Details
The FDA initiated this study to understand the differences in how key covariates are captured in electronic health records (EHR) versus claims data. This study assessed the identification of key covariates, such as body mass index (BMI) and smoking status, among users of sodium-glucose co-transporter 2 (SGLT2) inhibitors and sitagliptin. This examination was conducted using EHR data from the National Patient-Centered Clinical Research Network (PCORnet) as well as claims and Integrated Delivery System (IDS) data in the Sentinel Distributed Database, from March 1, 2013, through June 30, 2018.
EHR data identified more patients with documented records of BMI and smoking status than claims data. An IDS analysis of new SGLT-2 inhibitor users observed a good concordance (77%) between claims and EHR data for current tobacco use. A more sensitive definition of obesity in claims data increased concordance (44%) with EHR data, compared to a narrower, BMI-specific definition (14%).
This methods project characterized missingness and clinical measurement concordance between EHR and claims data for key covariates, which could inform quantitative bias analysis or imputation approaches for FDA’s drug safety assessments.