Online survey
In partnership with a local community-based organization (CBO) and the Guangzhou Center for Disease Control and Prevention (CDC), we conducted a cross-sectional online questionnaire survey with 777 MSM in Guangzhou, China in September 2018. The survey was distributed online to MSM through CBO and CDC social media accounts. Eligibility criteria included the following: being biologically male at birth; being 16 years old or above; reported ever having oral or anal sex with men; residing in Guangzhou in the past six months. All survey data were anonymous and confidential, and online consent was obtained prior to the survey. Each man who participated received either 7.5 USD (50 Chinese Yuan) or a free HIV self-test kit as an incentive to participate.
Survey instruments
We collected information about participants’ sociodemographic characteristics including age, residence permit, occupation, heterosexual marital status (never married, engaged or married, and divorced/separated/widowed), annual income, highest education obtained (high school or less, some college, university, and postgraduate), gender identity (male, female, transgender, and unsure), sexual orientation (gay, bisexual, heterosexual, and unsure) and sexual orientation disclosure to people other than their partner(s) (yes/no).
Homoprejudiced violence questionnaire
Twelve homoprejudiced violence survey items were designed based on previous literature (5, 21, 22). We selected 12 items to cover four domains – physical assault, verbal aggression, psychological abuse, and cyber violence (Supplementary file 1). We translated and adapted the 12 items in order for them to be relevant to Chinese men. These 12 items asked whether a participant had ever experienced any of the following due to their sexual orientation: being gossiped about, being name called, being deliberately alienated or isolated, being threatened, being maliciously called gay, being spat on, having personal belongings damaged, being deprived of economic resources or personal belongings by someone (including family members), having personal freedom restricted by someone (including family members), being physically harmed (such as being slapped, beaten or kicked), being harmed on social media (such as WeChat and Weibo, the Chinese substitutes of WhatsApp and Twitter), and being harmed through phone calls or messages. The items were field tested with 10 participants and minor amendments were made for better clarity.
All 12 items used three responses: “yes”, “no” and “do not want to tell”. A new summative variable was generated by adding up the responses (“yes” were coded as 1, “no” or “do not want to tell” were coded as 0) of the 12 items to assess the overall prevalence. The summed value 0 was recoded as 0 (no prior experiences of homoprejudiced violence), and the summed values 1 to 12 were recoded as 1 (prior experiences of homoprejudiced violence of any type) (outcome 1). Additionally, one follow-up item asked whether participants had ever committed any of the 12 violent behaviors aforementioned against others due to their sexual orientation (yes, no, do not want to tell) (outcome 2). The Cronbach alpha value was 0.89.
Data analysis
Descriptive analysis was used to describe sample characteristics, including sociodemographic backgrounds and frequencies of violence experiences. The two outcomes were dichotomized (with “no” and “do not want to tell” grouped together) in regression analyses. We conducted univariate and multivariable binary logistic regressions to examine sociodemographic factors associated with homoprejudiced violence. We reported odds ratios and 95% confidence intervals (CIs). Data were analyzed using SPSS, version 25.