Fig. 1

Overview of simulation framework and study design.

Using Monte Carlo simulation, longitudinal datasets are generated with pre-specified covariate distributions and correlation structures before applying assumed treatment assignment and outcome models. Matching methods are then applied within each dataset using covariates observed either at baseline (time-invariant propensity score matching) or per time interval (time-dependent propensity score and genetic matching). Treatment effects are estimated in each matched cohort. The results of each matching method are then compared based on bias and efficiency of treatment effect estimates and covariate balance metrics