Artificial Intelligence in Military Graduate Medical Education: Trainee Perceptions and Current Use

Interest in artificial intelligence (AI) has surged recently, with healthcare being no exception. As free, widely accessible AI interfaces grow more user friendly, AI's impact on medical education is likely to be significant. Research into graduate medical education (GME) has explored AI utilization, with strengths noted in diagnostic specialties, but few studies included multiple programs and, to our knowledge, none were within the Defense Health Agency. We aimed to assess GME trainees' perceptions and experiences with AI to help inform educational curricula and ensure the readiness of our graduates.

We administered a cross-sectional survey of all San Antonio Uniformed Services Health Education Consortium (SAUSHEC) GME trainees. Survey questions were adapted from previously published AI investigations in literature. Voluntary responses were captured via an electronic quick response ("QR") code that was distributed through program directors and the institution's house staff council. Responses were collected from November through December 2024. Demographic data such as age, branch of service, and GME program specialty were included. Five-point Likert scales, ranging from "Totally Disagree" to "Totally Agree," were used to gauge participant agreement with statements across multiple domains. Descriptive statistics were performed on all results. The Human Research Protections Office found the activity does not meet the definition of research.

Eighty-three GME trainees elected to participate in the survey. The majority of respondents were 26-30 years old (61%), male (64%), and in the Air Force (54%). There was a wide range of postgraduate year (PGY) participation: PGY1 (25%), PGY2 (28%), PGY3 (20%), PGY4 (22%), and PGY5 (5%). Ten respondents reported previous formal AI training. Eighty-four percent of trainees agreed with a need for formal AI training, while 73% disagreed that their current curriculum contained enough. Among all surveyed, 84% agreed that AI will become "essential" to the field of medicine, whereas 12% were "undecided." Perceptions that AI utilization could improve GME training were the overwhelming majority; only 7% responded that it would not improve training and education, and 8% responded it would not improve clinical decision-making. Current utilization of AI in daily work varied, with 30% of respondents reporting use for diagnosis, 27% for treatment, 25% for research, and 28% for teaching. In comparison, only 19% reported using AI for prognosis and 16% for clinical documentation. Interestingly 23% reported using AI for military administrative tasks. More than half of respondents (58%) agreed that AI may impact patient trust in physicians, but only 26% perceived it would make their work less valued. Attending physician role modeling of AI utilization was overall low, ranging from 11% in prognosis to 22% in teaching.

Trainee perceptions regarding AI were overall favorable and indicated a desire for formal training. Importantly, trainees reported utilizing AI in daily practice without training. Implementation of new technologies will require deliberate training to ensure that this emerging human-machine interaction is appropriated skillfully. These results help establish a baseline needs assessment for AI curriculum development at a military GME platform.