Analysis of the observational data often used in outcomes research is fraught with unique challenges. These include considerations like selection bias, highly skewed cost data, missing values, and patients switching treatments during the study. We have utilized a variety of methods to deal with these challenges, ranging from multiple regression with covariates, propensity score binning, optimal propensity score matching, instrumental variable analysis, variable transformation, two-part models, generalized linear models with gamma distributed errors, propensity-score stratified bootstrapping, mixed models for repeated measures, non-linear growth curve models, and marginal structural models. Let our years of experience in health outcomes research help you to identify an effective approach not only for accurately estimating the effects of your intervention, but also in communicating the results to your audience. Results of data analysis are only impactful when they are clearly understood by the intended audience.