Assessing and Mitigating Medical Knowledge Drift and Conflicts in Large Language Models

TL;DR

This paper evaluates how large language models in healthcare handle evolving medical knowledge, revealing challenges in maintaining accuracy over time and proposing mitigation strategies to improve their reliability.

Contribution

Introduces the DriftMedQA benchmark to simulate medical guideline evolution and assesses mitigation strategies like retrieval augmentation and preference fine-tuning.

Findings

Models struggle with outdated and conflicting medical advice.

Mitigation strategies improve model reliability.

Combined methods yield the most consistent results.

Abstract

Large Language Models (LLMs) have great potential in the field of health care, yet they face great challenges in adapting to rapidly evolving medical knowledge. This can lead to outdated or contradictory treatment suggestions. This study investigated how LLMs respond to evolving clinical guidelines, focusing on concept drift and internal inconsistencies. We developed the DriftMedQA benchmark to simulate guideline evolution and assessed the temporal reliability of various LLMs. Our evaluation of seven state-of-the-art models across 4,290 scenarios demonstrated difficulties in rejecting outdated recommendations and frequently endorsing conflicting guidance. Additionally, we explored two mitigation strategies: Retrieval-Augmented Generation and preference fine-tuning via Direct Preference Optimization. While each method improved model performance, their combination led to the most consistent and reliable results. These findings underscore the need to improve LLM robustness to temporal shifts to ensure more dependable applications in clinical practice. The dataset is available at https://huggingface.co/datasets/RDBH/DriftMed.