UCVA ontology: Standardizing local context factors to support the analysis of unwarranted clinical variation

Unwarranted clinical variation (UCV), defined as care provided to a patient that is not proportional to the patient's needs, clinical characteristics, or preferences, results in negative patient outcomes. UCV detection is performed to compare healthcare outcomes across regions or organizations and reduce UCV incidences. However, the lack of standardized UCV concepts and the local context factors hinder the interoperability of UCV analyses between organizations. Currently, UCV analysis relies on centralized analysis, posing challenges due to the proprietary nature of an organization's healthcare and operational data. Standardizing analysis factors enables interoperable analyses by establishing a common context between organizations. This work introduces a standardization of local context factors of UCV analysis.

We standardized local context factors for UCV analysis as an application ontology, UCV analysis ontology (UCVA). Using Protégé 5, UCVA references concepts with similar semantics from external ontologies and creates new concepts de-novo, with Microsoft Copilot as a contributor of synonym terms. Application reasoners and human evaluators are utilized to validate UCVA.

UCVA has 78 classes, 102 instances, and 20 object properties. There are 944 axioms: 227 declarations axioms, 359 logical axioms, and 358 assertions. Textual definitions for classes are provided using the "IAO:0000115" annotation from the Information Artifact Ontology. Additional related terminology is drawn from Medical Subject Headings and documented through the "rdfs:comment" property. UCVA references concepts from external ontologies including Ontology of Biomedical and Clinical Statistics and Ontology for microRNA Target prediction. These are referenced using the Simple Knowledge Organization System's "skos:closeMatch" and "skos:exactMatch" property.

UCVA is designed to support local context factor information as standardized variables for UCV analysis, enabling the interoperability of UCV analysis between organizations without the need to pool together data for UCV analysis.