems seem like “black
boxes” to clinicians, who may be uncomfortable accepting a recommendation when they
cannot see how it was derived or validated. It will also be necessary to ensure that AI
systems don’t simply add yet another layer of potential alert-fatigue onto clinicians.
Medical and nursing education, which have long stressed knowledge acquisition and
retention, will need to consider how to address knowledge management, interpretation
and appropriate application of AI [11]. The systems that are most likely to succeed, will
be those that can explain the basis for their conclusions and recommendations and can
improve processes that support clinical workflows.
From a societal perspective, issues of privacy loom large, although there is likely to
be a generational shift among younger patients already accustomed to digital life. Equity
and fairness are also important to consider both with respect to the role such systems can
play, as well as how and for whom they will be used.
 E. Koski and J. Murphy / AI in Healthcare 299

8. Discussion/Conclusions

The potential value of AI applications to healthcare have been well documented, and
such systems may become virtually indispensable as ever more precise and detailed data
continues to be amassed about every aspect of health. AI can help to reduce variability,
improve precision, accelerate discovery and reduce disparities. AI can empower patients
and potentially allow nurses and other clinicians to focus more on their patients than their
data. This will allow all healthcare professionals to fully relate to their patients not just
as caring healers, but as healers supported by the combined wisdom of the best of medical
research and analytic technology combined.
The challenges ahead will relate to understanding AI’s optimal uses; addressing the
technological, systemic, regulatory and attitudinal roadblocks to successful
implementation; and finally, appropriately integrating such systems into the fabric of
health care and society. As with all new technologies, an appropriate balance will evolve,
but we will need both visionaries moving us forward, and skeptics asking tough questions,
to assure that the greatest possible benefits of AI in healthcare are achieved.

References

[1] Weiss SM, Kulikowski C. Computer Systems that Learn. San Francisco, Morgan Kaufmann Publishers,
1991.
[2] Weber B. Swift and slashing, computer topples Kasparov. The New York Times May 12, 1997.
https://www.nytimes.com/1997/05/12/nyregion/swift-and-slashing-computer-topples-
kasparov.html?module=inline
[3] Sheldon N. This more powerful version of Alpha Go learns on its own. Wired October 2017.
https://www.wired.com/story/this-more-powerful-version-of-alphago-learns-on-its-own/
[4] Silver D, Schrittwieser J, Simonyan K et al. Mastering the game of Go without human knowledge. Nature
550(7676):354-359, 2017 Oct 18.
[5] Information Week. It’s about augmented intelligence not artificial intelligence. 4/5/2018
https://www.informationweek.com/big-data/ai-machine-learning/its-about-augmented-intelligence-not-
artificial-intelligence/a/d-id/1331460
[6] National Library of Medicine. Yearly citation totals from 2017 MEDLINE/PubMed Baseline.
https://www.nlm.nih.gov/bsd/licensee/2017_stats/2017_Totals.html
[7] Collins FS. The right drug at the right dose for the right person. In The Language of Life, Collins FS,
New York, Harper Collins Publishers 2010, pp. 231-250.
[8] Hwang DK, Hsu CC, Chang KJ et al. Artificial intelligence-based decision-making for age-related
macular degeneration. Theranostics 9(1):232-245, 2019 Jan 1.
[9] Rodriquez-Ruiz A, Lang K, Gubern-Merida A et al. Stand-alone artificial intelligence for breast cancer
detection in mammography: Comparison with 101 radiologists. J Natl Cancer Inst. 2019 Sep 1;11(9):916-
922.
[10] Sensmeier J. Harnessing the power of artificial intelligence. Nrsg Mgt 48(11): 14-19, 2017 Nov.
[11] Wartman SA, Combs CD. Reimagining medical education in the age of AI. AMA J Ethics 21(2):E146-
152, 2019 Feb 1.