Medical Heuristic Learning: An LLM-Driven Framework for Interpretable and Auditable Clinical Decision Rules

Medical Heuristic Learning (MHL) introduces a framework leveraging large language models (LLMs) for interpretable clinical decision-making, addressing the limitations of traditional black-box models in healthcare. MHL employs a combination of statistical and medical knowledge probes, rule synthesis, and iterative code refinement to generate transparent, auditable decision rules expressed in pure Python, which are adaptable to data drift and feature evolution. Experimental results demonstrate that MHL achieves performance on par with state-of-the-art methods while effectively managing small sample sizes and class imbalances, making it a significant advancement for practitioners seeking interpretable AI solutions in clinical settings.