MedRep introduces a novel approach to overcome a critical limitation in Electronic Health Record (EHR) foundation models - the inability to process medical codes outside their training vocabulary.

• Creates generalizable representations of medical concepts using the OMOP common data model
• Enables models to handle previously unseen medical codes without retraining
• Demonstrates improved performance across various medical tasks
• Facilitates integration between models trained on different vocabularies

This innovation significantly expands the practical utility of AI in healthcare by allowing medical AI systems to adapt to diverse clinical settings and terminologies without requiring complete retraining, potentially accelerating adoption across healthcare systems.

MedRep: Medical Concept Representation for General Electronic Health Record Foundation Models