Gravar-mail: Regression Discontinuity Designs in Epidemiology: Causal Inference Without Randomized Trials