Randomized clinical trials (RCT) have long been the gold standard for establishing the causal effect of treatments. However, RCTs typically enroll a narrow population of patients limited by strict inclusion and exclusion criteria. Thus, RCT populations generally include more adherent patients and exclude patients using concomitant medications and patients with comorbid diseases. Health care decision makers, however, need to understand whether RCT evidence can generalize to select usual care populations who may receive the intervention in usual care settings.
This presentation will begin with a review of the literature on the representativeness of RCT populations followed by a discussion of the concepts of transportability and generalizability. Second, we will discuss the analytical challenges in generalizing RCT evidence, including heterogeneity, non-overlapping populations, and missing confounders. Lastly, we will present current and novel analytic methodology for and applied examples of generalizing evidence to a target population. This will include inverse probability of treatment weighting, entropy balancing, and newer modeling methods.