New opportunities to rapidly evaluate societally relevant questions affecting health are emerging with the growing availability of large, observational datasets. The ability to evaluate the effect of interventions, exposures, or policies on health outcomes is increasingly accessible at reduced cost. A critical step when conducting an observational database study is ensuring that influential characteristics associated with the study main exposure or the study outcome are balanced between the two or more groups being compared to each other. If this equality of confounding is not achieved, the study results could be biased. To achieve greater balance of influential characteristics when comparing two or more groups in a health research study, statistical adjustment methods are commonly used. One approach for confounder balancing is usage of propensity score matching. However, selection of covariates in the propensity score model will impact the confounder balance between the groups being compared in the analysis. This process becomes more complex when interaction terms in the propensity score regression model as the subject matching procedure relies solely on the propensity score itself rather than the confounders that drive the score. In this case study, I will describe an analysis that utilized propensity score matching. I will describe how different specifications of a commonly used matching procedure impacted the confounder balance and the influence on the study odds ratios. This includes a demonstration of when improved matching of the propensity score led to reduced overall confounder balance. Propensity score matching methods are an effective and growing means of adjusting for confounders in observational data research. However, researchers must pre-specify the goals of their propensity score production algorithm as well as the level of acceptable individual confounder imbalance.