Addressing Data Quality Challenges in Lung Cancer Data Within the Observational Medical Outcomes Partnership Common...

The secondary use of health data is essential for advancing medical research and improving clinical practice. The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) enables large-scale, multicenter studies but faces challenges related to consistency, completeness, and transparency during data mapping from original data sources.

This study aimed to evaluate the quality of the mapping process for lung cancer data within the Federated Health Innovation Network project, with a focus on consistency, completeness, and challenges encountered throughout the process.

Clinical data from Ghent University Hospital were mapped to the OMOP CDM using a reference data dictionary. Consistency was assessed using Cohen kappa (κ) scores, while completeness was evaluated by comparing patient and record counts before and after mapping. Challenges, including unstructured data and an evolving reference standard, were documented and analyzed.

High consistency was observed for structured variables, while some unstructured variables, such as "Smoking status," were excluded due to their free-text format and the lack of suitable OMOP concepts. The completeness analysis showed minimal data loss for most structured variables but highlighted substantial challenges associated with unstructured data. Persistent issues included evolving data dictionary versions and mismatches in diagnostic code granularity between institutions, underscoring structural challenges in standardization.

The transformation of lung cancer data to the OMOP CDM highlighted both technical and systemic challenges, including the handling of unstructured data and the resolution of granularity discrepancies. A multidisciplinary approach involving clinical and technical expertise is crucial for ensuring reliable, high-quality datasets for multicenter research.