The 5R's of large language model-assisted diagnosis: A practical framework for hospitalists

Abstract

Diagnostic error remains a major patient safety challenge in hospital medicine. Large language models (LLMs) are increasingly used by clinicians to aid in diagnosis, yet most lack a structured approach for doing so safely and effectively. In this piece, we propose a practical, clinician-centered framework for LLM-assisted diagnosis. The 5R's Framework promotes a human-in-the-loop approach that harnesses LLMs to reduce diagnostic error while preserving critical reasoning skills.