Appendix 1
Post Hoc Analysis of the Mediating Effects of Other PrEP Attitudes
Analytic Approach:
We performed post hoc mediation and conditional process (moderated mediation) analyses to explore whether the relationship between PrEP beneficiary group and support for PrEP financial assistance identified in our primary analyses could be explained by any of the other four PrEP attitudes identified—respect for taking PrEP, predicted risk compensation, perceived community benefit/support for access, or predicted adherence—and whether any such explanatory pathways were contingent upon level of prejudice.
Mediation is a statistical method commonly used to examine the mechanism(s) by which one variable (X) affects another (Y); it allows for the quantification and significance testing of both a direct effect (XàY) and indirect effect(s) through intervening variable M (XàMàY) or through multiple intervening variables operating in parallel or sequentially. We constructed a parallel multiple mediation model in which the other four PrEP attitudes were included in separate, co-occurring indirect pathways between PrEP beneficiary group (X) and support for PrEP financial assistance (Y). This parallel mediation approach allowed us to establish and contrast the specific indirect effect of each hypothesized mediator adjusting for all others. We subsequently used conditional process analysis, a statistical method that estimates the significance and magnitude of the indirect effect(s) as moderated by another variable or variables (moderated mediation), to determine whether specific indirect effects varied by level of prejudice (60).
We performed two separate sets of post hoc analyses, dichotomizing PrEP beneficiary group as Gay Man Condition vs. General Population Condition and Black Gay Man Condition vs. General Population Condition, as we did for earlier analyses. For each set of analyses, we used Hayes’ PROCESS macro (60) to analyze the following conceptual models: the four other PrEP attitudes as parallel mediators of the relationship between PrEP beneficiary group and support for PrEP financial assistance (Model 1), Model 1 with all indirect and direct paths moderated by racism (Model 2), Model 1 with all indirect and direct paths moderated by heterosexism (Model 3), and Model 1 with all indirect and direct paths moderated by racism x heterosexism (Model 4; see Figure S1). Relevant background characteristics (race, gender, sexual orientation, education, prior knowledge of PrEP, and knowledge of disparities in HIV prevalence) were adjusted for in all models. The PROCESS macro generated bias-corrected and accelerated 95% confidence intervals (CIs) of the indirect effect from 10,000 bootstrapped samples as well as indices of moderated mediation. Significance was established by CIs that did not straddle zero.
Figure S1. Conceptual Models
Model 1 Model 2
Model 3 Model 4
Figure S1 Caption. Conceptual models of direct and indirect effects of PrEP beneficiary group on support for PrEP financial assistance, tested via post hoc mediation and conditional process (moderated mediation) analysis.
Results:
Testing of the parallel mediator model not conditioned on either form of prejudice (Model 1) failed to yield evidence for a significant indirect effect of PrEP beneficiary group on support for PrEP financial assistance through any of the four mediators, irrespective of the stigmatized group to which the general population was being compared (gay men or Black gay men; see Figure S2). For models involving gay men (i.e., Gay Men Condition vs. General Population Condition analyses), this held true even when all paths were conditioned on racism (Model 2), heterosexism (Model 3), or both (Model 4). For models involving Black gay men (i.e., Black Gay Men Condition vs. General Population Condition analyses), a conditional indirect effect was present relative to only one of the four mediators tested, risk compensation, and this indirect effect was only present in Model 3 (Index of Moderated Mediation = -.15, SE = .09, 95% CI [-.42, -.02]). According to this model, for participants reporting relatively low heterosexism (heterosexism score = 1.00[1]), framing PrEP as benefiting Black gay men led to lower predicted risk compensation, which, in turn, was associated with higher support for PrEP financial assistance (95% CI [.02, .39]). This indirect pathway was not significant for participants reporting average (heterosexism score = 1.55) or relatively high (heterosexism score = 2.25 = 1 SD above the mean) heterosexism (95% CIs [.00, .23] and [-.22, .05], respectively).
Figure S2. Parallel Mediator Models (Model 1)
Gay Men Condition vs. General Population Condition:
Black Gay Men Condition vs. General Population Condition:
Figure S2 Caption. Unstandardized coefficients and standard errors of all paths of the unconditioned parallel mediator models (Model 1), adjusting for relevant background characteristics (race, gender, sexual orientation, education, prior knowledge of PrEP, and knowledge of disparities in HIV prevalence; knowledge of disparities included disparities by sexual orientation only for the Gay Men vs. General Population model and disparities by sexual orientation and race for the Black Gay Men vs. General Population model). No significant indirect effects were detected for any of the four mediators in either the Gay Man Condition vs. General Population Condition model or the Black Gay Man Condition vs. General Population Condition model (bias-corrected and accelerated bootstrapped 95% confidence intervals for all indirect paths straddled zero).
*p < .05 **p < .01
Interpretation:
Our post hoc mediation analyses generally failed to implicate any of the other PrEP attitudes we measured—respect for taking PrEP, predicted risk compensation, perceived community benefit/support for access, and predicted adherence—as explanatory factors in the disparity in support for PrEP financial assistance observed. One exception was the finding of a significant conditional indirect effect through risk compensation (moderated mediation), suggesting that among the subset of participants reporting low heterosexism, Black gay men were predicted to engage in lower risk compensation than the general public, which was associated with greater support for PrEP financial assistance.
This single observed indirect effect has limited practical utility since it only applies to participants reporting low heterosexism, a subset of the sample who expressed high support for PrEP financial assistance across PrEP beneficiary conditions and would therefore be a low-priority target for interventions seeking to generate public support around PrEP. The finding is nonetheless theoretically provocative when considered in the context of promiscuity stereotypes ascribed to Black MSM (81, 82), which would likely predict greater anticipated risk compensation among Black gay men vs. the general population (91) and a consequent reduction in support for PrEP financial assistance rather than the conditional indirect effect observed. It is possible that this conditional indirect effect is simply a manifestation of socially desirable responding, particularly given the absence of a significant indirect effect through risk compensation at higher levels of heterosexism. In other words, those participants who responded in the most socially desirable way on our heterosexism measure may also have been especially reluctant to endorse unfavorable judgments about a minority group. Nevertheless, further research is needed to characterize stereotypes of Black gay men more fully and a nuanced investigation of the impact of racial prejudice and sexual stereotypes on predictions surrounding risk compensation would help to clarify this unexpected finding.
Appendix 2
PrEP Attitudes Measure
Note: Italicized text was not visible to participants completing the measure.
Background Information
HIV Basics:
· HIV stands for human immunodeficiency virus.
· HIV is the virus that leads to AIDS.
· HIV can be spread through bodily fluids such as blood, semen (“cum”), and breast milk. Therefore, behaviors like having sex without condoms and sharing needles when injecting drugs puts individuals at risk for getting HIV.
· In the U.S., most new infections occur through having sex without condoms.
PrEP Knowledge/Experience Items
1. Prior to this study, had you ever heard of HIV pre-exposure prophylaxis, “PrEP,” or a daily pill that can be taken to help prevent getting HIV?
o Yes
o No
o I don’t know/I don’t remember
2. Prior to this study, had you ever received a prescription for HIV pre-exposure prophylaxis, “PrEP,” or a daily pill that can help prevent HIV?
o Yes
o No
o I don’t know/I don’t remember
Note: Item wording was varied by condition where indicated: General Population Condition=person/people, Gay Men Condition=gay man/men, Black Gay Men Condition=Black gay man/men. Participants were unaware of other conditions.
Additional Background Information
· Many [people/gay men/Black gay men] in the U.S. become infected with HIV every year.
· A daily pill called HIV pre-exposure prophylaxis (“PrEP”) has emerged as a new prevention strategy that has been recommended for individuals at high risk of getting HIV.
· PrEP is an antiretroviral medication that can be prescribed to HIV-negative individuals to prevent them from becoming infected with HIV. If they take PrEP once a day before they are exposed to HIV (such as through having sex with someone who is HIV-positive), the medication lowers their chance of getting infected.
continued
· Four large-scale clinical trials have shown that PrEP significantly reduces the likelihood of getting HIV, especially for individuals who take their pills on schedule without missing doses.
· In July of 2012, the U.S. Food and Drug Administration (FDA) approved a medication called Truvada® as the first PrEP medication, stating, "Truvada is approved for use as part of a comprehensive HIV prevention strategy that includes other prevention methods, such as safe sex practices, risk reduction counseling, and regular HIV testing."
· Doctors can now prescribe Truvada as a once-a-day pill to individuals who are at risk for getting HIV.
PrEP Attitude Items
Based on this background information, please rate your agreement with the following statements related to PrEP (1 pill/day of Truvada).
Response Options:
1 2 3 4 5
Strongly Disagree Strongly Agree
Note: Items preceded by asterisks were included in the original measure but not retained in scales based on the factor analysis.
continued
Scoring Instructions
1. Reverse score Items 6, 8, 12, 15, 16, 19, 24, 26, 27
2. Calculate scale scores by averaging across items:
· Respect for Taking PrEP: Items 9, 13, 15, 16, 19, 20, 24, 25, 27
· Support for PrEP Financial Assistance: Items 5, 8, 10, 12, 14, 17
· Predicted Risk Compensation: Items 3, 11, 21, 22, 23
· Perceived Community Benefit/Support for Access: Items 1, 4, 7
· Predicted Adherence: Items 2, 6, 26
Appendix 3
PrEP Attitudes Scales Determined through Exploratory Factor Analysis (n = 238)
Analytic Approach:
We performed an exploratory factor analysis with oblique (Promax) rotation and maximum likelihood extraction on the 28 PrEP attitude items. All participants from the larger study who completed all of the PrEP attitudes scale items and passed the attention/manipulation check (n = 238) were included, including participants who were excluded from the other analyses presented in this paper due to their involvement in a separate message-framing experiment. The PrEP attitude items that the message-framing experiment participants completed were phrased identically to items completed in the Gay Men Condition. The adequacy of our sample for this analysis was confirmed by a Kaiser-Meyer-Olkin (KMO) value within the “superb” range (KMO = .91) and all KMO values for individual items ≥ .73, well above the minimum acceptable value of .50 (92). Bartlett’s test of sphericity indicated sufficiently large inter-item correlations to support investigation of underlying factors, X2 (378) = 3996.49, p < .01. The eigenvalue (i.e., latent root) of a factor is the sum of squared loadings for that factor and is indicative of the amount of variance that the factor accounts for (93). An initial analysis of eigenvalues revealed five factors with eigenvalues greater than Kaiser’s criterion of 1, which is a commonly used cutoff for determining the number of factors to retain, with values above 1 considered to account for significant variance (93). These five factors accounted for 64.4% of the variance collectively. Scales corresponding to these five factors were constructed from items with factor loadings > .40 (93) that did not cross-load on other factors (see Table S1 below). These scales included Respect for Taking PrEP, Support for PrEP Financial Assistance, Predicted Risk Compensation, Perceived Community Benefit/Support for Access, and Predicted Adherence. For calculation of mean scale scores, items should be reverse-scored such that higher scores indicate greater respect for taking PrEP, greater support for PrEP financial assistance, greater predicted risk compensation, greater perceived community benefit/support for access, and greater predicted adherence.
Table S1. Exploratory Factor Analysis of PrEP Attitude Items
continued
[1] Indirect effect for low heterosexism reported at heterosexism score = 1.00 instead of 1 SD below the mean because the latter fell outside of score range