A Review of Generalizability and Transportability

When assessing causal effects, determining the target population to which results are intended generalize is a critical decision. Randomized and observational studies each have strengths limitations for estimating effects in population. Estimates from randomized data may internal validity but often not representative of Observational better reflect population, hence be more likely external validity, subject potential bias due unmeasured confounding. While much inference literature has focused on addressing bias, both necessary unbiased estimates This article presents framework including synthesis approaches generalizability transportability, assumptions they require, as well tests heterogeneity treatment differences between study populations.