OBJECTIVE: Propensity scores (PS) are often used with the binary treatments. However, in day to day practice multiple treatment settings are experienced rather than binary treatments. Therefore extension of binary PS  analysis to multiple PS will add to the empirical knowledge of use of PS. We compared binary PS analysis with multiple PS approach by examining clinical effectiveness in patients with Chronic Heart Failure (CHF). METHODS: The study was a retrospective analysis of a national cohort of patients diagnosed with CHF identified from the Department of Veterans Affairs electronic medical records system. PS  analysis (binary and multiple) was used to balance 47 baseline patient characteristics between the different Angiotensin Converting Enzyme Inhibitors (ACEIs). For multiple PS we used multinomial logistic regression and for binary PS we split our cohort into separate models. Effect of different ACEIs on time to death was assessed using a multiple PS weighted Cox proportional hazard model and three separate binary PS weighted Cox proportional hazard models.  Captopril was used as reference in all models. The statistical significance of effect of individual ACEIs on mortality was compared between the two propensity approaches. RESULTS:  For binary propensity approach the adjusted hazards ratio from three different PS-weighted Cox models were 1.003 (95% CI: 0.724-1.390) for enalapril, 0.740 (95% CI: 0.688-0.796) for fosinopril and 0.823 (95% CI: 0.770-0.879) for lisinopril compared with captopril. For multiple propensity approach the adjusted hazards ratio were 1.033 (95% CI 0.739-1.445) for enalapril, 0.738 (95% CI: 0.685-0.796) for fosinopril, and 0.819 (95% CI: 0.767-0.875) for lisinopril. CONCLUSION: We found the 2 propensity approaches produced similar estimates of the effects of individual ACEIs on mortality. Multiple PS may be used more often if no information needed to predict outcomes is lost from sub sampling.