OBJECTIVES: Hybrid applications of conventional exact covariate matching and propensity score concepts have recently been explored in the literature. In this research, we examine the advantages and disadvantages of hybrid usage of these two principles. Specifically, we evaluate the impact on comparable samples and treatment persistence measures from retrospective prescription claims data. METHODS: Patient persistence on anti-diabetic agents (Exenatide and Insulin Glargine) was used to compare six hybrid matching algorithms proposed by Yang and Stemkowski (2008), using IMS' LifeLink longitudinal prescription database (LRx). Persistence was evaluated by persistent days on quartiles, persistence rate over time, survival censoring rate, Kaplan-Meier survival curves and Cox Proportional Hazard Model. The hybrid matching algorithms were compared on run time, matching rate and bias reduction. RESULTS: Directly matched cohorts resulted in more comparable samples and improved the evaluation on treatment persistence relative to pre-matched sample. When propensity score played as a caliper, the matching process resulted in un-balanced variables and exhibited the weakest ability in bias correction due to the least drop in standardized difference. This was the only method among the six that failed assumption tests in the survival analysis (P<0.05). Conversely, among all other algorithms where all factors were balanced, the algorithm in which propensity score acted in a parenting role had the least running time and greatest bias reduction, which was displayed by the largest decline in standardized difference. The smallest values in Fit-of-Statistics through the whole study period also indicated the strongest hold of assumptions in the Cox proportional hazard model, relative to the other five algorithms. CONCLUSIONS: In the assessment of treatment persistence through survival analysis, the role of propensity score as a parenting factor in the selection of matched samples outperformed alternative hybrid matching algorithms. In contrary, use of propensity score as a caliper factor, was least satisfactory.