A Trial Emulation Approach for Policy Evaluations with Group-level Longitudinal Data

To limit the spread of novel coronavirus, governments across world implemented extraordinary physical distancing policies, such as stay-at-home orders. Numerous studies aim to estimate effects these policies. Many statistical and econometric methods, difference-in-differences, leverage repeated measurements, variation in timing policy effects, including COVID-19 context. Although methods are less common epidemiology, epidemiologic researchers well accustomed handling similar complexities individual-level interventions. Target trial emulation emphasizes need carefully design a nonexperimental study terms inclusion exclusion criteria, covariates, exposure definition, outcome measurement—and those variables. We argue that evaluations using group-level longitudinal (“panel”) data take careful approach we refer emulation. This is especially important when intervention varies jurisdictions; main idea construct target trials separately for each treatment cohort (states implement at same time) then aggregate. present stylized analysis impact state-level orders on total coronavirus cases. estimates from panel methods—with right modeling diagnostics—can help add our understanding many though doing so often challenging.