Federated Multi-Agent Architecture for Harmonizing Public Health Datasets into OMOP and FHIR Standards

This study presents a federated system developed within the SHIELD project, part of Horizon Europe's effort to reduce non-communicable diseases. It integrates retrospective and prospective clinical data using LLM-based multi-agent systems for automated ETL and natural language querying. Harmonization of MIMIC-IV, ELSA, and synthetic data into OMOP CDM and FHIR enabled validation. Results show high mapping accuracy and demonstrate the feasibility of scalable and interoperable data integration supporting clinical decision-making.