(C-155) Transforming a Large-Scale Prostate Cancer Outcomes Dataset to the OMOP Common Data Model—Experiences with a Regional Health Record Database in China

Background: Diverse vocabularies across healthcare institutions covered by the Tianjin regional electronic health record (rEHR) database in China hinder semantic interoperability and data exchange. The Observational Medical Outcomes Partnership (OMOP)-Common Data Model (CDM) has been widely used to standardize and harmonize observational health data for international research.

Objectives: To assess the data structure for real-world evidence (RWE) studies regarding prostate cancer (PC) using the Tianjin rEHR database and to standardize medical terms into global vocabularies according to the OMOP-CDM.

Methods: Data tables from the Tianjin rEHR database, spanning from 2017 to 2021, were extracted and transformed into the OMOP-CDM schema. Patients with at least one diagnosis of PC during outpatient visits or hospitalizations in the Tianjin rEHR database were included in this study. The OMOP-CDM conversion involved profiling the source data to understand its structure and content, mapping source data elements to the specified target locations in the CDM schema through extract, transform, and load (ETL) operations, and applying vocabulary mappings to translate the codes and values used in the source data to those used in the CDM. The data vocabularies employed included ICD-10-CN for diagnosis codes, Anatomical Therapeutic Chemical (ATC) classification for drugs, and Logical Observation Identifiers Names and Codes (LOINC)-CN for laboratory tests and vital measurements. A total of 126 predefined data quality checks were performed to validate the quality of transformed data.

Results: Data profiling of the Tianjin rEHR database revealed that the dataset includes approximately 310,000 clinical visits, encompassing both outpatient and inpatient records. Around 70% of these visits occurred between 2019 and 2021. All clinical information required for the OMOP-CDM structure was contained in the Tianjin rEHR database, with over 85% of records were successfully mapped. A total of 18,745 male PC patients met the selection criteria, and their records were converted into 25 OMOP-CDM tables. The transformed database passed all 126 quality checks, including those for patient information, diagnosis, medication, procedures, and lab measurements.

Conclusions: The adoption of OMOP-CDM at the Tianjin rEHR database demonstrated that rEHR data in China are feasible for standardization with minimal information loss. This conversion highlights the significant potential for application to other rEHR or independent electronic medical record (EMR) databases in China, enhancing data interoperability in the era of the single large-scale health data platform promoted by the government.