3559 Generating interactive case reports for self-directing learning from published literature with a large language model

Le résumé fourni par la source

Background/ Objectives Case reports are a longstanding cornerstone of medical education and literature. The aim of this study was to evaluate the feasibility of using artificial intelligence (AI), namely large language models (LLM), to convert published case reports into an interactive format.Method This cross-sectional study utilised a sample of five published case reports with neuroinfectious diagnoses. The cases were converted into a ‘screenplay’ based on an existing template with the selected LLM (OpenAI’s GPT-4o). ‘Screenplays’ were then uploaded to a custom internet interface to enable users to engage with them in a question-and-answer format. Three users interacted with the selected cases through sequential requests for information regarding patient history, examination findings, and diagnostics. All AI question-answer pairs were then reviewed by a neurologist.Results A total of 803 question-answer pairs were generated. 1/803 (0.1%) responses included information not in the original case report (termed ‘hallucinations’). The remaining 802/803 (99.9%) responses were deemed appropriate, including 113/803 (14.1%) responses in which the LLM declined to answer. A ‘declined to answer’ response may occur, for example, following a request for ‘liver function tests’ when none were presented in the case; the response may be ‘These results are not available.’ The single error involved a question regarding the timeline of the illness, the answer to which appeared unclear upon human review of the published case report.Conclusion AI, namely LLM, can effectively convert published case reports into interactive cases, creating a novel and engaging way for readers to explore and learn from medical literature.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.