OBJECTIVES: To introduce a method that combines the propensity score matching and interrupted time-series models to measure drug therapy guidelines on outcomes.   METHODS:  Propensity score matching is used to balance groups before the trend is analyzed. The “kitchen sink” approach is used for propensity score matching.  Interrupted time-series models are applied over the matched sample. The time-series model contains two predictor variables: the binary intervention variable and an interval coding for time. This model controls for the confounding influence of any underlying trend and ensures that any estimated change in the mean level of the series after intervention is not simply due to the fact that the series was already decreasing or increasing. RESULTS: : To illustrate the model, changes in the utilization of two hypothetical drugs were analyzed after issuance of guidelines. Patients who used these two drugs were different at the baseline in terms of observable characteristics such as age, gender, and comorbidities (p<0.001). Samples were balanced with nearest neighbor matching. Then segmented time-series models were applied. There was a significant association between the onset of intervention and the level of utilization of these drugs. CONCLUSIONS: To isolate the effects of guidelines, we need to control for three different factors: 1) baseline differences between the two groups; 2) stepwise differences at the intervention point; and 3) trend differences after the intervention. We showed that propensity score matching can be used for the first factor, and the latter two can be controlled with the interrupted time-series model