Women in this study were only asked about sex with men. Based on responses to these items, the computer-based interview asked pertinent questions about sexual

behaviour. Participants were asked to provide the number of times they had engaged in insertive or receptive vaginal or anal sex with HIV-infected partners, HIV-uninfected partners and partners of unknown HIV status. Participants were also asked about the click here number of times they had used condoms (male or female) from the beginning to the end of penetration and the number of times sex was unprotected. Unprotected sex was limited in the questioning to any act of insertive or receptive anal or vaginal intercourse in which a participant did not use a condom, a definition that excludes risk acts produced by accidental condom slippage or breakage. Our primary outcome variable was TRB and was defined as unprotected anal or vaginal sex with HIV-negative or status unknown partners. The variable itself was binary (yes/no). We used bivariate correlations and, where appropriate, crosstabs to assess the extent to which our data replicated previously established

bivariate TRB risk and protective factors. In addition, we ran bivariate analyses selleck kinase inhibitor on all of the nonscale items of the ACASI interview (i.e. all items except those that were part of the Treatment Optimism and Self-Efficacy scales) to determine if any individual questions were viable predictors of TRBs. Because the TRB outcome measure was dichotomous, we chose binary logistic regression for the multivariate modelling. In addition to the variables we planned to test for a relationship with TRBs (i.e. self-efficacy, treatment optimism, age, substance use, engagement with medical care, awareness of risky behaviours and education), we initially entered other variables with reliable (P<0.05) or suggestive (P<0.10) associations with TRBs. After building the initial model we then removed the variable with the weakest association and re-ran the analysis. This process was repeated until all predictors had estimates with P<0.10. The primary purposes of the bivariate analyses were to validate that the present sample

was not dramatically different from previously described samples (i.e. that we could replicate established bivariate relationships) and to generate candidates for our multivariate models beyond those we Nintedanib (BIBF 1120) intended to test a priori. Therefore, we did not correct for type I error and individual analyses should be interpreted with caution. For all of the analyses described below, positive correlations suggest more TRB and negative correlations suggest less TRB. We were able to replicate bivariate associations in the hypothesized direction for age (r=−0.28, P<0.0005), frequency of alcohol use in the past 3 months (r=0.11, P=0.07), any methamphetamine use in the past 3 months (r=0.25, P<0.0005), any nonprescription sildenafil use in the past 3 months (r=0.20, P=0.001), any cocaine use in the past 3 months (r=0.11, P=0.