(129) Characterization of a Novel Oncology Electronic Health Records-Derived Data Linkage to Commercial Claims

Publication Number: 1174

Background: Real-world data from electronic health records (EHR) are enhanced by linkage to complementary data sources such as healthcare claims, enabling investigations that would be limited by the use of only EHR or claims alone.

Objectives: Linkage between an oncology EHR-derived dataset and commercial claims data was implemented and characterized to better understand which novel research questions the linkage might address.

Methods: This study used the nationwide (US-based) de-identified EHR-derived Flatiron Health (FH) database to select patients diagnosed with at least one of 20 cancer types, per ICD codes and confirmed by clinical documentation, concurrent with enrollment in healthcare plans with complete medical and pharmacy claims available from Komodo Health commercial claims data. Duration of enrollment in claims-observed healthcare plans was assessed before and during EHR-observed patient-time. Standardized mean differences in clinical and demographic characteristics were calculated between linked and unlinked patients, along with descriptive statistics about volume and variety of claims data.

Results: Sixteen percent (N = 44,702 / 282,203) of patients were observed in claims data at the time of EHR cancer diagnosis, ranging from 15% to 20% with type of cancer. Mean healthcare plan enrollment durations prior to EHR diagnosis ranged from 2.2 up to 3.0 years, and enrollment time overlapped with a mean 77% to 93% of EHR patient-time, depending on cancer type. Patient characteristics were generally similar between linked and unlinked patients, and observed volume of claims data for patients after cancer diagnosis was greater than before diagnosis.

Conclusions: We determined that robust linkage of two data sources, an EHR-derived oncology database and aggregated commercial claims, is feasible and may support cancer outcomes research by providing longitudinal follow-up from claims combined with clinical detail from EHR.