Large language models in internal medicine residency: current use and attitudes among internal medicine residents

Abstract

Abstract Background Artificial intelligence (AI) uses in healthcare changed dramatically with large language models. While work is underway to integrate generative AI safely and effectively into medical practice, little is known about how resident physicians are already using these tools in their work. Objectives To characterize the current usage patterns and attitudes towards large language models among internal medicine residents in North Carolina. Methods Cross-sectional survey-based study of Internal medicine residents in North Carolina. Results 152 of 523 total residents completed the survey (29% response rate). Many identified a role for LLMs in both clinical (56%) and non-clinical medicine (72%) today, although only 26% reported current use. More perceived a future role in clinical (71%) and non-clinical (80%) medicine. None had received formal training, but 94% expressed interest in learning more and 87% thought residencies maybe or definitely should have formal curriculum. Top clinical uses include forming a differential diagnosis, researching treatments, and creating educational materials. Top non-clinical uses included self-directed learning, scientific writing, and summarizing literature. Conclusions Medical educators and administrators should be aware that residents are using generative AI in professional settings. Residents believe in a current and future role for AI in medicine and desire curriculum and learning opportunities during training.