I recently attended a conference that previewed a still-embargoed report from the National Academy of Medicine titled Artificial Intelligence in Healthcare: The Hope, The Hype, The Promise, The Peril. Hosted by Stanford University School of Medicine in the heart of Silicon Valley, the conference drew more than 700 attendees from 11 countries, underscoring the fact that artificial intelligence is undeniably part of medicine's future. The question being asked at the event: "What role will AI take?"
This is a question our Academy has been asking, too, as it works to position family medicine to take full advantage of the coming AI revolution in health care. When so many of our colleagues are suffering from burnout -- or are at risk for it -- it's prudent that the AAFP has set its first priority related to AI as the reduction of administrative burden, a driver for projects such as the Academy's work with Suki, a digital assistant and scribe. I would describe these as important defensive plays -- a strategy that protects physicians in our daily practices.
Now, it is time for us family physicians to consider our offense. How do we make AI conform to our vision for health care? Family medicine, as a specialty, must reflect on its leadership role as AI expands its presence in our profession. As will be made clear later this month by the National Academy of Medicine report, AI offers tremendous opportunity for physicians and patients.
AAFP EVP and CEO Doug Henley, M.D., said in September during the 2019 Congress of Delegates that the real challenge is ensuring "that family medicine does not miss the opportunity to embrace this future and move upstream."
Through AI, family medicine has the opportunity to not just survive in the current work environment but to leverage the technology to transform the public's expectations of what primary care can and should deliver for patients.
Emerging technologies offer a preview of how AI could broaden family medicine's scope of practice. Imagine using AI to detect diabetic retinopathy without a referral to an ophthalmologist.
Patients with Barrett's esophagus could be monitored by their family physicians without the expense and risk of upper endoscopy by using clinic-collected esophageal cytosponge samples analyzed with deep learning techniques.
Point-of-care ultrasound may become a more powerful bedside tool when used, for example, to predict which infants with hydronephrosis need surgical referral.
What if family physicians could make lifesaving diagnoses for conditions such as melanoma earlier than ever using a handheld imaging device that performs pathology "staining" on living tissue?
AI represents an opportunity for family physicians to disrupt the current order by redefining primary care with an unprecedented scope. Family physicians already provide one in five outpatient visits and the majority of care to rural and underserved populations, so how will the health system transform when we can do more for our patients than ever before?
For one, it could relieve the supply constraints in our health system that come from lack of access to subspecialists. Increasing health care supply would give patients more choice and reduce wait times and, with incentives properly aligned, this care would be delivered at reduced cost.
I already discussed the Academy's current focus on looking for ways intelligent automation can allow family physicians to offload repetitive and laborious tasks that impede clinical workflow. What family physicians have in AI are the ingredients for the quadruple aim, four goals to optimize the delivery of value-based health care: improved patient experience, lower per-capita health care costs, improved population health and improved work life for physicians. Used correctly, AI just may be the tool family physicians need to fix U.S. health care.
Despite the potential for AI in health care, current artificial intelligence is not that intelligent. It cannot discern clinically relevant data from nonsense, as was seen when an algorithm learned to successfully identify melanoma from dermoscopic images because of the surgical marker surrounding the suspicious lesion. Whatever biases are present in the programming or data that go into AI will surface as biased answers. For example, an algorithm designed to identify patients that would benefit the most from care management programs was systematically biased against black patients because it used health care spending as a proxy measure for health risk.
Mia Keeys, M.A., the AMA's director of health equity policy and advocacy, served as a panelist at the National Academy of Medicine conference. She cited digital redlining -- when algorithms perpetuate inequities in marginalized populations because they do not work equally across all populations -- as a real threat to traditionally underserved populations. This underscores the need for strong physician leadership to ensure both patient safety and progress in health equity.
We physicians must be knowledgeable about any treatment to effectively and safely give it to our patients. We know why beta blockers are important in heart failure, and we must also know how they may hurt our patients who use them. Similarly, AI is a medical intervention like any other, and physicians must know how it works so we can safely help our patients. If family medicine wants to be a leader in stewarding the integration of AI in health care, we must learn how to use AI as a tool.
The question I would like to pose to my fellow family physicians, then, is what priority do we place on training present and future family physicians to become practitioners of artificial intelligence? How we confront this question will help decide how family physicians can fulfill our vision for health care. Answered correctly, I believe the AI revolution will also be the #FMRevolution.