HIV transmission in Ukraine is driven in part by unsafe injection drug use and sexual risk behaviors among people who inject drugs. We performed a random-intercept latent transition analysis on responses to 9 binary injection drug use and sexual behavior items from 1195 people who inject drugs with negative HIV status enrolled in a clustered randomized clinical trial of a social network intervention in Odessa, Donetsk, and Nikolayev, Ukraine. We identified 5 baseline classes: "Social injection/equipment-sharing" (11.7%), "Social injection" (25.9%), "High-risk collective preparation/splitting" (17.0%), "Collective preparation/splitting" (11.3%), and "Dealer-facilitated injection" (34.1%). After 12 months, intervention participants were more likely to transition to the "Collective preparation/splitting" class, which featured the fewest risk behaviors. Transitioning from the "Collective preparation/splitting" to the "Social injection/equipment-sharing" class was associated with HIV acquisition for control participants. Research to illuminate the stability of these patterns and how they may benefit from uniquely tailored programming to reduce unsafe behaviors is needed.

Shame may increase HIV risk among stigmatized populations. The Personal Feelings Questionnaire-2 (PFQ-2) measures shame, but has not been validated in Spanish-speaking or nonclinical stigmatized populations disproportionately affected by HIV in resource-limited settings. We examined the psychometric properties of the Spanish-translated PFQ-2 shame subscale among female sex workers in two Mexico-U.S. border cities. From 2016 to 2017, 602 HIV-negative female sex workers in Tijuana and Ciudad Juarez participated in an efficacy trial evaluating a behavior change maintenance intervention. Interviewer-administered surveys collected information on shame (10-item PFQ-2 subscale), psychosocial factors, and sociodemographics. Item performance, confirmatory factor analysis, internal consistency, differential item functioning by city, and concurrent validity were assessed. Response options were collapsed to 3-point responses to improve item performance, and one misfit item was removed. The revised 9-item shame subscale supported a single construct and had good internal consistency (Cronbach's α = .86). Notable differential item functioning was found but resulted in a negligible effect on overall scores. Correlations between the revised shame subscale and guilt (r = .79, p < .01), depression (r = .69, p < .01), and emotional support (r = -.28, p < .01) supported concurrent validity. The revised PFQ-2 shame subscale showed good reliability and concurrent validity in our sample, and should be explored in other stigmatized populations.