OBJECTIVES: To develop a systematic approach using disparate observational databases for identifying pre-exposure differences in condition incidence across drug cohorts. A case study to examine the utility of the method was conducted, comparing dutasteride with finasteride. METHODS: Two disparate databases an electronic health record (EHR) and an administrative claims database were used for analysis. We applied an unmatched cohort design to each source, capturing all persons within the two exposed populations. For all conditions, we calculate unadjusted incidence rates prior to exposure for each cohort and the associated incidence rate ratios (IRR) between cohorts. Three different IRR estimates were calculated using unique definitions of person-time: ‘6 months' prior to exposure, any time ‘before' exposure, and ‘variable' time based on length of exposure. Each method used statistical significance of the IRR as the threshold for identifying ‘candidate differences' (CD). A composite threshold requiring significance across both sources was also used. RESULTS: Using the ‘6 months' metric, 194 CDs within the EHR and 469 within the claims database were identified, with 108 conditions occurring in both sources. Expert review found the combined list contained all concepts previously hypothesized as important to consider when designing a dutasteride-finasteride study, as well as 10 conditions not hypothesized but deemed important for any evaluation, 40 unexpected pre-exposure conditions that warranted further consideration, and 10 terms that added no value. There was good concordance across metrics, with 70 of 129 ‘before' CDs and 95 of 161 ‘variable' CDs matching the ‘6 months' results. CONCLUSIONS: Exploratory analysis of pre-exposure cohort differences can enhance design of observational evaluation studies, guiding researchers to develop their conceptual model to assess the relationship between treatment and outcome by identifying potential sources of sample selection bias. Using multiple data sources allows independent verification of exploratory findings, raising confidence that CDs identified bear consideration.