Cross-Standard Health Data Harmonization using Semantics of Data Elements

Faced with heterogeneity of healthcare data, we propose a novel approach for harmonizing data elements (i.e., attributes) across health data standards. This approach focuses on the implicit concept that is represented by a data element. The process includes the following steps: identifying concepts, clustering similar concepts and constructing mappings between the clusters using the Simple Standard for Sharing Ontological Mappings (SSSOM) and Resource Description Framework (RDF), and enabling the creation of reusable mappings. As proof-of-concept, we applied the approach to five common health data standards - HL7 FHIR, OMOP, CDISC, Phenopackets, and openEHR, across four domains, such as demographics and diagnoses, and nine topics within those domains, such as gender and vital status. These domains and topics are selected to represent the broader range of topics in the health field. For each topic, data elements were found in the health data standards after a thorough search, resulting in the analysis of 64 data elements, identification of their underlying concepts, and development of mappings. Three use cases were implemented to demonstrate the role of data element concepts in data harmonization and data querying at varying levels of granularity. The approach helps overcome the limitations of context-dependent mappings and provides valuable insight for mapping practice within the health domain.