Code Like Humans: A Multi-Agent Solution for Medical Coding
This addresses the challenge of automating medical coding from unstructured clinical notes for healthcare professionals, representing a domain-specific advancement.
The paper tackles the problem of medical coding by introducing a multi-agent framework that implements official coding guidelines and supports the full ICD-10 system with over 70,000 labels, achieving state-of-the-art performance on rare diagnosis codes while noting that fine-tuned classifiers retain an advantage for high-frequency codes.
In medical coding, experts map unstructured clinical notes to alphanumeric codes for diagnoses and procedures. We introduce Code Like Humans: a new agentic framework for medical coding with large language models. It implements official coding guidelines for human experts, and it is the first solution that can support the full ICD-10 coding system (+70K labels). It achieves the best performance to date on rare diagnosis codes (fine-tuned discriminative classifiers retain an advantage for high-frequency codes, to which they are limited). Towards future work, we also contribute an analysis of system performance and identify its `blind spots' (codes that are systematically undercoded).