This study was approved by the Shanghai Medical Ethics Committee and the Shanghai Municipal Health Bureau. All of the participants were informed of the study objectives and provided consent.
Sample and selection criteria
Shanghai is one of the largest metropolitan areas in China, with a 40% itinerant population in whom HIV/AIDS transmission is largely attributed to heterosexual contact. Because provision of commercial sexual services is illegal in China, and most FSWs generally operate covertly, it is impossible to estimate the actual number of FSWs operating in Shanghai. Non-governmental organizations and community hospitals are the best links to the FSW population, and we relied on their knowledge to conduct our cross-sectional survey by cluster sampling. Between July and December 2009, 20 geographic sites among 5 districts were identified randomly where access to FSWs was considered feasible. One hundred thirty-two small sex establishments, including Xitou Fang, massage parlors, and hair salons from 20 geographic sites of 5 districts which explicitly provided sexual services, were enrolled in the study. We visited every hotspot that provided commercial sexual services and interviewed the FSWs working there. Currently, there are only 4–6 FSWs working in each hotspot. The participant selection criteria included FSW migrants from the mainland who reported that at least one of their last three sexual partners had paid to go out on a date.
Sample estimation
By conducting a detailed review of the literature and performing a small scale pre-survey with a limited portion of the FSWs of the itinerant population in Shanghai, we found that risky behaviors, such as unprotected sexual intercourse, were practiced by 34%-40% of FSWs [17]. By taking a cross-sectional sample size estimation approach, the error was calculated as follows: ϵ = 0.15P; P = 37%; Q = 1-P =63%; and α = 0.05 Z
α
= 1.96. The sample size calculation method used was as follows:
The intra-cluster correlation coefficient (ICC) of the primary outcome is known to play a key role in the design and analysis of cluster randomized trials, in which clusters, such as health care organizations, school classes, or geographic areas, are randomized to trial arms and outcomes are measured for individuals within those clusters [18]. The ICC for condom use ranged from 0.01-0.08 in other studies [19, 20]. Because it is very difficult to obtain a sufficiently large sample population of FSWs in China, we used the average ICC (0.045) to adjust the sample size, as follows: m = N/layers = 303/20 = 15, where
Our final sample population was comprised of 504 FSWs from the Shanghai district hotpots located in 20 geographic sites.
Data collection
Researchers from the School of Public health of Shanghai Jiaotong University surveyed the FSWs of small sex establishments using structured questionnaires between May 2010 and January 2011. Individuals completed the questionnaires by interview-led surveys to collect detailed information about condom use and specific knowledge, perceptions, attitudes, behaviors, and the physical, social, and policy environment related to condom use and HIV/AIDS prevention. Prior to participation, we explained the aim and major content of the survey to each of the FSWs and emphasized that participation was voluntary and anonymous. The procedure lasted approximately 30 minutes and was conducted face-to-face in a private room. Each participant was compensated 50 RMB ($7-8 USD) in cash.
Measure
The questionnaire was developed by referring to the entertainment guide to HIV/AIDS prevention in the China and World Health Organization guidelines. A preliminary research showed good fitness of the questionnaire by reliability and validity analyses. The questionnaire consisted of socio-demographic characteristics of the participants (the age, hometown, year school was completed, civil status, monthly income, age of sexual debut, age at which the participant started selling sex, duration of sex work, and average client dates per week),predisposing factors of condom use (perception of susceptibility to HIV/AIDS , perception of benefits of condom use, and safe sex self-efficacy), and enabling and reinforcing factors, which are usually environmental-structural factors. The method by which we scored the constructs of perception of susceptibility, condom use, and safe sex self-efficacy was based on the Thurstone scale [21].
The perception of susceptibility to HIV/AIDS and benefits of condom use were used to measure the individual perception of susceptibility to HIV/AIDS as follows: “everyone can be infected by HIV;” “HIV-infected people and ordinary people are difficult to separate;” “HIV can be spread through sexual transmission;” and “HIV also can be transmitted by oral sex.”. Correct answers were credited with a score of one, while incorrect answers or responses of “do not know” received a score of zero. The sum of the score of each question was converted into a total score, with a maximum of 4 (Cronbach’s alpha, 0.81). Another four items related to condom use during sexual intercourse were also used to measure the perception of benefits of condom use by the participants, and the sum of the score of each question was converted into a total score, with a maximum of 4 (Cronbach’s alpha, 0.75).
The safe sex self-efficacy was investigated as follows: CCU with clients when sexually excited; reject clients who looked clean and offered more money not to use condoms; reject clients who refused to used a condom; use condoms with intoxicated clients. The sum of the score of each question was converted into a total score, with the maximum of 4 (Cronbach’s alpha, 0.63).
Environmental-structural support was enabling and reinforcing factors supporting condom use in the establishment as follows: the perceived level of safe sex information exchange among employees; support from the establishment owner about the important of condom use during commercial sexual services; accessibility of condoms in the establishment; accessibility of HIV voluntary testing performed by government hospitals or the CDC; support from community physicians about health advise; health education and health promotion on HIV/AIDS from regular intervention for high-risk population; and getting free condom from family planning departments ( Cronbach’s alpha, 0.71).
Consistent condom use was measured using a five-point Likert scale, as follows: “Have you always, almost always, sometimes, almost never or never use condoms during commercial sexual service (the primary dependent variable of our study).” FSWs were considered to use condoms consistently if they answered “always” compared to inconsistently, which was defined as “almost always, sometimes, almost never and never” [22].
Data entry and analysis
Data were double-entered using Epidata 3.0 software. All statistical analyses were performed using the Statistical Package for Social Sciences (SPSS) for Windows. Generalized assessments were made using the mean, standard deviation, and percentages. A chi-square test was used to compare differences in socio-demographic characteristics between the suburbs and central city. For multivariate analysis, the continuous independent variables were examined for normalcy and were categorized by the median. The dependent variable (condom use) was dichotomized into consistent (always) versus non-consistent (almost always, sometimes, almost never, or never) and a binary logistic regression was used to determine the predictors of CCU. To address potential collinearity among the variables, we added the variables to the regression model at a time by a forward stepwise approach, and the significant level was 0.05.