Subjects were recruited in Shenyang (provincial capital of Liaoning Province), China between April 2017 and July 2017. According to the sixth national population census of the People’s Republic of China , Shenyang is the largest city in Northeast China by urban population, with 8.1 million resident population. The number of MSM population in Shenyang ranks the second in China with an estimated 140,000 MSM in 2006 . MSM population accounted for 81.2% of the city’s annual newly reported HIV cases , while the national average proportion is only 28.3% .
The inclusion criterion were: 1) older than 18 years; 2) male birth assigned sex; 3) self-identified as woman; 4) reporting having sexual intercourse experiences in the last 6 months; 5) agreeing to participate in the study and willing to provide written informed consent.
Snowball sampling [30, 31] was used for recruitment because TGW in China are hard to reach and hidden due to the universal discrimination . And in the absence of a sampling frame, it was not feasible to perform probability sampling. The participants were recruited through a non-governmental organization (NGO), China Love Aid, which dedicated to improving the physiological and psychological health of TGW. With the assistance of the NGO, five participants were recruited as the ‘seeds’. The five seeds then each recruited a maximum of five participants, thus, aggregating the snowball until saturation reached. All eligible participants were asked to read and sign informed consent documents for voluntary participation after being informed of the following: all data collected were anonymous and strictly confidential and would be used only for research; refusal would not affect their right to use any service; they could quit at any time without being questioned.
A total of 198 eligible TGW provided written informed consent and participated in this study. They were then asked to complete the questionnaires during a face-to-face interview in a private room by trained investigators, who were staff of the NGO, to ensure the quality and completeness of the response. The questionnaires were anonymous and took approximately 30 minutes to complete. All subjects were paid 200 CNY (about 30 USD) as remuneration after completion of the above procedure. The ethics committee of the School of Public Health, Shanghai Jiao Tong University, approved the recruitment procedure and the study protocol.
The following background characteristics were collected, including age, educational level, marital status, monthly income, and length of residence in Shenyang.
To assess knowledge deficits and misconceptions about HIV risks and self-protective behaviors, the 18-item true–false Brief HIV Knowledge Questionnaire  (Cronbach’s α = 0.683) was administered. The HIV knowledge score was obtained by summing the number of correct responses. Higher scores indicated greater information.
Motivation to engage in condom use was measured in accord with the principal constructs of the Theory of Reasoned Action  and the Health Belief Model . Two scales were developed by the authors to provide two separate indicators of motivation. A total score was obtained by summing scores for individual items. Higher scores on the two scales indicated stronger motivation.
The 6-item, Likert type Attitude Scale (Cronbach’s α = 0.528) assessed subjects’ attitudes toward condom use. For favorable statements (“I think condoms can prevent the transmission of HIV”; “I think condoms should be used during anal intercourse”; “I will try to persuade my partner into using condoms”), the strongly agree response is given a weight of 5 and the strongly disagree response a weight of 1, with other responses weighted accordingly. For unfavorable statements (“I think using condoms during anal intercourse is uncomfortable for both partners”; “I think condoms are unreliable in preventing HIV”; “In my opinion, condoms are too much trouble”), the scoring is reversed.
The 4-item Subjective Norms Scale (Cronbach’s α = 0.806) assessed perceptions of what others (family members, friends, peers who are engaged in the same work, and significant others) think about the participants’ consistent condom use during anal sexual intercourse, where 1 = strongly disapprove, and 5 = strongly approve.
To assess subjects’ HIV-preventive behavioral skills, the 6-item Condom Influence Strategy Questionnaire (CISQ-S)  (Cronbach’s α = 0.873) and the 6-item Self-Efficacy Scale (Cronbach’s α = 0.874) developed by the authors were administered. Each scale represented a separate indicator of behavioral skills. Higher scores on both scales indicated better behavioral skills.
The CISQ-S assessed subjects’ condom use negotiation skills and unprotected sex refusal skills. Questions were answered in a Likert format that ranged from 1 = never to 5 = always. An example item from the scale is “I would like to discuss condom use with my partner”.
The Self-Efficacy Scale assessed their ability to engage in condom use effectively by these questions: “I can carry a condom with me in case I need it”; “I can use a new condom every time I have anal sex”; “I can use condoms throughout my anal sex”; “I can avoid the slipping of the condom during anal sex”; “I can discuss condom use with my partner even after drinking alcohol”; “I can negotiate using condoms with my partner no matter who he is”. Participants answered on a 5-point Likert scale that ranged from 1 = definitely can’t do to 5 definitely can do.
The questionnaire contained three frequency rating items (all scaled from 1 = never to 4 = always) to assess subjects’ condom use with their regular sexual partners (defined as those who were in a stable relationship such as boyfriends that did not involve transactional sex), casual sexual partners (defined as those who were not regular partners where transactional sex was not involved), and transactional sexual partners (defined as those with whom transactional sex was involved) during the last 6 months. Similar definition has been widely used in other studies (e.g. ). To compare background characteristics between subjects with and without CCU, a dichotomous variable were created, with 1 indicating always using condom with all types of sexual partners, and 0 indicating not always using condom with at least one type of partners. Besides, a total condom use frequency was obtained by calculating the average frequency of use with the three types of sexual partners. Higher scores indicated more frequent condom use.
Descriptive analyses were conducted using IBM SPSS Statistics 23.0 (IBM Corp., Armonk, NY, USA). Comparisons were conducted between subjects showing CCU and those not showing CCU to detect potential risk factors using Student’s t-test and Pearson’s chi-square test. Spearman’s correlation analysis was conducted to examine correlations among IMB model constructs. Confirmatory factor analysis (CFA) and structural equation modeling (SEM) were conducted using IBM SPSS AMOS 21.0 (IBM Corp., Chicago, IL, USA). The CFA measurement model included two latent constructs (motivation, behavioral skills) that predicted its indicators (motivation: condom use attitude, subjective norms; behavioral skills: condom use skills, self-efficacy). Once an acceptable measurement model had been established, a structural equation model was built to examine correlates of condom use among TGW based on the IMB model. Model fit was assessed using the ratio of chi-square values to degrees of freedom (χ2/df), the comparative fit index (CFI), and the root mean square error of approximation (RMSEA) . A χ2/df ratio of 3 or less, a CFI greater than 0.90, and an RMSEA lower than 0.08 indicated acceptable model fit [38, 39].