China has experienced a drastic change in attitude toward sex in the past 3 decades. Concurrently, sex-related health issues, especially sexually transmitted infections (STIs) including HIV/AIDS, have become public health problems in China. Female sex workers (FSWs) should be considered a priority for prevention of HIV/STI.1
Multipurpose barrier methods have been the most effective preventive devices to date, and the most well known of them is the male condom (MC). However, MC may not be suitable or usable in many circumstances, especially for FSW.2 Although MC has been promoted among FSW for more than a decade in China, there remains much room for further improving their protected sex. Data from Chinese national HIV surveillance indicated the median proportion of self-reported 100% MC use with clients in the last 30 days was approximately 60% to 74% from 2004 to 2008 among FSW across 15 surveillance sites.3 Thus, MC-only intervention is insufficient for HIV/STI prevention.
Another barrier technology for dual protection, the female condom (FC), now with several products available with similar efficacy to MC,4,5 should be given greater consideration. Numerous clinical trials have shown that FC was effective in increasing protected sex acts and decreasing STI incidence among women6; however, FC is still underused in the real world nearly 2 decades after first introducing the product into the market.5,7 Although FC education, skills training, and counseling, accompanied with supply, are primary strategies used in FC promotion,8,9 these strategies produced very different acceptability and use in different settings. For instance, acceptability studies in family planning and STI clinic settings have reported high rates of using FC at least once, ranging from 39% to 78%, and their results suggested that FC adoption was associated with individual-level characteristics such as age, education, and sexual relationship.10–13 However, these studies might be insufficient to understand FC use in natural settings because they did not analyze the contextual factors that might have an impact on FC promotion. The exploration of these factors could help improve program development for FC interventions in the real world and lead to further success in increasing protection during sexual activities.
Very few FC studies have been conducted in China. An early intervention study with 330 FSW revealed that as many as 94% of participants liked FC very much.14 Previously, we reported on FC acceptability identified in the first 3 sites of our study15 and on preliminary analysis of factors related to FC adoption in the first 2 study sites.16 Here we will further explore the individual and contextual factors related to FC use in all 4 study sites of the project.
MATERIALS AND METHODS
Settings of Study Sites
Our project17 was conducted sequentially in 4 study sites in southern China, first in 2 rural towns within the same county in Hainan, then in a small urban center in Guangxi, and, finally, in a midsized city in Hainan, which are referred to in this study as FS, YF, PX, and QH, respectively.
These sites were selected primarily because they have had a well-established sex industry for approximately 2 decades, and sex establishments there are typical of the sex industry in rural and small urban areas in China. Generally, the 5 types of sex establishments in these sites are as follows: (1) roadside brothels (also referred to as roadside restaurants),15,16,18 which were found only in FS and QH and provide only sex services; (2) hairdressers/massage parlors; (3) boarding houses; (4) hotel-based massage, beauty, and sauna parlors, where the women might pick up clients by calling hotel guests, which were only found in PX; and (5) nightclubs, only in QH, where entertainment such as karaoke was offered, often accompanied with alcohol. The size of establishments varied within and across different types in the 4 sites. For example, the number of women in hairdressers/massage parlors varied from a handful in some towns to 30 to 50 women in one parlor in QH. Nightclubs were also large venues with 20 to 50 women. Of the 5 types of venues, all but boarding houses were operated with bosses or managers. Women in boarding houses do freelance business.
The HIV/AIDS epidemic and history of previous prevention projects were very different among these sites. The first HIV/AIDS case was discovered among injecting drug users in PX in 1996. Since then, 499 cumulative cases have been identified there through 2009. In recent years, approximately 60% to 70% of new HIV cases were contracted through heterosexual contact.19 HIV prevalence was relatively low in Hainan Province, although other STIs are of public health concern. For instance, only 24 HIV/AIDS cases have been detected in QH since the late 1990s; however, reported syphilis prevalence was 8% among FSW in 2005 to 2007.20 In the county where FS and YF are located, a handful of HIV/AIDS cases were reported, but none of them were from these towns.
Our project worked with different local organizations in the 4 sites. In the rural towns, we worked with the township hospitals. In the 2 urban sites, we collaborated with the municipal Centers for Disease Prevention and Control (CDC) in PX and a public clinic specializing in STI in QH. Compared with the other 3 collaborating organizations, PX CDC had significantly more experience with HIV/STI prevention projects for FSW, drug users, and HIV-infected people since the late 1990s. Although prevention projects for FSW in FS also started in the mid-1990s, the local township hospital had changed its staff, organizational duties, and administration when our project started. Relatively few projects have been conducted in YF and QH. In particular, local QH staff had had very little outreach experience at the beginning of our project.
No FC had been introduced, and no women reported having used it in the 4 sites before our project started, although a few women had heard of FC by name.
Design of the Project
The FC intervention was designed as a 2-phase program with a first 6-month phase of intensive intervention and a second 6-month phase of maintenance. Before the intervention, 4 to 6 months of formative ethnographic research and then a baseline survey were carried out. Another 2 cross-sectional surveys were conducted after each intervention phase. The project followed this timeline in the first 3 study sites (FS, YF, and PX). However, in QH, it took much longer for the ethnographers and local staff to build rapport with women in the sex venues to conduct formative research and the baseline survey. Therefore, we only completed the first phase of intervention and 1 postintervention survey there in the 18 months.
The intervention was conducted according to a standard protocol mainly through outreach to establishments by trained local staff. This protocol required outreach staff to teach and demonstrate FC use in the establishments, often using a flip chart and a plastic vagina-uterus model as aids, then to provide 4 free FCs to all women and replenish their supply on the basis of need. In the first 3 sites, the project provided women with the first generation of FC produced by the Female Health Company, but the second generation of FC by the same company (i.e., FC2) was provided in QH. We will not distinguish them in this study because the 2 products have the same design and similar properties. Other relevant information on HIV/STI prevention and free MCs were also delivered during outreach intervention.
Based on ethnographic mapping of sex establishments, there were approximately 100 FSW in each of FS and YF, 200 in PX, and 800 in QH. Owing to the limited local staff in QH, the project selected 2 areas in the vicinity of the local collaborative clinic as the intervention target neighborhoods, which included the full array of sex establishments, a total of 18 establishments with 200 to 250 women.
All cross-sectional surveys were designed to recruit 70% to 80% of the total estimated women working in the targeted sex establishments. The number of participants from each venue type was roughly proportional to the estimated number of women in the different types of establishments at each study site.
Women working in the targeted establishments who were 16 years or older and sexually active in the prior 30 days were eligible for the survey at any time point. Women were also eligible in both postintervention surveys if they could show a project card with a unique personal ID code that was written by an interviewer from a previous survey. Recruitment was conducted by local staff. Eligible women were invited to a private room where formal informed consent form was completed before administering a face-to-face interview using a structured, anonymous questionnaire. Survey participants received a noncash gift worth about RMB80 yuan (about US $13). The protocols for this study received full review and approval by institutional review boards at both the Chinese and the US research institutions.
In this study, the primary comparison was conducted between women who used or never used FC after intervention in each of the 2 postintervention surveys. Therefore, we only included study participants who had heard of FC at those 2 time points. For these analyses, in each postintervention survey, eligible women were categorized into 3 groups of FC use according to self-reported number of times they had ever used FC. “Nonusers” referred to the women who had heard of FC but never used it; “1-time users” were defined as the women who reported having used FC only once; and “multitime users” were those who reported using FC twice or more times.
The following domains were also measured: demographics, reproductive history, sexual relationships and activities in the last 30 days, attitudes about FC and HIV/STI risk, and indications of prior research participation and intervention exposure. Sexual relationships were categorized into 3 types according to whether the participant had a primary or paying male partner(s) or both. The proportion of protected sex was calculated as the number of sexual acts using MC or FC divided by the total number of sexual acts in the last 30 days. Attitude measures included willingness to use FC, a 17-item FC beliefs and attitudes scale measuring responses to perceived FC characteristics (range, 1–4; higher score is more positive attitudes, negative items reverse coded; Cronbach α = 0.77),21 and worries about HIV/STI infection (range, 0–2). An FC intervention exposure score was constructed to measure involvement in our intervention. The participant would get one point on this score for each of the following 4 items to which she answered yes: (a) knew the project by name; (b) recognized the flip chart used during the intervention to illustrate FC use; (c) learned FC insertion in a vaginal model, which health care workers encouraged the women to do during outreach; and (d) practiced FC insertion by herself.
Participant baseline characteristics were described by study site. Then women’s characteristics and intervention exposure were compared among the 3 subgroups of FC users from all study sites combined, that is, nonusers, 1-time users, and multitime users at each of the 2 postintervention surveys. Next, 2 polytomous logistic regression models were used to assess the independent contributions of study site, establishment types, and other individual factors to FC 1-time use or to FC multitime use, both with nonusers as the reference group, for both the first and second postintervention surveys. The models, both using the same group of explanatory variables, were built based on theoretical considerations regardless of the degree of bivariate association. Three dummy variables of study sites were entered into the model with YF (a small sized site with less experienced staff) as the reference group. A dichotomous variable of boarding house versus nonboarding house was used as the type of establishment, which generally distinguished freelance women from nonfreelancers. Age was dichotomized into women 30 years or older and women younger than 30 years because older women were more likely to be married, have children, and work in the boarding houses. To understand the relation of MC use to FC use, a dichotomous variable of MC use in the previous 30 days was included in the model. Because the median proportion of sex protected by MC was 87% in the first postintervention and 94% in the second postintervention survey, we use 80% as the cutoff point for this variable. Finally, the 4-point exposure score was included as a continuous variable.
It should be noted that analysis of the second postintervention survey included only data from 3 sites because only the first postintervention survey was completed in QH. All statistical tests were only done across subgroups within each cross-sectional survey with conventional statistical tests. We did not do statistical comparisons across surveys because approximately 40% to 60% of the women reported participating in a previous survey, which made cross-survey samples related to each other. Statistical significance was accepted at P < 0.05. All analyses were conducted using SAS9.2 software (SAS Institute, Cary, NC).
At the 3 cross-sectional surveys, we interviewed 445, 437, and 290 women. At baseline a handful of participants had heard of FC by name, and two women reported having used it before. At the first and second postintervention surveys, 364 (83.3%) and 237 (81.7%) reported knowing about FC, respectively, and 126 (28.8%) and 106 (36.6%) had used FC at least once.
In addition to all 445 baseline surveys, we include in postintervention analysis only the women who reported knowing about FC, that is, 364 first postintervention surveys and 237 second postintervention surveys from all study sites. Table 1 shows the study site and establishment characteristics of the 3 samples included in our analysis.
Baseline Characteristics of the Women by Study Site
The demographic characteristics, reproductive history, and sex activities at baseline survey by study site are shown in Table 2. Women from PX were much older and more likely to be married with more complex reproductive experience in general than women in the other study sites.
A large variation in proportion of protected sex was identified across the 4 sites at baseline. Women from FS and PX had significantly more 100% protected sex and less nil protected sex than did women from YF and QH.
Characteristics of 3 Levels of FC Users at the Postintervention Surveys
Table 3 compares demographics, reproductive history, attitudes, and sexual activities among FC nonusers, 1-time users, and multitime users. The 2 types of FC users were, on average, older and more likely to be from PX than nonusers. Multitime users were more likely to have used FC in the prior 30 days than 1-time users, and they reported significantly less unprotected sex than did nonusers in the last 30 days (3.0% vs. 17.2% at first and 3.2% vs. 16.8% at second postintervention survey, P < 0.01).
Multitime users were more worried about getting infected with HIV than the others and had more positive attitudes toward FC. Approximately 84% of them were willing to use FC for prevention of HIV/STI in the future at both postintervention surveys, which was significantly higher than those proportions among the nonusers and 1-time users.
Intervention Participation and Its Association With FC Use
Staff conducted nearly 800 outreach encounters to more than 100 establishments in the 4 sites during the 12 months of 2 intervention phases (only approximately 6 months in QH). More than 3000 FC and 38,000 MC were delivered. Self-reported participation in project activities is shown in Table 4. Multitime users were more likely to report having participated in previous surveys, indicating they had stayed in the study site longer, and were more involved in the intervention compared with nonusers. Intervention score was correlated with self-reported times participated in FC education (Spearman correlation coefficient r = 0.59 at first and r = 0.62 at second postintervention survey, P < 0.01).
Results of polytomous logistic regression of variables associated with FC use levels are shown in Table 5. Better participation in the intervention was consistently associated with being any type of FC user in both surveys, controlling for study site, establishment, and other individual factors in the model. In comparison with nonusers, multitime users were 3 times more likely to be from PX than from YF and 3 times more likely to be from boarding houses than from other types of establishments. Using FC was not associated with the level of MC use in the prior 30 days.
After conducting 6 months of intervention to promote FC along with MC among women working in sex establishments in our 4 study sites, approximately 29% of participants reported using FC either once or multiple times. At the second postintervention survey, this proportion increased to approximately 37%. In comparing 3 levels of FC use, we found that participation in the intervention, being from PX, and working in a boarding house were 3 consistent independent contributors to use of FC at both postintervention cross-sectional surveys, controlling for the effects of age and level of MC use in the prior 30 days.
Female sex workers are an important group for prevention of sex-related ill-health consequences. Although MC use has improved to different degrees in our study sites,22–24 sexual transmission of HIV/STI continues to grow. This calls for further efforts to increase protected sex among women. This project started with an average of 40% of baseline participants reporting 100% MC use and a median proportion of protected sex of 76% in the previous 30 days. Adding FC achieved a significant reduction in the proportion of nil protection comparing multitime FC users to nonusers, although increases in proportions of protected sex were not as significant across the 3 levels of FC users. This result is consistent with other studies6 and our previous analysis for the first 2 of 4 study sites (FS and YF).16
A standard intervention protocol was used in all sites; however, levels of FC uptake were different across sites. This effect was independent of type of establishments and individual factors such as age and level of recent MC use. Our previous publications reported that FC users were more likely to be from boarding houses and level of FC use was higher in PX than in FS and YF,15,16 but relatively few study sites and small sample size prohibited us from exploration of independent contributions of study site to key outcomes. This finding revealed that an intervention protocol standardized in terms of core content of education, method of delivery, and timeline interacted with the diverse contexts of the 4 study sites and women in different types of establishments, producing a different level of FC adoption.17
Although systematic exploration of the role of site and establishment types based on our quantitative data is beyond the scope of this study, several explanations may add deeper understanding of the implications of intervention contexts. The most obvious and important differences among the 4 study sites included differences in the prevalence of HIV/STI and past experience with prevention projects (which represents difference in capacity of outreach workers), in addition to scale and types of sex venues and features of the women. Since the outbreak of the HIV epidemic in PX in the mid-1990s, quite a few HIV prevention programs have been conducted successively with the local CDC as the primary implementation organization. The PX CDC maintained a relatively stable staff team that gained a wealth of experience in intervention implementation and fostered a trusting rapport with women in sex venues. By comparison, outreach workers in the other 3 study sites were inexperienced, and many of them had never worked with women in sex establishments before this project. At least one implication of this is that when we translate an effective intervention such as the FC promotion program into the real settings of diverse communities, we must recognize that community interventions are complex interactions that act on different individual potential users but also in diverse microsetting and macrosettings, such as sex venues and communities.25 We need to tailor the intervention protocol and address community capacity building and supervision to fit each unique community system to achieve optimal impact.
A higher level of FC acceptability among women from boarding houses may have 2 explanations. One is individual characteristics of freelancers from boarding houses. As our univariate analysis showed, FC users were more likely to be older and married with children, which was similar to characteristics of women from boarding houses. These women had more sex acts and fewer 100% protected sex in the last 30 days, and they charged less for each sex act but bore greater obligation to their families. They were also more worried about getting infected with HIV, which indicates their significant need of FC in reality. Although the dichotomous age group was not independently associated with level of FC use, we would say that it is worth further exploration whether age and other factors related to age, such as reproductive history, sexual relationships, and sexual activities, facilitate intervention and how. The other was the microcontext of boarding houses in which freelancers worked and the nature of freelance sex work. Our outreach workers reported that FC education and delivery were conducted more effectively in boarding houses than in other types of establishments because women there could arrange their time more freely and liked to share their experience of FC use with each other during the intervention. Mechanisms of how individual-level and microsetting-level factors affect intervention implementation should also be further analyzed with our qualitative data or explored in future studies. In particular, we could not explain the difference in FC attitudes scores across 3 groups in our 2 surveys. Understanding mechanisms of change at multiple levels would be very valuable for improving intervention design and conduct.
Several limitations should be kept in mind when interpreting the findings of this study. First, the types of sex establishments of the 4 study sites were generally similar to the sex industry in rural and small- to middle-sized urban settings in China, and we recruited approximately 70% to 80% of women working in the sex industry in our study sites or sampled study districts in the midsized city. However, the details of features of the same type of sex venues (e.g., massage parlors) and other contextual factors that affect intervention implementation and outcomes might be different from other communities. Second, the level of lifetime FC use was measured in this study, but some explanatory factors were measured in the prior 30 days. Therefore, the results should be interpreted with care. Third, free FC provision might inflate the initial acceptability; however, providing free product might always be the first step to introduce a new product. Finally, sex workers’ mobility undermined our attempts to track individual women and only allowed us to compare different groups in each of the cross-sectional surveys, which limited our ability to accurately explore causal relationships regarding FC use.
Although our study suggests that adding FC into MC-only intervention can reduce unprotected sex among women working in the sex industry in the relatively natural setting of communities, our findings may uncover more complications of combining FC into existing MC intervention. This supports the perspective that the scientific paradigm for community intervention should be collaborative, multilevel, and culturally situated25 and reaffirms the call for interdisciplinary and system sciences to improve population health.26
1. Li X, Hong Y, Poston D. Preventing HIV in women: A top priority in China’s efforts in fighting AIDS. AIDS Care 2011; 23 (suppl 1): 1–4.
2. Huang YY, Henderson GE, Pan SM, et al.. HIV/AIDS Risk among brothel-based female sex workers in China: Assessing the terms, content, and knowledge of sex work. Sex Transm Dis 2004; 31: 695–700.
3. Wang L, Ding ZW, Ding GW, et al.. Data analysis of national HIV comprehensive surveillance sites among female sex workers from 2004 to 2008. Zhonghua Yu Fang Yi Xue Za Zhi 2009; 43: 1009–1015.
4. Macaluso M, Blackwell R, Jamieson DJ, et al.. Efficacy of the male latex condom and of the female polyurethane condom as barriers to semen during intercourse: A randomized clinical trial. Am J Epidemiol 2007; 166: 88–96.
5. Gallo MF, Kilbourne-Brook M, Coffey PS. A review of the effectiveness and acceptability of the female condom for dual protection. Sex Health 2012; 9: 18–26.
6. Vijayakumar G, Mabude Z, Smit J, et al.. A review of female-condom effectiveness: Patterns of use and impact on protected sex acts and STI incidence. Int J STD AIDS 2006; 17: 652–659.
7. The female condom: still an underused prevention tool. Lancet Infect Dis 2008; 8: 343.
8. Choi KH, Hoff C, Gregorich SE, et al.. The efficacy of female condom skills training in HIV risk reduction among women: A randomized controlled trial. Am J Public Health 2008; 98: 1841–1848.
9. Latka M, Gollub E, French P, et al.. Male-condom and female-condom use among women after counseling in a risk-reduction hierarchy for STD prevention. Sex Transm Dis 2000; 27: 431–437.
10. Rasch V, Yambesi F, Kipingili R. Acceptance and use of the female condom among women with incomplete abortion in rural Tanzania. Contraception 2007; 75: 66–70.
11. Vieira EM, Machado AA, Duarte G, et al.. The use of the female condom by women in Brazil participating in HIV prevention education sessions. Rev Panam Salud Publica 2004; 15: 373–379.
12. Napierala S, Kang MS, Chipato T, et al.. Female condom uptake and acceptability in Zimbabwe. AIDS Educ Prev 2008; 20: 121–134.
13. Lara DK, Grossman DA, Munoz JE, et al.. Acceptability and use of the female condom and diaphragm among sex workers in Dominican Republic: Results from a prospective study. AIDS Educ Prev 2009; 21: 538–551.
14. Yimin C, Zhaohui L, Xianmi W, et al.. Introductory study on female condom use among sex workers in China. Contraception 2002; 66: 179–185.
15. Liao SS, Weeks MR, Wang YH, et al.. Inclusion of the female condom in a male condom-only intervention in the sex industry in China: A cross-sectional analysis of pre- and post-intervention surveys in three study sites. Public Health 2011; 125: 283–292.
16. Liao SS, Weeks MR, Wang YH, et al.. Female condom use in the rural sex industry in China: Analysis of users and non-users at post-intervention surveys. AIDS Care 2011; 23 (suppl 1): 66–74.
17. Weeks MR, Liao S, Li F, et al.. Challenges, strategies, and lessons learned from a participatory community intervention study to promote female condoms among rural sex workers in Southern China. AIDS Educ Prev 2010; 22: 252–271.
18. Wang Y, Liao SS, Weeks MR, et al.. Acceptability of hypothetical microbicides among women in sex establishments in rural areas in Southern China. Sex Transm Dis 2008; 35: 102–110.
19. Nong LP, He B, Li WL, et al.. HIV/AIDS epidemic characteristics and trends in Pingxiang, 1996–2009. J Prevent Med Inform 2011; 27: 3.
20. Xu HW, Pei DN, Zhong N, et al.. Analysis of infectious status of syphilis in special population in Qionghai City, Hainan. China Trop Med 2008; 8: 2.
21. Neilands TB, Choi KH. A validation and reduced form of the Female Condom Attitudes Scale. AIDS Educ Prev 2002; 14: 158–171.
22. Liao SS, Schensul J, Wolffers I. Sex-related health risks and implications for interventions with hospitality women in Hainan, China. AIDS Educ Prev 2003; 15: 109–121.
23. Liao SS, He QY, Choi KH, et al.. Working to prevent HIV/STIs among women in the sex industry in a rural town of Hainan, China. AIDS Behav 2006; 10: S35–S45.
24. Nong LP, Nie L, He B, et al.. Condom use among female sex workers in Pingxiang city of Guangxi Zhuang Autonomous Region. Chin J AIDS STD 2011; 17: 4.
25. Trickett EJ, Beehler S, Deutsch C, et al.. Advancing the science of community-level interventions. Am J Public Health 2011; 101: 1410–1419.
26. Mabry PL, Olster DH, Morgan GD, et al.. Interdisciplinarity and systems science to improve population health: A view from the NIH Office of Behavioral and Social Sciences Research. Am J Prev Med 2008; 35: S211–S224.