OBJECTIVES:   In infectious disease, treatment decisions are often influenced by concerns about antibiotic resistance, which often leads to restriction of newer agents to sicker patients (i.e., selection bias).  PSM is often used to control for this problem in “real-world” comparisons. We examined the adequacy of PSM in a “real-world” comparison of vancomycin versus daptomycin as treatment for complicated skin and skin structure infections (cSSSI). METHODS: Using a database comprising >100 US hospitals, we identified admissions (1/1/2007 - 6/30/2010) with cSSSI who received initial antibiotic therapy with vancomycin or daptomycin. A propensity score model was estimated, using demographics, comorbidities, laboratory values, and receipt of vancomycin ≤30 days prior to hospitalization. Vancomycin patients were matched 1:1 to daptomycin patients in stepwise fashion to minimize the difference in propensity scores for each matched pair (i.e., “greedy” matching). RESULTS:   We identified 347 patients who received daptomycin and 8963 patients who received vancomycin as initial antibiotic therapy for cSSSI. Four hospitals contributed 54% of daptomycin patients, but only 17% of vancomycin patients. Daptomycin and vancomycin patients differed significantly in a number of respects. Only 47.6% of daptomycin patients could be matched to vancomycin patients (i.e., most patients had nonoverlapping propensity scores). Unmatched daptomycin patients were older than those in the matched subset (mean age: 57.3yrs vs. 52.3yrs); they also were more likely to have chronic/ulcerative infections (23% vs. 10%), comorbidities (e.g., diabetes [19% vs. 0%], malnutrition [4% vs. 0%], alcohol/drug abuse [11% vs. 1%]), and to have been hospitalized previously (63% vs. 39%) (all p<0.01). CONCLUSIONS: While PSM is often used to control for selection bias, the problem of nonoverlapping propensity score distributions is often overlooked and can adversely impact generalizability. Use of PSM to control for selection bias in “real-world” comparisons of initial antibiotic therapy for infectious diseases may be limited; alternate study designs may be needed.