Using electronic health record data to identify comparator populations for comparative effectiveness research

Electronic health records (EHRs) can define real world patient populations with high levels of clinical specificity, potentially addressing some of the shortcomings of other types of real world data (RWD) when informing decisions about the comparative effectiveness of medical technologies. An important but under-recognized concern for EHR-derived RWD, however, is that the rich clinical data permits creation of very homogenous subpopulations from the larger group of eligible patients, thereby reducing the representativeness of the cohort relative to clinical practice. In this article, we discuss the tradeoffs between choosing clinical specificity versus representativeness in population sampling for comparative effectiveness research. Using EHR-derived RWD, we provide an example in non-small cell lung cancer to illustrate the concepts, showing wide variation in outcomes among potential comparator cohorts. We close with several recommendations for selecting comparator populations from EHRs that address the balance between matching clinical guidelines and capturing practice variability in comparative effectiveness research.