Clinical Data Harmonization: Two Silos, One Layer
Clinical Data Harmonization: Two Silos, One Layer | Axiomera
Summary
This case study demonstrates how patient data siloed across different systems (hospital lab and physician EHR) can be standardized and integrated into a single 'Domain Intelligence Layer' using multiple AI engines. This process enables the construction of one longitudinal clinical record from two separate silos.
Details
The article details a sophisticated approach to unifying patient data fragmented across disparate sources, such as a Snowflake lab warehouse and a Google Cloud EHR. The process involves three main stages: Classification (SCE), Mapping (DMM), and Harmonization (DHM). Initially, raw site data is semantically classified by the SCE, which assigns clinical meaning to local codes or free text. Next, the DMM maps this data into FHIR R4 using standardized ontologies like ICD-10 and SNOMED CT. Finally, the Domain Intelligence Layer receives these standardized datasets. The DHM executes identity resolution, de-duplication, and unification of clinical information to generate a single 'longitudinal record.' This entire architecture is managed by an orchestration layer called Emergent Mapping Intelligence (EMI), which tracks pipeline improvement using the DERMS score. This methodology provides a highly advanced technical framework applicable to solving data interoperability challenges in Japanese healthcare settings involving legacy systems or departmental silos.
Original content copyright by respective publishers