Ontology-driven identification of inconsistencies in clinical data: A case study in lung cancer phenotyping.

Abstract

To illustrate the use of an ontology in evaluating data quality in the medical field, focusing on phenotyping lung cancers. We crafted an ontology to encapsulate crucial domain knowledge, leveraging it to query the Clinical Data Warehouse (CDW) of Bordeaux University Hospital. Our work aimed at accurately representing domain knowledge and identifying inconsistencies through ontological axioms. Specifically, our aim was to pinpoint lung cancer patients with EGFR or ALK mutations treated with tyrosine kinase inhibitors (TKIs). We evaluated the ability of this ontology to retrieve and characterize patients in comparison with a traditional SQL queries executed on the CDW. The ontology's results closely aligned with those of the SQL queries. A sub-cohort of 60 lung cancer patients with conflicting information was identified, highlighting inconsistencies in the data. Moreover, the ontology complemented the existing data, uncovering additional information and enriching the dataset. This work has highlighted challenges in managing temporal data and handling imperfect data. Addressing these challenges is essential for the effective use of CDW in phenotyping. Ontologies improve data quality by identifying inconsistencies, enhancing data completeness, facilitating complex SQL queries, and standardize processes. Developing a framework to manage inconsistent healthcare data, considering its temporal nature, is essential.