The practice of medicine is experiencing a profound shift, as technological innovations fundamentally change how patients are diagnosed and treated. At the forefront of these technological breakthroughs is the continued development of artificial intelligence. While the use of artificial intelligence for medical treatment is still in its infancy, this technology has the potential to revolutionize the clinical practice of medicine in coming years. By understanding the various ways that artificial intelligence is being used to improve healthcare, physicians can prepare themselves for its influence on practices of all sizes and specialties.
What is AI?
Commonly known as “AI,” artificial intelligence refers to a computer system or software that is able to perform tasks that typically require human intelligence. A prevalent topic in today’s media, the concept behind artificial intelligence (and its potential to replace humans for various tasks) is familiar to most. However, throughout the medical community, many physicians are approaching the use of AI technology within clinical settings as a way to improve, rather than replace, their ability to diagnose and treat patients. This concept of employing artificial intelligence alongside human expertise is known as augmented intelligence, and it allows physicians to maintain a lead role with machines that further enhance their expertise and ability to diagnose and treat.
Most current healthcare AI tools are either too limited or too early in their development process to be standalone clinical decision-makers, which means that most current AI tools are a form of augmented intelligence. However, as technology advances, this state of affairs could potentially change rapidly, drastically altering the healthcare landscape. At the same time, no matter how advanced the technology becomes, laws and regulations will decide whether AI tools remain in the role of enhancing physician decision-making or gain the ability to operate independently. For example, the FDA must approve medical devices that use artificial intelligence for clinical decision-making before they can be marketed. Similarly, state medical boards and future legislation may be able to regulate the use of devices and AI software that is engaged in the clinical practice of medicine. Given that human beings will make these decisions, it seems likely that physicians will maintain at least some supervisory role over AI devices for the foreseeable future, although the scope of this oversight and control may depend on how quickly the technology advances.
How advanced is AI now?
AI is already affecting many aspects of healthcare delivery, including clinical care, and every week brings news of additional potential uses of AI technology in the medical setting. For example, researchers at Harvard Medical School and the Perelman School of Medicine at the University of Pennsylvania recently developed an algorithm that was able to predict 30-day mortality for patients starting chemotherapy across a wide range of cancers and treatments. Similarly, after completing a successful pilot program, Geisinger Medical Center is implementing an AI tool that can reduce avoidable readmissions related to chronic obstructive pulmonary disease. Mount Sinai Health System has announced a new initiative that uses AI to identify patients at the greatest risk of developing advanced kidney disease, allowing doctors to proactively prevent this outcome through early intervention. The potential value of these AI-driven clinical interventions is also illustrated by a recent study by researchers at Cedars-Sinai Medical Center and Optum Advisory Services, showing that the odds of a patient experiencing a complication increased by 29 percent when physicians did not follow clinical decision support alerts embedded in patient EHRs.
Perhaps the area in which the potential of AI-based technologies is most apparent is in imaging-related fields. Results from a study earlier this year indicate that researchers from Germany, France, and the United States have trained an AI system to diagnose skin cancer more accurately than trained dermatologists. Researchers at New York University have also trained an AI algorithm to diagnose two different types of lung cancer using only images of the tumor.
The FDA has approved the use of an AI-based software program that analyzes x-ray images to detect signs of distal radius (wrist) fracture and highlights these locations to aid physicians with their diagnoses. At the same time, the FDA has also approved software that uses AI to detect diabetic retinopathy based on digital images of a patient’s eyes, enabling physicians to refer the patient to an eye care professional if moderate to severe diabetic retinopathy is detected.
What does this mean for physicians?
While some of these AI-enabled enhancements are already in use, many currently remain in the developmental stage. However, given how quickly technology is evolving, physicians need to start thinking about the potential effect AI may have on their practices. A few key issues to consider include:
- Physician Role and Autonomy. Physicians are currently the ultimate decision-makers on clinical issues, but AI systems are being developed that may offer more accuracy when making diagnoses. As a result, practices need to consider how they will handle a conflict between a physician and the AI system. Should the practice ultimately defer to the physician’s opinion? Will policy require the physician to get a “tie-breaker” opinion from another professional? Physicians’ daily routines may also be affected when AI systems are able to handle standard clinical tasks. For example, as AI systems improve their ability to accurately read images, it may make sense to task more junior radiologists with reviewing images for which the AI system has made a diagnosis with a high degree of certainty, allowing experienced radiologists to review only those images that produced a higher level of uncertainty. Productivity gains from AI-enabled diagnoses may also require practices to determine whether their staffing levels are appropriate.
- Business Decisions. AI-enabled changes in how physicians practice medicine may have a significant effect on practice revenues. For example, if AI software enables a physician to treat more patients (e.g., analyze more images), practices may see an increase in yearly revenues. Conversely, if payers begin factoring AI-enabled productivity gains into per-treatment reimbursement rates, physician practices that do not adopt AI-enabled solutions may see revenues drop substantially. The introduction of AI systems into the clinical workflow also has potential implications for malpractice liability risks, as physicians may face increased liability for treatment decisions made in reliance on recommendations from AI systems. Physicians also may face increased liability risks if they go against an AI system’s recommendation and the patient experiences an adverse outcome. In light of these concerns, practices may need to consider what types of indemnification provisions (and other provisions) need to be included in contracts with manufacturers of AI systems before the adoption of this technology.
- Burnout and Wellness. More than half of U.S. physicians experience some symptoms of burnout. To the extent that the implementation of AI systems has the potential to make physicians feel that they are becoming mere babysitters to machines, the adoption of these technologies may increase physician burnout. However, some believe that if the adoption of AI systems that can handle the more routine parts of a physician’s job allows physicians to focus their time and efforts on the more complex, interesting issues confronting their patients, then implementation of AI systems could potentially reduce levels of burnout.
AI technology is only just beginning to influence clinical practice, and many new developments are poised to leave a lasting impression on the future of medicine. Physicians and their practices should be on the lookout for future developments in this area so that they are best positioned to succeed in an AI-enabled world.
Rediscover the joy of medicine Subscribe
Curi explores the implications of additional stress and anxiety on healthcare workers during the COVID-19 pandemic.
Details on how a physician’s failure to complete colon cancer screening resulted in a patient’s untimely death and key takeaways to avoid a similar outcome.
Curi Capital explores three important factors to consider when coordinating multiple investment relationships.