Purpose
Artificial intelligence (AI) has sparked tremendous interest over the last several years regarding its diverse applicability in medical education. A notable challenge is the abrupt transition from the preclinical to clinical years. This presents a promising opportunity to use AI to enhance student learning in patient interviews and history-taking. We developed a novel large language model (LLM) tool designed to help students to improve their patient interview skills. This abstract describes the app’s development and initial feedback from students and faculty.

Methods
Example patient actor scripts and write-ups were created by VT and the digital actor’s behavioral heuristics and prompt engineering were designed by AP. This work was developed through the medium of an in-house platform “medicalsimchat” (https://medicalsimchat.vercel.app/). The frontend was built using React, Tailwind and Radix UI libraries. The OpenAI GPT-4o-mini model and Realtime API were utilized.The backend was developed with Node.js and Next.js API routes, along with Typescript 5/Javascript client libraries for integration and publishing.

15 preclinical medical students and 2 faculty members provided feedback after completing an example case.

Results
Students strongly appreciated the ability to practice a traditionally time-consuming learning process at their own pace. AI-driven feedback on missing or incorrect elements in their write-up helped them consolidate their mistakes more effectively, Faculty members were impressed with the tool’s ease of use, patient dialogue flexibility, and the feedback system. Data collection for this study is ongoing.

Conclusion
The next step of AI’s role in medical education is to optimize the integration of medical knowledge transfer with softer skills like empathetic conversation, pertinent questioning, and history taking. The encouraging early responses support a combined student/faculty relationship in developing these AI tools effectively.