Is the medical profession on an inevitable path to obsolescence, similar to the fate of travel agents, typesetters, and cartwrights? Technology has transformed workforce requirements through the ages, reducing the need for manual labor and, more recently, for knowledge workers. Although the practice of medicine is complex and varied, some have predicted that technology will dramatically reduce the need for physicians.
As has been discussed in a series of editorials recently, artificial intelligence is clearly a threat to the profession as it exists today.1–3 Still in its infancy, the field of artificial intelligence is accelerating rapidly, and tasks that seemed impossible a few years ago are addressed now with surprising facility. A particularly promising form of artificial intelligence, convolutional neural networks, can identify tuberculosis infections in chest radiographs with nearly complete accuracy.4 Trained neural networks can diagnose melanoma from images of skin lesions more accurately than dermatologists5 and identify metastatic cells in images of lymph node tissue more accurately than pathologists.6 And, yes, we do have surgical robots, though they are hardly autonomous.
Medicine as Knowledge Management
Since the introduction of the personal computer and the Internet, information technology has altered the practice of medicine at an accelerating pace. Physicians now use Google and other search engines more frequently than textbooks to aid in diagnosis and treatment.7 The breadth of knowledge about medicine and the pace of its development has also grown exponentially, with the time needed to double medical knowledge decreasing from an estimated 3.5 years in 2010 to a predicted 0.2 years by 2020.8 This is further exacerbated by the large array of variables we collect about our patients, with the addition of detailed genetic information making it clear that our intellects can no longer handle the array and complexity of important data. Knowledge is much more plentiful and easier for us to access than ever before, but it is also impossible for the unaided physician to retain and process.
Some resist the intrusion of information technology into the practice of medicine, which is hardly surprising when the largest investment in health technology, the electronic medical record, is frequently cited by physicians as their greatest source of job dissatisfaction and as obstructing high-quality care.9 Nonetheless, information technology is already improving patient outcomes in many areas and will ultimately prevail. The next 20 years are likely to see further acceleration in the capabilities of computers to analyze complex data and mimic human cognition. Although the types of activities handled by computers are currently limited to targeted areas of knowledge access and constrained analytical tasks like diagnosing individual conditions, the array of activities handled by computers will progressively increase, dramatically changing the role of the physician.
Medicine as the Art of Caring
The nonanalytical, humanistic aspects of our profession—most importantly, the art of caring—will be much more difficult to replace with technology. Our profession has existed for more than 2,000 years, well before we had reliable knowledge of physiology and disease. In those days, interventions often did more harm than good. In spite of this, physicians have been called to the bedside and revered by societies for centuries. The art of caring has always been central to the profession. Even today, when patients select and recommend physicians, they rely heavily on bedside manner and trust rather than on measures of patient outcomes, which are generally unavailable to them.
Studies have shown that the skills of caring are also associated with improved patient outcomes. In a trial of 262 adults with irritable bowel syndrome all receiving sham acupuncture, a patient–practitioner relationship augmented by warmth, attention, and confidence was found to be superior in improving outcomes compared with a typical patient–practitioner relationship.10 Furthermore, some practitioners were particularly effective at producing benefit through a caring approach. Clearly, how we care matters to patients and is important to their outcomes. The skills associated with effective caring—an array of communication skills—can be identified through validated measures. They can also be taught.
Medical schools have generally underemphasized the nonanalytical aspects of the profession. Still adhering to Abraham Flexner’s directive to devote two years of study to the basic sciences and two years to apprenticeship on clinical rotations, most medical schools allocate substantial time in the curriculum to memorization and analysis, tasks that will become less demanding as artificial intelligence improves. The art of caring with all its components—such as communication, empathy, shared decision making, leadership, and team building—is usually a minor part of the medical school curriculum, sometimes relegated to clinical rotations or a short session on medical humanities.
Some medical schools have already shortened the amount of time spent studying basic science, and some have deemphasized memorization, though the board examinations required to become a doctor continue to apply pressure in the opposite direction. Reducing time spent on memorization of facts could allow for greater training in the art of caring, particularly in the aspects of effective communication. However, the “hidden curriculum” introduced during clinical rotations is likely to encourage the current status quo, a method of training that is too variable to serve as a standard for future physicians.
Artificial intelligence is also unlikely to alter the growing need to address system problems through team-based approaches. Effective leadership and creativity are distant aspirations for artificial intelligence but are growing needs in a system of care that is ever more complex. Future physicians should be prepared to take responsibility for the systems of care that are critical to patient outcomes, leading the changes necessary to provide the best possible care.
As several other medical schools have also done, at the Dell Medical School at the University of Texas at Austin, we have reduced the duration of basic science instruction to 12 months and have emphasized group problem solving while deemphasizing memorization. This has freed up additional time for instruction in the art of caring, leadership, and creativity. We have emphasized communication skills throughout the curriculum and, in partnership with the Moody College of Communication, have created the Center for Health Communication, which advances the underlying science of evidence-based health communication. Human-centered design and leadership skills are taught through didactics and specific projects targeting health system issues, and there is a nine-month innovation and leadership block centered around team projects designed to produce measurable improvements in patient and population health through system changes in hospitals, clinics, or directly in the community.
As machines gain preeminence in the retention, access, and analysis of knowledge, it has never been as important for physicians to recognize the particularly human aspects of the profession encapsulated in the art of caring. Listening, tone, touch, and counsel are critical components of medicine and always have been. Furthermore, physician leadership and creativity are required to address recalcitrant system problems. Our educational systems should rebalance the curriculum toward these components and create opportunities for practicing physicians to recalibrate and prepare for the future. Studies have suggested that physician job satisfaction is improved with more opportunities for meaningful patient communications and for the delivery of higher-quality care.9 Therefore, one side effect of a rebalancing hastened by artificial intelligence may be a reintroduction of additional meaning and joy into our professional lives.
1. Jha S, Topol EJ. Adapting to artificial intelligence: Radiologists and pathologists as information specialists. JAMA. 2016;316:2353–2354.
2. Obermeyer Z, Emanuel EJ. Predicting the future—Big data, machine learning, and clinical medicine. N Engl J Med. 2016;375:1216–1219.
3. Beam AL, Kohane IS. Translating artificial intelligence into clinical care. JAMA. 2016;316:2368–2369.
4. Lakhani P, Sundaram B. Deep learning at chest radiography: Automated classification of pulmonary tuberculosis by using convolutional neural networks. Radiology. 2017;284:574–582.
5. Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542:115–118.
6. Liu Y, Gadepalli K, Norouzi M, et al. Detecting cancer metastases on gigapixel pathology images. arXiv. 2017;1703.02442. https://arxiv.org/abs/1703.02442
. Accessed January 23, 2018.
8. Densen P. Challenges and opportunities facing medical education. Trans Am Clin Climatol Assoc. 2011;122:48–58.
10. Kaptchuk TJ, Kelley JM, Conboy LA, et al. Components of placebo effect: Randomised controlled trial in patients with irritable bowel syndrome. BMJ. 2008;336:999–1003.