UC San Diego health trial suggests promising role of ChatGPT in easing physician workloads
A recent study conducted by UC San Diego Health suggests that ChatGPT, a powerful AI model, may offer more empathetic responses to patients’ queries. This finding has spurred the launch of a pilot program wherein Epic and Microsoft’s generative AI technology autonomously generate responses to messages.
Dr. Christopher Longhurst, Chief Medical Officer and Chief Digital Officer at UC San Diego Health, expressed enthusiasm about the pilot programme after discussing the study results with one of Epic’s leaders. “Given our preliminary experience with ChatGPT, we decided to participate in the pilot,” he said.
The pilot program, initiated by UC San Diego Health, UW Health based in Madison, Wisconsin, and Stanford Health Care located in Palo Alto, California, marks the first instance of health systems using AI, courtesy of Microsoft and Epic, to aid physicians in addressing patients’ questions on online portals.
This nascent project is designed to alleviate the burden of documentation on physicians. Dr. Longhurst expressed concern over the increasing volume of messages doctors have to manage nationwide, underscoring the pressing need to address this issue. Excessive EHR documentation, a factor contributing to physician burnout, is reported by 57% of providers. However, generative AI and ChatGPT are indicating potential to ease this problem. “Our existing research already affirms that ChatGPT can be beneficial,” said Dr. Longhurst. “The recent integration of this AI into our clinical workflow and electronic health record system is noteworthy.”
However, while the AI’s integration is encouraging, Dr. Longhurst added that it would be deployed cautiously with clinicians reviewing all AI-generated responses before they are relayed to patients. “In the pilot, ChatGPT drafts a response to a patient’s query, which a doctor can choose to use as a starting point, modify, or opt to craft their own response,” explained Dr. Longhurst. “Every auto-generated message is accompanied by a disclaimer stating that the message was created in a secure environment and was reviewed and edited by the patient’s care team.”
Given the apprehension patients may harbour towards AI, Dr. Longhurst emphasised the importance of transparency in AI usage by UC San Diego Health. He said, “While the technology holds potential for various applications, we plan to study and pilot each application thoughtfully to ensure that it is beneficial and does not inadvertently cause harm.”
The UC San Diego Health team is scrutinising the AI for potential bias and the risk of exacerbating health inequities. They are also gathering data to determine if the tool enhances clinicians’ efficiency and if patients find ChatGPT’s responses useful. While preliminary feedback from physicians and patients has been favourable, Dr. Longhurst stated that an additional two to three months would be necessary to evaluate whether the tool truly delivers on its promises.