Fig. 1.

Conceptual models. Two patient characteristics were systematically manipulated in the 2 × 3 experimental design: race (Black or White) and risk behavior (man who has sex with men [MSM], man who has sex with women [MSW], and man who injects drugs [MID]). Patient race, patient risk behavior, and the interaction of patient race and risk behavior were examined relative to the set of clinical judgments reported by providers in a multivariate model (Model A). The indirect effects of patient characteristics on intention to prescribe PrEP for the patient were tested with all other clinical judgments of the patient considered as mediators in a parallel multiple mediator model (Model B). Finally, the moderating effect of provider prejudice on the indirect effects in the parallel multiple mediator model was examined for different patient characteristics and corresponding forms of prejudice (Model C)