Data analyses

Sample characteristics were described overall. The mean and standard deviation for each BTI item, as well as the n and percent that endorsed each response type (Disagree, Uncertain, Agree) were calculated. Each individual BTI item was described (rather than aggregated in previously identified BTI domains [15]) in order to describe the full range of barriers reported in this sample for this hypothesis-generating study. We then coded qualitative responses to the open-ended question of additional barriers using template analysis [27] (i.e., content analysis or thematic coding). In this approach, data are coded both inductively and deductively, thus allowing for coding of both a priori domains and emergent domains. The initial coding template was based on the seven factors (domains) of the BTI reported by Rapp et al. [15]. Two coders (M.D.O., J.A.C.) independently coded open-ended responses; discrepancies then were discussed by all authors until consensus was reached. Consistent with the recommendations by Sandelowski [28], we opted not to report frequencies of qualitative results, as the use of numbers may have detracted from making well-rounded interpretations of the data, particularly in the context of data generated from a single, open-ended survey question as opposed to an in-depth interview.