We are exploring the use of synthetic controls, a type of external controls method that constructs a synthetic control arm for a treatment group by leveraging information across “control'' patients whose outcomes are observed. We start with data from Phase III and Phase IV Clinical trials to estimate comparative effectiveness and safety profile as measured by adverse events for each group of patients. We then compare its performance to standard causal inference methods (propensity score and matching, machine learning) in quantifying the treatment effect of transfusion policies on clinical outcomes for patients with acute upper gastrointestinal bleeding and biologic therapies for patients with ulcerative colitis. The estimated treatment effects can then be extended to real world patient registries containing similar covariates for patients to personalize treatment strategies.