Improving Causal Inference in Observational Studies: Propensity Score Matching