This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Education, is properly cited. The complete bibliographic information, a link to the original publication on http://mededu.jmir.org/, as well as this copyright and license information must be included.
Health care is evolving and with it the need to reform medical education. As the practice of medicine enters the age of artificial intelligence (AI), the use of data to improve clinical decision making will grow, pushing the need for skillful medicine-machine interaction. As the rate of medical knowledge grows, technologies such as AI are needed to enable health care professionals to effectively use this knowledge to practice medicine. Medical professionals need to be adequately trained in this new technology, its advantages to improve cost, quality, and access to health care, and its shortfalls such as transparency and liability. AI needs to be seamlessly integrated across different aspects of the curriculum. In this paper, we have addressed the state of medical education at present and have recommended a framework on how to evolve the medical education curriculum to include AI.
Global health care expenditure has been projected to grow from US $7.7 trillion in 2017 to US $10 trillion in 2022 at a rate of 5.4% [
From clinicians’ perspective there are many important trends that are affecting the way they deliver care of which the growth in medical information is alarming. It took 50 years for medical information to double in 1950. In 1980, it took 7 years. In 2010, it was 3.5 years and is now projected to double in 73 days by 2020 [
Artificial intelligence (AI) is a scientific discipline that focuses on understanding and creating computer algorithms that can perform tasks that are usually characteristics of humans [
Today, AI is being piloted in health care [
The rise of AI in health care and its integration into routine clinical practice is going to be a challenge. Along with changing the conventional ways physician work, the
Researchers at Mount Sinai Hospital have created a deep learning algorithm that was trained on the data of 700,000 patients. This algorithm was able to predict onset of a disease such as schizophrenia with high accuracy [
Finally, many of the AI systems attempt to mimic aspects of human and animal central nervous systems that are, at large, still a black box. In a recent paper, Zador [
The development of AI algorithms almost as a rule requires data from a large number of patients. Google, for example, is using 46 billion data points collected from 216,221 adults’ deidentified data over 11 combined years from 2 hospitals to predict the outcomes of hospitalized patients [
Another unresolved question related to the use of AI in health care is liability for the predictions of an algorithm. It is unclear who is liable when a patient experiences serious harm because of an inaccurate prediction. One could argue for any of the involved parties: the physician, the hospital, the company that developed the software, the person who developed the software, or even the person who delivered the data. Standards for use of AI in health care are still being developed [
As use of AI and proactive use of tools such as chatbots [
With medical information growing at a breakneck speed, physicians are having trouble keeping up. This is leading to information overload and creates pressure to memorize all this content to pass the United States Medical Licensing Examinations (USMLE) to qualify for residency positions. Physicians today are working longer hours and are also expected to deliver coordinated care [
AI could help physicians by amalgamating large amounts of data and complementing their decision-making process to identify diagnosis and recommend treatments. Physicians in turn need the ability to interpret the results and communicate a recommendation to the patient. In addition, AI could have an impact by alleviating the burden from physicians for performing day-to-day tasks [
Physicians go through extensive periods of training before they can eventually register as specialists. Although medicine has seen major changes over the last decades, medical education is still largely based on traditional curricula [
Initiatives for artificial intelligence in medical education.
Institution | Project |
Duke Institute for Health Innovation | Medical students work together with data experts to develop care-enhanced technologies made for physicians |
University of Florida | Radiology residents work with a technology-based company to develop computer-aided detection for mammography |
Carle Illinois College of Medicine | Offers a course by a scientist, clinical scientist, and engineer to learn about new technologies |
Sharon Lund Medical Intelligence and Innovation Institute | Organizes a summer course on all new technologies in health care, open to medical students |
Stanford University Center for Artificial Intelligence in Medicine and Imaging | Involves graduate and postgraduate students in solving heath care problems with the use of machine learning |
University of Virginia Center for Engineering in Medicine | Involves medical students in the engineering labs to create innovative ideas in health care |
Another important technology-related aspect that is often overlooked in medical training is working with electronic health records (EHRs). EHRs have many benefits, such as improved patient safety, but also assist the implementation of AI in health care. AI algorithms use information from the EHR, and therefore, the knowledge on how to input unbiased data into the EHR is essential. Otherwise, the AI algorithm will likely be biased as well [
With the rapid digitization of health care, EHRs facilitate new ways to acquire and process valuable information that can be used to make an informed decision [
Physicians will have a crucial role in deciding which of these tools is best for their patients. In turn, this will likely change the physician-patient relationship [
Future physicians will need a broad range of skills to adequately use AI in clinical practice. Besides understanding the principles of medicine, physicians will also need to acquire satisfactory knowledge of mathematical concepts, AI fundamentals, data science, and corresponding ethical and legal issues. These skills will help them to use data from a broad array of sources, supervise AI tools, and recognize cases where algorithms might not be as accurate as expected [
Some of the time that was originally spent on memorizing medical information will now have to be devoted to other skills. This will have a major impact on the way students and residents will experience their training. The system has to change in such a way that competence will no longer be judged based on factual knowledge but rather on communication skills, emotional intelligence, and knowledge on how to use computers.
With an overfull curriculum, there is limited interest in adopting new topics [
To achieve a change in curriculum, many political and bureaucratic hurdles have to be overcome.
However, one of the most compelling arguments for the implementation of AI training in medical education is that this training will augment existing curriculum rather than replace existing coursework. When students are trained to use AI tools, focus should shift from acquiring basic knowledge on how to use the tool to a basic understanding of the underlying principles. This will enable the students to use this fundamental knowledge when current tools get outdated and new tools are introduced.
Another practical problem is that traditional medical training revolves mainly around the interactions between an attending physician and the residents or medical students. When AI is increasingly introduced into clinical practice, this could be problematic. Many senior physicians have little to no experience with AI. AI training could be delivered via Continuing Medical Education (CME) programs and might need to be also taught by educators from outside the medical community. For example, a 2-credit CME course on AI and the Future of Clinical Practice is delivered by a computational biologist and business economists [
The traditional medical curriculum, which is mostly memorization based, must follow the transition from the information age to the age of AI. Future physicians have to be taught
In the core phase of preclinical didactics, time should be devoted to working with health data curation and quality [
During clinical rotations and residency, focus should shift toward relevant applications of AI in practice. With advancements in digital biomarkers [
List of Continuing Medical Education programs on artificial intelligence in health care.
Program | Faculty; organization | Number of Continuing Medical Education credits |
Artificial Intelligence and the Future of Clinical Practice [ |
Computational biologist, Business economist; Massachusetts Medical Society | 2.0 |
Intro to AI and Machine Learning: Why All the Buzz [ |
Medical Informatics, Radiology; The Radiological Society of North America | 1.0 |
Current Applications and Future of Cardiology [ |
Healthcare Technologists, Bioinformatics, Cardiology; Mayo Clinic | 10.0 |
Artificial Intelligence and Machine Learning: Application in the Care of Children [ |
Pediatric Medicine; University of Pittsburgh School of Medicine | |
Artificial Intelligence in Healthcare: The Hope, The Hype, The Promise, The Peril [ |
Medical Informatics, Business Administration; Stanford University School of Medicine | 6.0 |
AI skills must also be balanced with nonanalytics and person-centered aspects of medicine to develop a more rounded doctor of the future. Other skills such as
To enable clinicians to think innovatively and create technology-enabled care models, multidisciplinary training is needed in implementation science, operations, and clinical informatics. The Stanford medical school has created such a program to train clinician-innovators for the digital future by introducing a human-centered design approach to graduate medical education [
As not all of these interventions can be introduced simultaneously, we suggest a few first steps that will lay the foundation for the upcoming years. We suggest to start off by introducing questions on mathematical concepts into the MCAT similar to the mathematics section in the Graduate Record Examination. High quality Web-based courses on data sciences and AI fundamentals should be freely offered in the core phase of medical education. This might lead to students focusing on applications of these subjects more naturally in following years of training.
For residents and medical students who have already finished this phase of training, courses on the fundamental subjects should be available and mandatory throughout the remaining part of their medical education. For students interested in creating new technology-enabled care models, dedicated training in health care innovation during a gap year during the clinical years or after residency should be encouraged. For attending physicians, introductory courses and refresher courses should also be made available. Extensive training is especially necessary for this group so that they can partly take back the task of educating medical students and residents on these subjects in the future.
Recommendations per stage of medical education.
Medical education stage | Recommendations | Suggested content |
MCATa | Introduce questions on linear algebra (vectors, linear transformations, and matrix, solutions for linear systems), calculus (limits, differential calculus, and integral calculus), probability (joint, conditional, and distribution) |
Education Testing Services’ Graduate Record Examination mathematics test [ |
Medical school—core phase | Working with medical data sets (curation, quality, provenance, integration, and governance), EHRsb, AIc fundamentals, and Ethics and Legal |
Datasets: HealthData [ Public datasets in health care [ University of California San Francisco Data Resources [ AI fundamentals AI 101 course from MITd [ Ethics and Law Teaching AI, Ethics, Law and Policy [ AI Law [ EHR training [ |
Medical school—clinical phase | Familiarize with AI-based clinical applications and expand knowledge beyond basic principles of data and AI |
Clinical utility: Overview of Clinical applications of AI [ AI for Health and Health Care (US Department of Health and Human Services) [ Center for AI in Medicine and Imaging [ AI in Healthcare Accelerated Program [ |
USMLEe | Introduce questions on data sciences, AI, and working with EHRs |
Data science courses [ |
Residents | Detailed knowledge on clinical applications and attend conference in health care AI |
|
Specialist | Stay up-to-date on data/AI through CMEf credits and attend conference in health care AI |
|
aMCAT: Medical College Admission Test.
bEHR: electronic health record.
cAI: artificial intelligence.
dMIT: Massachusetts Institute of Technology
eUSMLE: United States Medical Licensing Examinations
fCME: Continuing Medical Education.
List of artificial intelligence in health care conferences.
Name of conference | Topics |
Ai4 AIa Healthcare Conference [ |
Exploring top use cases of AI and MLb in health care |
AI in Healthcare [ |
Business value outcomes of AI and experience in clinical care and hospital operations |
Machine Learning and AI forum (Healthcare Information and Management Systems Society—HIMSS) [ |
Data, analytics, and real-world applications of ML and AI |
AI in Healthcare @ JP Morgan Healthcare Conference [ |
AI applications—drug discovery, secure data exchange, insurer coordination, medical imaging, risk prediction, at-home patient care, and medical billing |
Radiology in the age of AI [ |
AI in medical imaging |
American Medical Informatics Association Clinical Informatics Conference [ |
AI in medical informatics |
Association for the Advancement of AI [ |
“Increase public understanding of AI, improve the teaching and training of AI practitioners, and provide guidance for research planners and funders concerning the importance and potential of current AI developments and future directions” |
aAI: artificial intelligence.
bML: machine learning.
Physicians and machines working in combination have the greatest potential to improve clinical decision making and patient health outcomes [
AI will enable faster and accurate diagnosis, augment radiology, reduce errors due to human fatigue, decrease medical costs, assist and replace dull, repetitive, and labor-intensive tasks, minimally invasive surgery, and reduce mortality rates.
With the global health care expenditure projected to reach US $10 trillion by 2022, AI has the invaluable potential to advance the quadruple aim in health care—enhance the patient experience, improve population health, reduce costs, and improve the provider experience [
artificial intelligence
Accreditation Council for Graduating Medical Education
American Medical Association
Continuing Medical Education
electronic health record
Medical College Admission Test
natural language processing
United States Medical Licensing Examinations
KP has written this paper as part of his PhD studies. He is a vice president at Roche. There is no conflict of interest with his employment at Roche. None of the rest of the authors declare any conflicts of interest.