The Rubin Causal Model is reviewed and its relationship to indirectly and directly standardized mortality rates historically used for provider quality assessments made. The potential outcome for patient, j, treated by provider i, is the outcome had the patient been treated by a provider other than i. A variety of causal parameters are defined, assumptions for identification discussed, and an assortment of estimators introduced. These include outcome regression modeling, provider assignment modeling via the propensity score and the generalized propensity score, and matching and weighting approaches using the propensity score, all based on parametric modeling approaches. Approaches that work with the confounders constrained optimization procedures without specifying a model are reviewed. Nonparametric machine learning algorithms that can be used in regression modeling or provider assignment modeling are introduced.