A recent UC San Diego Health study found that ChatGPT may be better at providing more empathetic answers to patient questions. Using that information as a catalyst, the health system launched a pilot in which Epic and Microsoft's generative AI technology automatically drafts message responses.
"I spoke with one of the leaders at Epic and shared the results of the study and they were really interested," Christopher Longhurst, MD, chief medical officer and chief digital officer at UC San Diego Health, told Becker's. "Since we already had an early experience using ChatGPT, we decided to join the pilot."
UC San Diego Health, Madison Wis.-based UW Health, and Palo Alto, Calif.-based Stanford Health Care are the first health systems to use AI — built from Microsoft and Epic — to help physicians respond to patients' questions in online portals.
The pilot is in its beginning stages and is aimed at reducing the documentation workload on physicians.
"Across the country, doctors are being inundated with messages, and it's a real problem we need to solve," Dr. Longhurst said.
Fifty-seven percent of providers said excessive EHR documentation is one of the contributing factors leading to burnout, but generative AI and ChatGPT are showing promising signs to relieve this problem.
"We have real research that already shows that ChatGPT can be helpful, and it was just released in November," Dr. Longhurst said. "Now we've got a live pilot where it's actually been integrated into clinical workflow and in our electronic health record where care occurs; that's pretty impressive."
While the integration is promising, Dr. Longhurst said it will use the AI cautiously and will have clinicians review all of its responses before they are sent to patients.
"In our pilot, ChatGPT is being used to generate a draft response to a patient inquiry, but a doctor has the choice to use the draft as a starting point, revise it or decide to write their own," Dr. Longhurst said. "We've also got a disclaimer that goes with every message saying, 'This message was generated automatically in a secure environment but was reviewed and edited by your care team.'"
Dr. Longhurst said as patients have been found to have some apprehension toward AI, it is important that UC San Diego Health ensures it is transparent about its usage.
"I see a lot of potential applications for this technology, but for each one of its applications, we're going to be very thoughtful," he said. "We're going to try to study it and pilot it so we determine its benefit and, most importantly, ensure that it's not unintentionally creating harm."
The team at UC San Diego Health is examining the AI for potential for bias and worsening of health inequities, as well as collecting data to see if the tool actually makes clinicians' jobs more efficient and to see if patients find ChatGPT's responses helpful.
Dr. Longhurst said initial feedback from physicians and patients from the pilot has been positive, but it will take two to three more months to assess if the tool can actually deliver on its promises.