What’s needed to build physician trust in health care AI

Kevin B. O'Reilly , Senior News Editor

AI and its potential implications to reshape the world for better—or worse—is the hot topic among thought leaders in every industry and was in the spotlight at CES, the powerful tech event held in Las Vegas each year.

The AMA’s leadership to ensure that health care AI improves patient care in a way that is transparent and trustworthy was reflected in the presence of AMA President Jesse M. Ehrenfeld, MD, MPH, at two CES panel discussions on the issue.

Membership Moves Medicine™

  • Free access to JAMA Network™ and CME
  • Save hundreds on insurance
  • Fight for physicians and patient rights

One panel, on the health care workforce, wellness and AI, was moderated by Acting U.S. Labor Secretary Julie Su.

She asked the panelists about one big worry and one big hope they have about health care AI, which the AMA calls augmented intelligence to emphasize the human element of health care that is so critical in the patient-physician relationship.

“There is a lot of room for hope,” Dr. Ehrenfeld told the crowd at CES. “One area is related to wellness and burnout among the healthcare work force. Many tools show tremendous promise in helping alleviate physician administrative burdens and allowing physicians to get back to what they do best: caring for patients.”

A worry, he added, “is related to bias and trust.”

Dr. Ehrenfeld noted that “we’ve already seen too many examples of systems found to have biased training data or a failure to imagine other design flaws.”

If such problems are “not addressed at the outset, these can lead to systems that invisibly and unintentionally reproduce and normalize the racial and other biases of their training sets,” added Dr. Ehrenfeld, an anesthesiologist who co-chairs the Association for the Advancement of Medical Instrumentation’s AI committee. “That is why we must intentionally apply an equity lens from the beginning development stages.”

Doctors see upsides, flaws in AI

Dr. Ehrenfeld’s remarks reflected the organization’s recent survey (PDF) results finding that nearly two-thirds of the 1,081 physicians responding said they see advantages to using AI. Meanwhile, 41% said they were both equally excited and concerned about potential uses of AI—often called artificial intelligence—in health care.

The AMA is influencing the development of health care AI by developing standards and a common language and has recently released a set of AI principles (PDF).

The survey results “should serve as a bit of a warning to developers and regulators,” said Dr. Ehrenfeld, who also took part in the Health Industry Leaders Forum at CES, an invitation-only, informal meeting of about 50 thought leaders. 

The biggest physician concern is that health care AI, “if not designed and implemented correctly, has enormous potential to negatively impact and in fact undermine the patient-physician relationship.”

What physicians—more than 90% of those surveyed—demand is “clear information about AI to help explain how decisions are made, to demonstrate efficacy, information on intended use and limitations, how bias is managed, and AI’s adherence to standards and performance validation,” he noted.

“At the end of the day, we want safe and reliable products in the marketplace,” Dr. Ehrenfeld said. “That is what will allow companies to earn—and keep—the trust of consumers and physicians.”

Dr. Ehrenfeld also moderated a separate panel discussion on digital health that featured, among others, Maria Ansari, MD, CEO and executive medical director of The Permanente Medical Group Inc., which is a member of the AMA Health System Program that provides enterprise solutions to equip leadership, physicians and care teams with resources to help drive the future of medicine.

Dr. Ansari highlighted her medical group’s recent introduction of ambient AI scribe technology to help enhance the care experience and address physician burnout.

For physicians, residents and medical students to best engage in the development and deployment of AI tools, they need a foundational knowledge base. This can be found on the AMA Ed Hub™ with the “AMA Artificial Intelligence Learning Series: AI in Health Care.”