Emergency Medicine Simulation Training Through Generative Artificial Intelligence: A Pilot in Resident Education

Experiential learning is important in emergency medicine, yet high-quality in-person simulation remains limited by cost, faculty time, physical infrastructure, and limited scalability across institutions. Advances in artificial intelligence and large language models offer opportunities to deliver scalable, adaptive, learner-centered education. We developed an AI-driven simulation platform that provides dynamic emergency medicine scenarios, enabling flexible, low-resource practice of clinical reasoning, decision-making, and workflow skills.

The purpose of this innovation is to leverage artificial intelligence and large language models to expand access to high-quality emergency medicine simulation while preserving humanistic elements of care. The platform provides adaptive scenarios with real-time feedback on clinical decision-making and physician–patient communication. By integrating clinical reasoning with interpersonal skills training, this innovation promotes early exposure to emergency medicine, supports competency-based education, and could enhance learner engagement across diverse training environments.

AI patients and a virtual simulation platform were developed by a multidisciplinary team including emergency medicine physicians, basic scientists, computer science engineers, and creative writers. An initial set of eight AI-driven cases addressed cardinal emergency department presentations, including chest pain, abdominal pain, shortness of breath, and weakness. These simulations were implemented for emergency medicine PGY-1 residents during residency orientation in place of traditional in-person simulation. During the trial phase, residents completed one case weekly.

PGY-1 residents completed most assigned simulations, with average case completion times supporting feasibility within the existing orientation curriculum. Learners reported increased confidence in managing each chief complaint following simulation completion, with confidence improving incrementally across cases. Residents expressed high satisfaction with the interactive format, perceived realism, and real-time feedback on clinical decision-making and physician–patient communication. Performance data demonstrated variability across learners, supporting the platform’s ability to differentiate clinical reasoning and communication skills.

AI-driven simulation can serve as a safe, scalable modality for early emergency medicine training. In this pilot, AI simulations were integrated into residency orientation, offering an efficient, low-resource approach to assessing resident clinical reasoning and physician–patient communication. An advantage of this platform is portability, enabling standardized deployment without reliance on local simulation infrastructure. Future work will include expansion to a 24-case curriculum, head-to-head comparison with traditional simulation, and evaluation of longer-term educational outcomes.