Epidemiologists spend a considerable amount of time designing and analyzing the results of observational studies. There are many challenges in the realm of observational studies with respect to controlling for confounding and addressing bias. Traditional methods such as matching and stratification have been used to address confounding but their limited usefulness is obvious when it comes to the large number of potential sources of confounding in observational studies. One is naturally limited by the number of variables upon which one can either match or stratify. One methodological tool in the epidemiologist’s toolkit is the use of propensity scores to adjust for large numbers of potential confounders. A general overview of propensity score methods will be presented. In addition, a presentation of considerations regarding choosing the appropriate variables and propensity score method will be provided. As an example, results from an analysis of administrative data for evaluating the comparative safety of individual anti-psychotic medications in the non-institutionalized elderly will be presented.