Expert panel discusses how AI can transform medicine

Experts from the Kaiser Permanente Division of Research, Mayo Clinic Platform, Veterans Health Administration, and Community Health Systems convened recently to share perspectives on artificial intelligence (AI) in medicine — specifically, how AI can transform medicine and contribute to innovation in health care.

The expert panel discussion, published in The Permanente Journal, touched on topics ranging from President Biden’s October 2023 executive order to the accuracy of generative AI. The group also shared impressions and early experiences with AI garnered from within the panelists’ respective organizations.

How AI can improve care safely and effectively

John Halamka, MD, president of the Mayo Clinic Platform, kicked off the conversation with a nod to the ubiquity of AI while also acknowledging the many “unanswered questions” about the benefits and harms of applying AI within the health care field. Halamka, who oversees Mayo’s AI initiatives, suggested that public–private collaborations will underpin the safe and effective rollout of AI as a means of transforming medicine.

Susan Kirsh, MD, who had direct experience regarding Biden’s executive order, shared her thoughts about trustworthiness in AI. “We all want AI to be safe and effective,” said Kirsh, who is the deputy undersecretary for health discovery, education, and affiliated networks for the Veterans Health Administration. She added that establishing policy and processes to support trustworthy AI and operationalizing workforce expectations will be early hurdles within the broader effort of AI as a component of health care innovation.

Related AI health care story: Navigating AI’s impact on health care

How generative AI can reduce physician workload  

Dr. Halamka also posed questions regarding the accuracy and quality of generative AI. Vincent Liu, MD, of the Kaiser Permanente Division of Research in Northern California, shared early results of ambient AI technology that helps capture notes from patient visits. Along with his colleagues, Dr. Liu previously published findings on AI-based ambient listening that showed promise in reducing physician administrative tasks. “We used an instrument designed to evaluate the quality of human medical scribe documentation to evaluate the quality of the notes taken by the ambient AI scribe tool,” Dr. Liu said. He went on to detail the sampling surveys gathered from patients and physicians and the amount of time users spent in the electronic health record to show that results were favorable. In case of omissions and inconsistencies introduced in the documentation, Dr. Liu noted that “those could be guarded against by ensuring that the physicians were still examining every output.”  

Related AI health care podcast: The promise of AI in health care 

Considerations for implementing AI at scale  

Beyond ethics and regulations, the rollout of AI also comes with practical business implications. Lynn Simon, MD, MBA, president of Healthcare Innovation and chief medical officer of Community Health Systems, spoke to the challenges of organizational adoption of AI tools. “I think that approaching it organizationally requires taking our most important business objectives and outcomes over 1 to 3 years and determining what tools can help us reach our goals,” Dr. Simon said. “That being said…it is wise to begin with the tools that are more likely to produce the biggest benefit within the lowest areas of risk.”  

Dr. Halamka spoke to the importance of collaboration across sectors as a means of ensuring prudent and appropriate expansion of AI in health care. “All of us need to share with each other our early experiences. We should discuss what worked and what didn’t work,” he said, referring to academic, government, and industry organizations.

In discussing the “hype versus reality” of AI-based innovation in clinical medicine, the panel debated whether the technology could replace the function of physicians, nurses, and other clinicians. The ultimate consensus was that AI cannot offer a suitable replacement; rather, it offers the promise of supporting workforces by augmenting and enhancing existing workflows.

Read the full and lively expert panel discussion in The Permanente Journal.