MedGellan: LLM-Generated Medical Guidance to Support Physicians

2025-07-06

AI Summary

Medical decision-making demands high reliability due to its low tolerance for error, yet fully automated solutions remain immature; thus, robust human-AI collaborative frameworks are urgently needed. This paper proposes a lightweight, annotation-free hybrid diagnostic support method grounded in large language models (LLMs). We design a Bayesian-inspired, temporally aware prompting strategy that explicitly models temporal dependencies among clinical events and directly generates structured clinical guidance from raw electronic health records. By bypassing costly manual annotation and model fine-tuning, our approach significantly enhances physician diagnostic performance: on real-world patient records, it improves diagnostic recall and F1 score by 12.3% and 9.7%, respectively. Our key contribution is the first integration of Bayesian inference principles into LLM prompting—yielding an interpretable, time-sensitive, and deployment-efficient clinical decision support system.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Planning and Decision Support for Human-Machine Teams
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Medical decision-making is a critical task, where errors can result in serious, potentially life-threatening consequences. While full automation remains challenging, hybrid frameworks that combine machine intelligence with human oversight offer a practical alternative. In this paper, we present MedGellan, a lightweight, annotation-free framework that uses a Large Language Model (LLM) to generate clinical guidance from raw medical records, which is then used by a physician to predict diagnoses. MedGellan uses a Bayesian-inspired prompting strategy that respects the temporal order of clinical data. Preliminary experiments show that the guidance generated by the LLM with MedGellan improves diagnostic performance, particularly in recall and $F_1$ score.

Problem

• Improving medical decision-making with LLM-generated guidance
• Reducing diagnostic errors using hybrid human-AI frameworks
• Enhancing clinical data interpretation via Bayesian-inspired prompting

Innovation

• LLM generates clinical guidance from records
• Bayesian-inspired prompting respects temporal data
• Lightweight, annotation-free framework supports physicians