An OMOP-Compatible Exposure Infrastructure for GeoAI-Enabled Environmental Health Research: The OHDSI GIS Toolchain

Abstract Environmental and socio-economic contexts shape population health. Despite increasing availability of data about such contexts, no open infrastructure standardizes these data into person-place-time exposure records for federated health research. We present the OHDSI GIS toolchain, an open-source infrastructure integrating geospatial data with the OMOP Common Data Model through a FAIR catalog, a GIS vocabulary, local geocoding, and a geometry-attribute index that registers every source uniformly. We tested the architecture on a synthetic cohort of 40,000 persons in real US census tracts, with eight simulated PM2.5 effects and six null outcomes. Derived exposures matched an independent answer key, and known effects were recovered. Displacing residences by empirical geocoding error changed estimates little for a smooth regional exposure but attenuated them by up to half for exposures varying over 100 m. These results show that the toolchain and its standards enable reproducible, transparent linkage of geospatial sources with health records.