Men who have sex with men (MSM) and transgender women (TW) are at increased risk of HIV infection worldwide, including those in low- and middle-income countries (LMIC) [1,2]. Stigma toward sexual and gender minorities is an important structural driver of the HIV epidemics among MSM and TW in LMIC, with evidence suggesting that such stigma is linked to increased sexual risk behaviors [3-6] and limits the provision and uptake of HIV prevention and treatment services within these socially marginalized populations [7-10]. Use of sex-seeking websites and mobile apps among MSM and TW has grown in popularity globally, with prevalence estimates ranging from 36% in Latin America [11], 44% in North America [12], 39% to 62% in Africa [13,14], and 41% to 77% in Asia [15-18]. This has sparked interest in and led to the development of interventions that harness these Web-based platforms to engage MSM and TW in safer sex practices [19], HIV testing [20,21], and HIV care [22]. Given that sex-seeking websites and apps allow MSM and TW to meet new sexual partners with more anonymity than physical venues where they may risk being outed, discriminated against, or harmed because of their sexual or gender identity [13,14], interventions delivered through Web-based sex-seeking platforms may be particularly effective for engaging MSM and TW in HIV prevention and treatment services in regions with widespread stigma toward these vulnerable populations [19-21].

Stigma stems from power structures that promote or maintain social inequity [23-25] and manifests at the structural, interpersonal, and individual levels [24,25] as prejudice and discrimination toward persons with socially devalued characteristics, such as sexual and gender minorities [23,24,26-28]. Sexual and gender minorities often face significant stigma in heteronormative societies where cultural values enforce traditional gender roles [23,26-29]. In these settings, sexual and gender minorities may experience discrimination [25,30], internalize or adopt negative feelings and shame about their own sexuality or gender identity over time [28], and anticipate future stigmatizing experiences [28,31], which may cause some to conceal their sexual or gender identity [32]. Although several exploratory studies in LMIC have examined the relationship between factors that shape or are influenced by stigma and Web-based sex seeking [11,13-18], findings are mixed regarding whether these Web-based platforms are being used by MSM who are more or less affected by stigma. Studies in Peru [11] and China [15,17,18] found that Web-based sex seekers were more likely to be MSM who identified as gay, which is more common among MSM less affected by stigma. Studies that measured a history of discrimination and perceived or anticipated stigma among MSM in Nigeria [14] and Vietnam [16], on the other hand, found that Web-based sex seeking was associated with a history of discrimination in Nigeria and perceived or anticipated stigma in both Nigeria and Vietnam, suggesting that Web-based sex seeking may be more common among MSM more affected by stigma. Furthermore, a study conducted among MSM in Swaziland and Lesotho [13] measured gay identification, history of discrimination, and perceived or anticipated stigma and found that all 3 were associated with Web-based sex seeking. However, few studies have examined how a range of factors that shape or are influenced by stigma are related to Web-based sex seeking, and less is known about the role of internalized stigma or social norms that perpetuate stigma toward sexual and gender minorities.

Mexico’s HIV epidemic is concentrated within key populations [33,34], including MSM and TW in the Mexico-United States border region. HIV prevalence among MSM and TW in Tijuana, which lies along Mexico’s northern border with San Diego, California, is estimated to be approximately 20%, with nearly 90% of those testing HIV positive reporting no prior knowledge of their HIV status [35,36]. Stigma toward sexual and gender minorities is common in Mexico [37-40] and may be partially attributed to the cultural norm of machismo [41-44], which defines rigid gender roles and reflects heteronormative expectations of male behavior [41,45]. Findings from a recent cross-sectional study in Tijuana suggest that MSM who are less out about their same-sex sexual behavior and have higher levels of internalized stigma are less likely to seek HIV testing, which may contribute to the high prevalence of undiagnosed HIV infection among sexual and gender minorities in this setting [46]. Given that MSM and TW most affected by stigma in Tijuana are least likely to undergo HIV testing, Web-based strategies may help increase their frequency of HIV testing and support their timely linkage to HIV prevention and treatment services. Web-based strategies may also represent a feasible method to increase uptake of these services, considering that 54% of adults in Mexico reported owning a smartphone or using the internet as of 2015 [47].

To assess the potential utility of this approach and inform the development of HIV prevention efforts for MSM and TW in Tijuana and other comparable low-income settings, we aimed to determine the prevalence of Web-based sex seeking and examine the effect of stigma on Web-based sex seeking among MSM and TW in Tijuana. We hypothesized that Web-based sex seeking would be more common among MSM and TW most affected by stigma toward sexual and gender minorities (ie, MSM and TW with greater endorsement of traditional machismo values, greater levels of internalized stigma related to same-sex sexual behavior or gender identity, and lesser outness about same-sex sexual behavior or gender identity and MSM who do not identify as gay or report a history of discrimination related to their same-sex sexual behavior).

Data for this analysis came from 2 studies conducted in Tijuana, Proyecto Enlaces (Links Project) and Proyecto Redes (Networks Project). Proyecto Enlaces was designed to compare the effectiveness of 2 recruitment methods—venue-based sampling (VBS) and respondent-driven sampling (RDS)—for the identification of undiagnosed HIV infection among MSM and TW. Proyecto Redes was embedded in Proyecto Enlaces and was designed to characterize the sexual networks of MSM and TW. Proyecto Enlaces was conducted between March 2015 and November 2018, whereas Proyecto Redes was conducted between March 2016 and September 2017.

VBS was performed using time-location sampling across 36 venues identified during formative research as locations frequented by MSM and TW in Tijuana (eg, nightclubs, bars, public spaces, and motels). Individuals identified via VBS were eligible for HIV testing if they were aged at least 18 years, cisgender male or transgender female, reported anal sex with a cisgender male or transgender female in the past 4 months, and did not report a previous HIV diagnosis. RDS is a chain-referral sampling technique often used to recruit hard-to-reach populations [48,49]. A total of 33 individuals identified through VBS or referrals from Tijuana’s municipal HIV clinic were selected to initiate RDS recruitment chains (ie, seeds). Seeds were further selected to be diverse with respect to HIV status, age, socioeconomic status, sexual orientation, gender identity, and recruitment venue. Individuals were eligible to be seeds if they were aged at least 18 years, cisgender male or transgender female, reported anal sex with a cisgender male or transgender female in the past 4 months, lived in Tijuana, and reported social networks that included at least 15 MSM or TW who also lived in Tijuana (changed to 5 MSM or TW in April 2017 to boost RDS recruitment). Seeds were given 3 coupons to recruit their MSM or TW peers (eg, sex partners, acquaintances, friends, and family members) to participate in the study. Peer recruits were then given 3 coupons to recruit peers themselves if they provided their referral coupon, were aged at least 18 years, were cisgender male or transgender female, and reported anal sex with a cisgender male or transgender female in the past 12 months. Those who did not report a previous HIV diagnosis were also eligible for HIV testing. Beginning in January 2018, seeds and peer recruits were given 6 coupons to boost RDS recruitment. A Microsoft Access database was used to track peer recruitment and store biometric information to prevent duplicate enrollment. Seeds and peer recruits were given Mexican pesos (Mxn) $100 (approximately US $5) for every eligible peer they referred to the study.

As individuals could be identified multiple times via the same or a different recruitment method, those who were identified more than once and who remained eligible for HIV testing were retested if it had been at least 3 months since their last test. Eligible individuals identified via VBS underwent rapid HIV testing (Advanced Quality HIV 1/2 Test Kits, Intec Products Inc) at recruitment venues or at the study site if they preferred, whereas those identified via RDS underwent rapid HIV testing at the study site. All rapid HIV test results and appropriate posttest counseling were delivered within a few days (VBS) or 20 min (RDS) at the study site. All rapid test–negative individuals identified via VBS and rapid test–negative individuals identified via RDS who reported anal sex with a cisgender male or transgender female in the past 4 months (for comparability with those identified via VBS) were offered enrollment in Proyecto Redes. Rapid test–positive individuals provided an additional blood sample for confirmatory testing via immunofluorescence assay at the San Diego County Public Health Laboratory and were offered enrollment in Proyecto Enlaces. Confirmatory HIV test results were delivered to rapid test–positive individuals within 2 weeks, and those confirmed HIV positive were referred for free HIV care at Tijuana’s municipal HIV clinic.

All study procedures occurred at the study site located in an unmarked office building staffed by local Spanish-speaking individuals with extensive experience working with sexual and gender minorities in Tijuana who were also trained to foster a nonjudgmental environment. In our previous research with key populations in Tijuana, our study site was recognized as a nonstigmatizing setting where participants felt comfortable visiting without fear of discrimination. All participants provided written informed consent, and all study procedures were approved by the Human Subjects Protection Committees at the Universidad de Xochicalco in Tijuana and the University of California, San Diego.

Surveys collected sociodemographic, psychosocial, and behavioral data and were interviewer administered using computer-assisted personal interviewing.

This analysis includes all Proyecto Redes participants (n=396; excluding 11 identified via RDS or VBS more than once who later tested HIV positive and enrolled in Proyecto Enlaces) and newly diagnosed HIV-positive participants enrolled in Proyecto Enlaces (n=165). Analyses considering 3 of the 5 stigma measures (ie, traditional machismo, history of discrimination related to same-sex sexual behavior, and sexual orientation) were restricted to specific subgroups of our sample because (1) HIV-positive Proyecto Enlaces participants completed baseline and supplemental (approximately 2 weeks post baseline when they returned for their confirmatory HIV test results) surveys, whereas HIV-negative Proyecto Redes participants only completed baseline surveys and (2) some survey questions were phrased for both MSM and TW, whereas others were only phrased for MSM. Data on traditional machismo and history of discrimination related to same-sex sexual behavior were collected via the supplemental survey among HIV-positive Proyecto Enlaces participants. As 27.3% (45/165) of Proyecto Enlaces participants did not return to complete the supplemental survey, analyses examining traditional machismo and history of discrimination related to same-sex sexual behavior were restricted to HIV-negative Proyecto Redes participants to reduce the potential for selection bias that could have been introduced by only including those Proyecto Enlaces participants who had returned. Analyses examining history of discrimination were then further restricted to MSM participants because the phrasing of the question was specific to experiences of MSM and not TW (Multimedia Appendix 1). Finally, analyses examining sexual orientation were restricted to MSM because we hypothesized that MSM and TW less affected by stigma would identify as gay or homosexual and heterosexual, respectively. Restricting these analyses to MSM allowed us to examine sexual orientation without potentially including participants who were more or less affected by stigma in the same category.

Descriptive statistics were calculated to characterize the study population by Web-based sex seeking. Next, bivariate associations between factors that shape or are influenced by stigma toward sexual and gender minorities were examined using linear and logistic regression models for continuous and categorical dependent variables, respectively. Separate unadjusted and adjusted logistic regression models were then used to examine the total effect of each stigma measure on Web-based sex seeking. A review of the literature was conducted to identify covariates that may confound the total effects of interest. Directed acyclic graphs (DAGs) [60] were then generated to depict interrelationships among each exposure of interest, Web-based sex seeking, and the identified covariates. Westreich et al have shown that it is inappropriate to interpret effect estimates for exposures of interest from a single adjusted model (ie, one that includes all exposures of interest) as their total effects on the outcome of interest because adjusting for an exposure of interest that lies on the causal pathway between another exposure of interest and the outcome of interest would yield an estimate of the direct effect of that exposure of interest on the outcome of interest [61]. Therefore, given the highly interrelated nature of our exposures of interest and the cross-sectional nature of our data (making it difficult to determine the directionality of the relationships among our exposures of interest), our DAGs considered 1 exposure of interest at a time to facilitate the identification of confounders for inclusion in adjusted models for the total effect of each exposure of interest on Web-based sex seeking. On the basis of these DAGs, sociodemographics (gender identity, age, education, and monthly income) and social support were identified as confounders and selected for inclusion in adjusted models. We were unable to stratify models by gender identity because of the small number of TW included in our sample (n=32). However, as controlling for gender identity in our adjusted models could have resulted in sparse data bias (because of the small number of TW), we conducted a sensitivity analysis excluding TW and found qualitatively similar results. In another sensitivity analysis, we examined the potential impact of grouping MSM who reported being unsure of their sexual orientation with those who did not identify as gay by excluding those who reported being unsure of their sexual orientation from models examining the effect of sexual orientation on Web-based sex seeking, and we found qualitatively similar results. Finally, we examined whether any of the relationships of interest were modified by HIV status or age by including product terms between the potential effect measure modifier and the exposures of interest in each of their respective models. Unstratified results are presented because none of the product terms were statistically significant. All statistical analyses were conducted using SAS V.9.3. (SAS Institute, Inc).

Approximately half of our sample was recruited via VBS (311/561, 55.4%; Table 1). Most participants identified as male (529/561, 94.3%), and participants had a mean age of 37 years (SD 11.2). Nearly half of the participants (239/561, 42.6%) reported at least a high school education, and 69.2% (387/561) of the participants reported an average monthly income of at least Mxn $3000 (approximately US $150). A total of 29.4% (165/561) of participants reported seeking sex partners on the Web in the past 4 months.

Bivariate associations between factors that shape or are influenced by stigma toward sexual and gender minorities were in the expected directions (Table 2). Positive associations were observed between traditional machismo and internalized stigma related to same-sex sexual behavior or gender identity (beta coefficient =.019; 95% CI 0.015 to 0.024). Outness about same-sex sexual behavior or gender identity was positively associated with gay identification (beta coefficient=2.35; 95% CI 1.96 to 2.73) and history of discrimination related to same-sex sexual behavior (beta coefficient =.93; 95% CI 0.42 to 1.43). Traditional machismo was negatively associated with outness about same-sex sexual behavior or gender identity (beta coefficient=−.04; 95% CI −0.06 to −0.03) and gay identification (beta coefficient=−.25; 95% CI −0.33 to −0.17). Internalized stigma related to same-sex sexual behavior or gender identity was inversely associated with outness about same-sex sexual behavior or gender identity (beta coefficient=−1.84; 95% CI −2.08 to −1.59) and sexual orientation (beta coefficient=−8.94; 95% CI −10.25 to −7.63).

On average, Web-based sex seekers reported less endorsement of traditional machismo values (mean 1.9 vs mean 2.2, range 1 to 4; P<.001; Table 1), less internalized stigma related to same-sex sexual behavior or gender identity (mean 20.5 vs mean 25.3, range 9 to 45; P<.001), and they were more out about their same-sex sexual behavior or gender identity (mean 5.1 vs mean 3.8, range 1 to 7; P<.001). Compared with MSM participants who did not seek sex partners on the Web, a greater proportion of Web-based sex seekers identified as gay (61.9% vs 28.3%, P<.001) and reported a history of discrimination related to their same-sex sexual behavior (46.0% vs 35.2%; P=.05).

After adjusting for sociodemographics and social support (Table 3), the odds of Web-based sex seeking were lower for participants who reported greater endorsement of traditional machismo values (adjusted odds ratio [AOR] 0.36, 95% CI 0.19 to 0.69) and greater levels of internalized stigma related to their same-sex sexual behavior or gender identity (AOR 0.96, 95% CI 0.94 to 0.99), whereas the odds of Web-based sex seeking were higher for those who reported greater outness about their same-sex sexual behavior or gender identity (AOR 1.17, 95% CI 1.06 to 1.28). Among MSM participants, the odds of Web-based sex seeking were higher for gay-identifying participants (AOR 2.13, 95% CI 1.36 to 3.33) and those who reported a history of discrimination related to their same-sex sexual behavior (AOR 1.83, 95% CI 1.08 to 3.08).