2.1 Participants
Between September 2017 and January 2018, a baseline survey of a four-year cohort study was conducted in three provincial capitals of China: Wuhan, Nanchang, and Changsha. The cohort study aims to examine the effect mechanisms of perceived stigma on risk behaviors, psychology and HIV infection among Chinese MSM. Participants were recruited through several ways (routine HIV testing services, peer recommendation, and outreach activities), with the help of the local Centers for Disease Control and Prevention (CDC), and MSM organizations (Tongxing organization, Xingyuan organization and Qingcai organization). The inclusion criteria for the study were: (1) they were 16 years or older; (2) they had experience of having sex with men; (3) they were sexually active during the past 6 months; (4) they self-reported no HIV infection; (5) they provided written informed consent. The participants were asked to complete the electronic questionnaire on tablets (e.g. iPads) on their own in a quiet room in order to protect the individual privacy, and the trained students or volunteers were available to provide clarity if participants were unsure what the question was asking. All participants who finished the survey received 50 RMB (approximately US $7) as compensation for their time in this baseline survey. This study was approved by the Medical Ethics Committee of Wuhan University, China.
2.2 Data Collection
Socio-demographic Characteristics
Socio-demographic information included: age, ethnicity (Han, others), current education level (high school or lower, college, graduate or higher), current employment (employed, unemployed), current marital status (unmarried, married, divorced), and monthly income (<1000 Yuan, 1000-3000 Yuan, 3001-6000 Yuan, >6000 Yuan). Sexual orientation was also measured by a single item question, “what’s your sexual orientation?”, and participants could choose from four options, “gay”, “bisexual”, “heterosexual”, and “unsure”.
Gender identity
Gender identity (the primary “exposure” variable for the analysis) was measured by a single item, in order to identify transgender individuals, the question was asked “Do you agree with your sex assigned at birth?”, with three options included: “yes”, “no”, and “unsure”. The questions have been widely used both in China and other countries [7, 20]. In the study, we divided the MSM population into three categories according to gender identity, including cisgender MSM (those who agreed with their birth-assigned sex), transgender MSM (those who disagreed with their birth-assigned sex), and gender non-conforming MSM.
Psychosocial characteristics
The Center for Epidemiological Studies Depression (CES-D) was used to screen the symptoms of depression during the past week [21]. The CES-D has 20 items rated on a 4-point Likert scale ranging from “rarely or none of the time” (0) to “most or all of the time” (3). A higher total score indicates more depressive symptoms. The Chinese version of CES-D has been validated [22, 23] and the Cronbach’s alpha was 0.92 in this research.
The Multidimensional Scale of Perceived Social Support (MSPSS) was used to measure social support [24]. The scale consists of three subscales and each subscale includes four items. Response option for each item ranges from 1 (very strongly disagree) to 7 (very strongly agree). A higher total score indicates greater perceived social support. This scale has shown a good validity and reliability in the Chinese population [25]. The Cronbach’s alpha was 0.94 in this research.
The 10-item Connor-Davidson Resilience Scale (CD-RISC10) was used to measure resilience [26]. Responses were rated on a 5-point Likert scale ranging from “not true at all” (0) to “true nearly all the time” (4). A composite score was calculated, with higher total score indicating higher capacity of resilience. This scale has been confirmed validated in Chinese adults [27, 28] and the Cronbach’s alpha was 0.87 in this research.
Identity concealment was measured by a six-item, 5-point Likert subscale. This subscale was based on nondisclosure subscale developed by Testa [29] and further modified by Outland [30], which was suitable for LGBT population [30]. Items were summed to yield a composite scale score, with a higher score suggesting a higher likelihood of sexual orientation concealment. It has been confirmed validated among Chinese adults [31] and the Cronbach’s alpha was 0.91 in this study.
HIV-related sexual behaviors
Information on HIV-related risky sexual behaviors for the past six months was collected, including multiple sexual partners (2 or more were classified as having multiple sexual partners), male partner types (all were regular partners/acquaintances, all were one-night stand/occasional partners, or both regular partners/acquaintances and one-night stand/occasional partners), sex after drugs (yes or no), sex after drinking (yes or no), commercial sexual behavior (yes or no). Participants were also asked to answer frequency of condom use during anal sex in the past six months. The options included “never”, “sometimes”, “often”, “almost always” and “every time”. In data analysis, “never” “sometimes”, “often”, “almost always” were classified as inconsistent condom use.
2.3 Statistical Analysis
All analysis of data was conducted using SPSS19.0. The prevalence of gender minorities and 95% confidence interval (95%CI) were described. The Chi-squared tests and Fisher’s exact test were performed to explore the differences of socio-demographic characteristics and HIV-related risky sexual behaviors by gender identity. In terms of psychosocial variables, multiple factor variance analysis was used to explore the relationships between different gender identity groups, which was adjusted for age, monthly income, and sexual orientation based on previous research [32-34]. The Bonferroni correction method was used to conduct a post-hoc analysis. Binary logistic analysis was used to explore how gender identity predicted HIV-related risky sexual behaviors (dichotomous variables, e.g., multiple sexual partners), and the Hosmer-Lemeshow test was used to assess the binary logistic regression. Multinominal logistic analysis was used to assess the relationship between male partner types and gender identity, and -2Log likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) were used to evaluate the model. In both binary logistic regression and multinomial logistic regression, gender identity was coded as independent variable, HIV-related risky sexual behaviors in the past six months were coded as dependent variables, and socio-demographic and psychosocial characteristics as covariates. The level of P<0.05 (two-sides) was set for statistical significance in this study.