Coverage-Aware Reasoning with Medical Tokens for Diagnosis Prediction

Large language models (LLMs) offer promising potential for next-visit diagnosis prediction, owing to their ability to integrate longitudinal clinical evidence and reason over it in natural language.