Remote patient monitoring (RPM) generates huge amounts of data. With the assistance of artificial intelligence (AI) and machine learning, a major focus of ongoing research is identifying how to best leverage this data in ways that improve clinical care. First, however, RPM technology has to make it into the hands of patients with diabetes to actually have an impact on population outcomes, said Mindy Lee, MD, PhD, Instructor of Pediatrics-Endocrinology and Diabetes at Stanford University.

Dr. Lee and other experts will review recent developments in medical technology and AI during the panel discussion, Clinical Applications of Artificial Intelligence and Diabetes Technology, on Saturday, June 21, from 8:00–9:30 a.m., in Room W183 BC of the McCormick Place Convention Center. On-demand access to recorded presentations will be available to registered participants following the conclusion of the 85^th Scientific Sessions, from June 25–August 25.

Dr. Lee will discuss an initiative at Stanford to bring care to patients through RPM.

“Our Stanford 4T Program is a research protocol where we offered continuous glucose monitors (CGMs) early, during the first month of diagnosis, for kids with type 1 diabetes and supported them with the RPM protocol,” Dr. Lee explained. “We have translated the protocol to standard of care for all the kids with type 1 diabetes who come to our clinic.”

“Medicine is very high stakes, so naturally, leaders have an abundance of caution.”

– Yaa Kumah-Crystal, MD, MPH

The 4T Program shows that RPM is feasible in real-world clinical care, that it is billable, and the costs associated with implementation in terms of manpower and infrastructure can be recouped, she said.

Medicine takes lessons from other fields, Dr. Lee noted, and as such can be a couple of steps behind.

“That’s fine because we are working with such sensitive data,” she continued. “But technology moves at a rapid pace, and we need to find ways to bridge that translation gap so we can use these advancements effectively to improve outcomes.”

The role of generative AI in caring for patients with diabetes is not yet clear, and most clinicians have yet to incorporate it into their routine. There are concerns over accuracy and even the provenance of the data used to train the models.

“Medicine is very high stakes, so naturally, leaders have an abundance of caution,” said Yaa Kumah-Crystal, MD, MPH, Associate Professor of Biomedical Informatics and of Pediatric Endocrinology at Vanderbilt University Medical Center. “Some researchers are doing small pilots to test how these tools can be used. Openly available AI platforms, such as ChatGPT, are not Health Insurance Portability and Accountability Act (HIPAA) compliant.”

This makes it challenging to interface with the tools for medical scenarios, as the user has to strip out personal health information, limiting its use. Dr. Kumah-Crystal, who will address some of the opportunities and pitfalls of using AI in clinical care, encourages health care organizations to implement internal HIPAA-compliant AI tools for providers to explore. 

“I encourage using hypothetical cases and testing different platforms to see how well the technology is able to respond,” she said. “This way, users can get a sense for how the technology adapts to various medical situations without compromising patient privacy.”

She has a three-times-a-day rule she applies to the use of generative AI.

“Anyone will have at least three questions that they’re asking themselves throughout a given day,” she explained. “I suggest throwing those questions into a large language model and compare the thought process and solution. That provides a good sense of how they work, what their strengths and limitations are, and how they might be useful.”

Dr. Kumah-Crystal is seeking to understand the edge of generative AI’s possibilities and where it falls short.

“I’m fortunate to work with a HIPAA-compliant version, so I can use true clinical information to compare to my own thought process,” she said. “I can see where its strengths are and where we need to find a way to beef up its knowledge.”

The panel also includes Marc Breton, PhD, Associate Professor of Research at the Center for Diabetes Technology, University of Virginia. He will discuss the future of automated insulin delivery systems.