Leveraging Large Language Models for Prediction and Explanation on Electronic Health Records

This thesis enhances the doctorXAI explainable pipeline for Electronic Health Records by integrating new, transformer-based, state of the art model, to improve prediction accuracy and by replacing ontological knowledge with a generative model for explanation. This shift removes the reliance on manually curated domain-specific resources, boosting the system’s flexibility, and ability to provide richer, data-driven explanations.