Introduction:
Artificial
intelligence (AI) in
healthcare is the use of complex
algorithms and software to
estimate human cognition in the analysis of complicated medical data.
Specifically, AI is the ability for computer algorithms to approximate
conclusions without direct human input. (Russell S and Norvig,
2020; Esteva, 2019) Simply, it is the ability of
computer systems to perform tasks that would usually require human levels of
intelligence. (Shameer et al., 2018) A subfield of AI
is machine learning, which can be used to teach a computer to analyze a vast
amount of data in a rapid, accurate, and efficient manner through the use of complex computing and statistical algorithms. (Legg
and Hutter, 2007)
What distinguishes AI technology from traditional
technologies in health care is the ability to gain information, process it and
give a well-defined output to the end-user. (Bakkar
et al., 2018; Trivedi
et al., 2018) The primary aim of health-related AI
applications is to analyze relationships between prevention or treatment
techniques and patient outcomes. AI programs have been developed and applied to
practices such as diagnostic processes, treatment protocol development, drug
development, personalized
medicine, and patient monitoring
and care. Additionally, hospitals are looking to AI solutions to increase cost
saving, improve patient satisfaction, and satisfy their staffing and workforce
needs. (Kent, 2018)
Despite the early success of AI in performing
and assisting with clinical tasks, there is still skepticism about the
potential of this technology. Although the scientific literature supporting AI
in clinical medicine is accumulating quickly, few previous studies have
described how physicians perceive the advent of this technology (Sarwar et
al., 2019) It is expected that AI will be used extensively in the medical
field in the future leading to fundamental changes in the role of physicians
and the way they practice medicine. The use of AI for the development of
decision support systems is an emerging area that may help to refine
algorithms, but more research is needed on acceptability, feasibility and
ethics. (Jha, 2016) However, there is no research, to our knowledge, on the
opinions and perception of the application of AI programs in the medical field
among Kuwaiti physicians. The purpose of this study is to highlight primary health
care physicians’ perception of AI.
SUBJECTS AND METHODS:
Primary health care (PHC) in Kuwait is
provided through 104 centers distributed within 5 administrative health regions
corresponding to the number of populations in each region. Two health regions
were selected randomly to conduct this study (Capital and Hawalli).
All health centers in the selected health regions participated in the study (41
centers). The studied population consisted of all physicians working in these
centers (531 during the period of study (1 August to 30 September 2019)
A
cross-sectional multicenter approach was
used in this survey. A predesigned self-administered questionnaire was used to collect data
from the physicians. The questionnaire was derived from other published studies
dealing with the same topic as well as from our own experience. It was sent out to all physicians working in
all PHC centers in the selected health regions who were asked to participate in
the study. Beside participants’ personal data, the questionnaire included four
sections aiming to evaluate their perception of AI. The first section of the survey enquired about
physician’s attitude toward the medical application of AI. The second included
the medical fields in which AI could be applied.
The third one entailed the possible problems regarding the application of AI in
medicine and the fourth section asked the physicians if they believe that AI
would replace physicians in the various tasks of PHC. Participants who believed
that replacement of a given task was likely were asked to give an estimate of
the timescale required for replacement. Participants were asked to respond to
every question to complete the survey.
All the necessary approvals for carrying out
the research were obtained. The Ethical Committee of the Kuwaiti Ministry of
Health approved the research. A pilot
study was carried out on a small sample of physicians to test the clarity,
applicability and suitability of the used questionnaire, and to identify the
difficulties that could be faced during the application. The Statistical
Package for Social Sciences (SPSS-27) was used for data processing. Simple
descriptive statistics were used (mean ± standard deviation for quantitative
variables, and frequency with percentage distribution for categorized
variables).
RESULTS:
Within 531
physicians working in the selected settings and invited to participate in the
study, 369 responded with a response rate of 69.5%. Twelve questionnaires were
excluded as they were not complete. The final analysis included 357 participants.
General
characteristics of the participating physicians:
Within the
participants, 37.3% were males, 42.9% were < 35 years old (mean = 42.4 ±
9.3, Min – Max: 27.0 – 69.0), 57.1% were Kuwaitis, 42.3% had a Bachelor degree.
General practitioners (GP), family physicians (FP), diabetologist, maternity
physicians and dentist constituted 47.9%, 25.8%, 3.9%, 5.0%, and 17.4 of the
studied population. The years of working experience ranged from 1 to 46 with a
median of 9 years. Half of participants worked for 30-39 hours /week with a
median of 36 hours. The number of patients seen / day ranged from 12 to 100
with a median of 25 patients, whereas 38.7% of participants examined ≥40
patients / day.
DISCUSSION:
While recent studies have shown that AI holds
the potential to improve performance with regard to health care, the pathway
toward widespread adoption of new technologies is complex and physician’
perceptions are likely to play a significant role in the pace of technology
adoption in clinical practice. The results of the present study suggest that
the perception of AI by PHC doctors was low whereas less than half of
participating physicians (46.2%) considered themselves familiar with AI.
However, 71.7% believed that AI had useful applications in the health care
field. These results are comparable
with a similar study that was conducted in Korea where the authors reported
that familiarity with AI was very low, only 6.0% believed that they were
familiar with AI in a good degree and many participants answered that AI is useful
in health care and medical field (73.4%). (Oh, 2019)
As regards the advantage of artificial
intelligence, 71.7% agreed that it speeds
up processes in health care, 59.1%
agreed that it helps reducing medical errors, 51.3% agreed that it can support very fast plenty of relevant
high-quality information. However, only 43.4%
and 20.4% support the idea that AI is protective against emotional stress
and space-time constraint respectively. In another study, the physicians agreed
that the advantages of AI include supporting good relevant data in short time (62.3%),
rapid acceleration of processes in health care (19.1%), and lower down the
number of medical errors (9.6%). (Oh et al., 2019)
Only 11.8% stated
that the diagnostic precision of AI is superior to the clinical opinion of a
human doctor. A higher proportion was reported in a previous study whereas 44% of participants agreed that “AI is superior to a
doctor’s experience”. (Oh et al., 2019) On the other hand, 27.2% of physicians agreed that, in the future, they will
always use AI for their final medical decisions, a figure that was relatively
low when compared with similar study as 42% of participants stated that to
reach a medical decision AI should be used. (Oh et al., 2019). When physicians were asked that in case of
difference between AI judgment and doctor’s opinion during medical practice,
85.2% chose doctor’ opinion, and only 3.9% chose AI opinion. This goes with a
previous study as 79% of participants
would follow the doctor’s opinion. (Oh et al., 2019)
Only 5.0% of
participants in the present study agreed that AI could fully replace their
jobs. The proportion of physicians who believed that AI would replace
physicians showed variations in literatures. It may be as high as in a Korean
study where 35.4% of physicians reported that AI could
replace a doctor, (Oh et al., 2019) or very low as in Doraiswamy et al (2020) study who
reported that though half of participants believed that AI would substantially
change physicians’ jobs, only 3.8% felt that it would make their jobs no longer
in use. In accordance with our results regarding difficulties in replacing
doctors. Krittanawong (2018) stated
that AI cannot replace doctors as it cannot conduct specific conversation with
patients to reach a point of trust. Also, in spite of the valuable data given
by AI, still interpretation of data, medical examination and history should be
conducted by physicians. (Krittanawong,
2018; Inkster et al., 2018)
In another study, only 3.8% of physicians
believed that future AI would make their jobs out of date and the majority
(83%) of respondents were sure that AI would never have the ability to provide
empathic care as well as or better than the average. Another 47% predicted that
their jobs would be moderately changed by AI over the next 25 years. (Doraiswamy et al., 2020)
When physicians were asked about the most useful expected application of
AI in the field of medicine, came at the top of the list biopharmaceutical research and development (30.5%) followed by diagnosis (24.1%)
and medical assistance in areas with less services. (18.5%) Less frequent
applications were mentioned by physicians as direct treatment (including
surgery) (11.2%), and making treatment decisions. These results are relatively
lower than that obtained in other study whereas respondents felt that in
medical field, AI would be most useful in the area of diagnosis (83.4%) as well
as formulation of the treatment plan (53.8%). (Oh et al., 2019) According to
Doraiswamy et al. (2020), most participants considered
it unlikely that AI would fully replace doctors in the field of psychology
particularly getting medical history during interview (58%), determination of
the department to which the patient should be referred (55%), planning of personalized
treatment plans for each patient (53%). On the other side, most of them (83%) believed
that the task of documentation will be likely replaced by AI.
In this study physicians believed that commercialization of AI will be through university
hospitals (33.15) followed by governmental primary centers (32.5%) which goes
in accordance this Oh et al. (2019)
study who reported that most participants
(66.2%) thought also university hospital would be the first.
In the present
study, physicians ‘opinion regarding the potential risks of AI when applied in
medicine revealed that the most frequent answer was inflexibility to be applied
to every patient (65.3%). This was followed by inability to give definite
decision in case of inadequate information
(59.1%), neglecting patient’ emotional well-being (55.2%) and difficulty of
application to controversial subjects (48.7%). At the bottom of the list came when
AI is developed by a person with little clinical experience in medicine or
health (24.6%). Comparable with our results, Oh et al. (2019) reported
that the possible risks with AI are that AI is deficient in providing a
decision in case of inadequate information (29.3%) and that it would not be applicable
to every patient (34.1%). Precision medicine is “an emerging approach for
disease treatment and prevention that takes into account individual variability
in genes, environment, and lifestyle for each person” With this approach physicians
are allowed to decide treatment plan and prevention strategies for patients.
(James et al., 2017) The mental status examination, evaluation of
dangerous behavior and formulation of a personalized treatment plan, were also
felt to be tasks that a future AI technology would be unlikely to perform as
well. This could be as doctors may be overvaluing their skills and/or
underestimating the rapid pace of progress in intelligent technologies. (Topol, 2019)
Blease et al. (2018) reported that the most GPs were not sure that AI
technology could perform physicians’ tasks as well as or better than humans.
However, they think that technology would be able to replace physicians administrative
work related to documentation.
The present study revealed that the proportion
of physicians believed that it is likely that technology will be able to
replace human physicians were 79.5% regarding providing documentation about
patient, 55.5%
regarding analyzing patient information to establish
prognoses, just less than half of participants regards evaluation of time of
referral of patients to other health professionals, analysis of patient
information to reach diagnoses. The least proportions were encountered
regarding formulating personalized treatment plans for patients (36.9%) and
providing empathetic care to patients (19.6%) Most of participants in the present study who believed
that AI is likely to replace physicians in certain tasks, indicated that this would
happen within ten years. Our results go in accordance with a previous study
conducted by Blease et al., (2018), who
found that most GPs believed it unlikely that technology will ever be able to
fully replace physicians when it comes to diagnosing patients (68%), referring
patients to other specialists (61%), formulating personalized treatment plans (61%),
and delivering empathic care (94%). On the other hand, 53% considered it likely
that technology will be fully capable of replacing physicians in prognostic
process and 80% believed it likely to fully replace humans to undertake
documentation. Blease et al., (2018) In
another previous study, most physicians expected that AI would be
helpful with diagnoses and in planning treatment by providing the latest
clinically relevant data. However, doctors were skeptical that future
technologies could perform most tasks as well as or better than human doctors.. This is may be due to the fact that doctors may not feel
confident in their knowledge of AI and may find it difficult to separate
marketing hype from ground truth. (Doraiswamy
et al., 2020)
According to some AI experts, machine level intelligence
will be capable of replacing the basic functions of medical professionals by
2050. (Walsh, 2017) Some physicians working in biomedical informatics
and related fields predicted that developments in AI are poised to
revolutionize the delivery of health care, (Hinton, 2018; Naylor, 2018;
Mandl and Bourgeois, 2017) with some suggesting advancements may eventually
obviate the need for physicians altogether. (Obermeyer and Lee, 2017) Physicians
may have multiple, possibly conflicting, beliefs about the impact of technology
in primary care and research is needed to explore these issues further. (Blease et al., (2018)
Some technologists argue that advancements in
AI may one day obviate the need of physicians altogether. (Darcy et al.,
2017; Naylor, 2018) Others forecasting that the role of doctors can never
be fully replaced, and that the future of medicine will likely become a team
between humans and machines. (Verghese et al.,
2018; Obermeyer and Lee, 2017) This view is also shared by authors of a
more recent working paper from the Office of Economic Cooperation and
Development which looked at skills data from over 30 countries. (Nedelkoska and Quintini, 2018)
Our results together with that in literatures provide powerful insights into
how AI could be optimally deployed to work with physicians, rather than replace
them, to enhance health care.
Our study had a limitation regarding
generalization of the results as our sample of PHC physicians may not be
representative of all Kuwaiti physicians and participation was voluntary. Another
limitation is the self-reporting nature of the study with the risk of recall
bias.
References:
Bakkar N, Kovalik
T, Lorenzini I, Spangler S, Lacoste A, Sponaugle K, Ferrante P, Argentinis F, Sattler R, Bowser R. (2018). Artificial
intelligence in neurodegenerative disease research: use of IBM Watson to
identify additional RNA-binding proteins altered in amyotrophic lateral sclerosis.
Acta Neuropathol; 135(2)
:227-47.
Blease C, Bernstein MH, Gaab J, Kaptchuk YJ, Kossowsky J, Mandl KD, Davis RB. DesRoches CM.
(2018). Computerization and the future of primary care: A survey of general
practitioners in the UK. PLOS ONE | https://doi.org/10.1371/journal.pone.0207418
Darcy AM, Louie AK, Roberts LW. (2016). Machine learning
and the profession of medicine. JAMA, 315(6), 551–2.
Doraiswamy M, Blease
C, Bodner K. (2020). Artificial intelligence and the future of psychiatry:
Insights from a global physician survey (2020). Artificial Intelligence in
Medicine; 102. https://doi.org/10.1016/j.artmed.2019.101753
Esteva A, Robicquet
A, Ramsundar B, Kuleshov V,
DePristo M, Chou K, Cui C, Corrado
G, Thrun S, Dean J. (2019). A guide to deep learning
in healthcare. Nat Med;25(1): 4-29.
Hinton G. (2018).
Deep learning–A technology with the potential to transform health care. Jama. ; 320(11): 1101–2.
Inkster B, Sarda S, Subramanian
V. (2018). An empathy-driven, conversational artificial intelligence agent (Wysa) for digital mental well-being: real-world data
evaluation mixed-methods study. JMIR Mhealth Uhealth;6(11):e12106. [doi:
10.2196/12106]
James R. National Institute of on Minority Health and
Health Disparities. (2017). National Institutes of Health Precision medicine
initiative URL: https://www.nimhd.nih.gov/about/legislative-info/clips/pmi.html
[accessed 2019-02-05] [WebCite Cache ID 75yNuNGg5]
Jha S (2016). Will computer
replace radiologist? https://www.medscape.com/viewarticle/863127
Kent J ( (2018). Providers Embrace Predictive Analytics for Clinical,
Financial Benefits". Health IT Analytics. https://healthitanalytics.com/news/providers-embrace-predictive-analytics-for-clinical-financial-benefits
Krittanawong C. (2018).. The rise of
artificial intelligence and the uncertain future for physicians. Eur J Intern
Med; 48: e13-e14. [doi: 10.1016/j.ejim.2017.06.017]
[Medline: 28651747]
Legg S, Hutter
M. (2007). A Collection of Definitions of Intelligence (Technical report). IDSIA.
https://arXiv:0706.3639. Bibcode:2007arXiv0706.3639L. 07-07.
Mandl KD, Bourgeois FT. (2017). The Evolution of patient
diagnosis: From art to digital data-driven science. Jama; 318(19): 1859–60
Naylor CD. (2018). On the prospects for a deep learning health
care system. JAMA, 320(11): 1099-100.
Nedelkoska L, Quintini G. (2018).
Automation, skills use and training. OECD Social, Employment and Migration
Working Papers, 2018. OECD Publishing, Paris, 202, 1- 124.
Obermeyer Z., Lee T. H. (2017). Lost in thought – the
limits of the human mind and the future of medicine. New England Journal of
Medicine, 377(13), 1209-1211.
Oh S, Kim JH, Choi S, Lee HJ, Hong J, Kwon SH. (2019).
Physician Confidence in Artificial Intelligence: An Online Mobile Survey. J Med
Internet Res; 21(3): e12422) doi:10.2196/12422
Russell S, Norvig P. (2020). Artificial Intelligence: A Modern
Approach (4th ed.). http://aima.cs.berkeley.edu/.
Sarwar S, Dent A, Faust K, Richer M, Djuric U, Van
Ommeren R, Diamandis P. (2019). Physician perspectives on integration of
artificial intelligence into diagnostic pathology. NPG Digit Med; 2: 28 https://www.nature.com/articles/s41746-019-0106-0
Shameer K, Johnson KW, Glicksberg BS,
Dudley JT, Sengupta PP. (2018). Machine learning in cardiovascular medicine:
are we there yet? Heart; 104(14): 1156-64.
Topol, E. (2019). Deep medicine: how artificial intelligence
can make healthcare human again. Basic Books. Hachette UK
Trivedi H, Mesterhazy J,
Laguna B, Vu T, Sohn JH. (2018). Automatic determination of the need for
intravenous contrast in musculoskeletal MRI examinations using IBM Watson's
natural language processing algorithm. J Digit Imaging;31(2):245-251.
Verghese A, Shah NH, Harrington RA. (2018). What
this computer needs is a Physician: Humanism and AI. JAMA, 2, 319-21.
Walsh T. (2017). Android dreams: the past, present, and
future of artificial intelligence. UK: Hurst.
