OBJECTIVES: Asenapine (ASE), an oral Atypical Antipsychotic (AA), was initially used for more severe bipolar I mixed disorder. Different propensity score (PS) methods were investigated to achieve balanced baseline characteristics between ASE and four oral AA cohorts for eventual outcomes analyses.  METHODS: Adults with ≥1 asenapine, aripiprazole, olanzapine, quetiapine, or risperidone prescription fill (Aug 2009 to Dec 2010) and diagnosis of bipolar I mixed disorder (ICD-9-CM: 296.6x) from MarketScan® claims databases, yielded 230 ASE, 2726 aripiprazole, 984 olanzapine, 3056 quetiapine, and 1623 risperidone patients. PS were derived using logistic regression models for ASE and each AA with baseline demographic and clinical characteristics as covariates. PS, inverse probability treatment weight (IPTW: 1/PS ASE; 1/(1-PS) AA), and standard mortality ratio weight (SMR: 1 ASE; PS/(1-PS) AA) distributions were evaluated. ASE-AA un-weighted, IPTW, and SMR baseline characteristics were compared using standardized differences, chi-squares, and t-tests. RESULTS: Un-weighted asenapine patients had pre-index greater bipolar I episodes rates, psychiatric drug use, dyslipidemia and obesity (all comparators). PS distributions for asenapine-olanzapine overlapped to some degree while PS of asenapine and the other comparators overlapped little to not at all. Comparing IPTW baseline characteristics, asenapine more resembled the AA cohorts. Demographic imbalance increased between asenapine and each AA. IPTW improved clinical characteristic balance for asenapine versus olanzapine and risperidone, but only slightly improved imbalance versus aripiprazole and quetiapine. However some clinical characteristics not previously balanced in the un-weighted analyses for asenapine versus each AA were now imbalanced. Applying SMR, AA cohorts more resembled the asenapine cohort and all baseline demographic and clinical characteristics were finally balanced.  CONCLUSIONS: SMR, a less common PS method, resulted in balanced baseline characteristics. SMR should be considered when IPTW leaves imbalance and the cohort of primary interest differs significantly from the broader underlying population to which it’s being compared.