Introduction

Female sex workers (FSWs) are considered a “core” group for HIV transmission because of high infection rates and large numbers of sexual partners [1]. Their clients then act as a “bridge” population, linked into a larger sexual network with spouses, girlfriends and casual partners who are generally considered at low risk of infection [2–8]. In India, adult HIV prevalence was 0.31% in 2009, and most HIV transmission is through heterosexual networks accounting for 88% of the total reported HIV cases [9] Nearly two thirds (57%) of HIV infections occur in four contiguous states in the South (Andhra Pradesh, Karnataka, and Tamil Nadu) and West (Maharashtra) [10]

Both interventions and research is more extensive among FSWs because they are easier to reach, even if the context is complex. There is great diversity in typology of FSWs at national and state level in India [11]. Studies use various criteria to define type of female sex work based on ‘place of sex’ [12,13] and ‘primary place of solicitation’ [14,15]. In a review, the place of solicitation of sex workers seemed the most suitable criteria to classify FSWs for programming purposes, but the authors acknowledge that it is still largely unknown how this captures the variability in risk and vulnerability of sex workers [11]. Since the scale-up of targeted intervention, several studies have documented a shift towards a more diffuse form of sex trade, with women selling sex from home and soliciting sex via mobile phones, with or without the help of agents who may also make sex workers available at hotel and lodges [11,16–18]. There has been a decline in the more visible forms of sex work solicitation on the streets and at brothels and little is known about how this influences risk reduction in commercial sex.

The size of both the FSW population and their client group, and the rate of unprotected sexual contacts between them are important determinants of the rate of HIV transmission [19]. High levels of sexual risk behaviours have been reported among male sub-populations like attendees of sexually transmitted disease (STD) clinics [20], homeless men [21], contract migrant workers [22] and truckers [23]. Using nationally representative behavioural surveillance data from 2006, Gaffey et al [24] estimate that 8.5 million Indian men had sex with a female sex worker in the last year. In the high HIV prevalence states, 8% of sexually active men reported commercial sex in the last year, compared to only 3% in the low prevalence states.

The first Integrated Bio-Behavioural Assessment (IBBA) survey in Andhra Pradesh, Maharashtra and Tamil Nadu, showed that 28% of clients reported ‘every time’ condom use with sex workers [7]. A quarter of the clients (27%) reported commercial partners only, 47% reported also having a main sexual partner and 27% reported casual and main female partners [7].

Early HIV prevention efforts in India have targeted clients of FSWs through HIV/AIDS awareness campaigns for general population and for truckers through the ‘Healthy Highways’ program [25]. The coverage of bridging populations especially truckers and migrants was expanded under phase III (2006-2011) of India’s National AIDS Control Program (NACP-III) with a strong focus on condom social marketing, STI and HIV testing, treatment and care services [26]. Avahan, the India AIDS Initiative of the Bill and Melinda Gates Foundation, reached out to men visiting FSWs in two main ways. Men were targeted at ‘hotspots’ where clients were most likely to look for sex. The aim was to improve consistent condom use and to create demand for, and access to, STI services [27]. From November 2004 information on HIV, STI treatment, safe sex and condoms was disseminated via a variety of mid-media activities including drama, street theatre, and small group discussions in 100 priority towns in the four high prevalence states [27,28]. Access to condoms was increased by using non-traditional out-lets frequented by men including betel leaf and tobacco shops. During 2007 and 2008 men in the general population were also targeted by a mass-media campaign via television, radio and online public service advertisements designed to normalise condom use as smart and responsible [29].

In this paper we document the profile of clients, during the second round of IBBA surveys in 2009, in terms of demographics, sexual networking and condom use across different solicitation points: from the more visible (public place/brothel) to the less visible (home -based or via phone/agents). We thus indirectly assess sex workers’ vulnerability by typology. We use propensity score matching (PSM) to estimate the impact of intervention messages on clients’ consistent condom use with FSW.

Methods

Results

Discussion

While targeted interventions among FSWs in India have been widely studied, our analyses focuses on the under-researched group of the clients of these sex workers and how their risk profile differs according to the place of solicitation. Heterogeneity is evident in their socio-demographic profile, the extent of their sexual networks, consistency of condom use, exposure to intervention messages and impact on levels of protection. Those usually accessing sex workers from public places and from brothels were contrasted to those using sex services from less visible home-based sex workers or in the last category women contacted via the phone or operating through agents, at hotels/lodges, bars and night clubs. Clients in this last category tended to be younger, more likely to be single, and started buying sex on average only 5 years ago compared to 7 years among all clients. This group and men soliciting home-based sex workers were significantly better educated, especially when compared to clients of brothel based sex workers. Men visiting brothels were least likely to cohabit with a sexual partner, and more likely to live away from their spouse.

The extent of the clients’ sexual networks, the riskiness of sexual practice they engage in and the risk reduction behaviour they adopt combine to affect their own risk and that of their partners: commercial and non-commercial, male and female. Overall, men reported nearly as many male or hijra partners in the last 6 months (15%) as casual female partners in the last year (16%). Clients of FSWs thus form an important bridge between several groups at higher risk (FSW, MSM and hijras) and to casual and marital partners.

Men visiting brothels stand out as reporting fewer casual and commercial female partners and less anal intercourse with men/hijras and FSWs compared to other clients groups. However, they did report the lowest levels of consistent condom use with sex worker (occasional and regular) and with male and hijra partners. Only with casual female partners was condom use higher than other client groups. Men soliciting sex-workers at public places had more extensive sexual networks than brothel-based sex workers and condom use levels were only marginally higher.

Consistent with their higher educational status, men soliciting the more hidden sex workers did report higher levels of condom use, not only with sex workers but also with male and hijra partner. However, a larger proportion engaged in anal sex either with FSW (16-17%) or with male/hijra partners (21%). Sexual networking was most extensive among men who solicited home-based sex workers, with nearly a third reporting casual female partners. While the different components contributing to the risk of HIV transmission varied significantly across these four diverse groups of clients, they combined into HIV prevalence rates that were not significantly different.

The low levels of protection in commercial sex deserve further discussion. The overall level of 45% consistent condom use varied more by state of residence than by place of solicitation. Clients in Andhra Pradesh reported consistent condom use at 65%, compared to 36% in Maharasthra and 29% in Tamil Nadu. This is substantially lower than what was measured in November 2008 in the end line survey which evaluated the Avahan-funded component of integrated behaviour change programme targeted at clients implemented by Population Services International (PSI) [28]. Consistent condom use was 76% in Andra Pradesh, 95% in Maharashtra and 79% in Tamil Nadu. Factors to explain such divergent levels include 1) eligibility criteria (men reporting commercial sex in the last 12 months (PSI) versus in the last month (IBBA), 2) recruitment (data collection during daylight (PSI) while IBBA carried on at night-time) and 3) most importantly the sampling strategy. The PSI survey [28] aimed for a representative sample of men who buy sex at the same hot spots where PSI implemented a mix of interpersonal communication and mid-media activities supported by static outdoor promotional materials. Sampled at the end of an intense campaign of nearly 4 years, it is inherently biased towards men with highest intervention exposure. Clients in the 2009 IBBA are more representative of all men who buy sex at randomly selected hotspots, with less intense intervention exposure on average, and consequently less affected by social desirability bias. Hence there seems no reason to doubt the validity of the clients’ condom use levels reported in this study.

In the 2009 IBBA, 85% of the sex workers reported consistent condom use with occasional and regular clients: 82% in Andhra Pradesh, 95% in Maharashtra and 87% in Tamil Nadu [36]. These levels are not borne out by what clients reported. While clients were sampled in fewer districts than the FSWs one would expect more agreement between the estimates. Social desirability bias is likely to have inflated reports by FSWs since they have been so intensely targeted by interventions on an almost daily basis. On the contrary, exposure to intervention for clients was far less intense, less frequent and less based on interpersonal communication making their responses probably nearer to the truth. Trends in HIV prevalence in Tamil Nadu indeed lend more credibility to reports from men as prevalence among clients was higher in the second round (0.7%-10.2% in 2009 compared to 2%-4.2% in 2006 vis-a-vis that among sex workers that did not decline but remained stable at 6.1% between 2006 and 2009 [36–38]. Furthermore, HIV prevalence in Maharashtra increased among FSWs from 26% to 27.5% which was statistically significant among street-based FSWs in Mumbai and brothel-based FSWs in Thane districts [36,39].

The most visible sex workers (brothel and street-based) are the most intervened, and the most monitored and surveyed group. They have been the focus of most published papers discussing targeted interventions in India which have repeatedly pointed to the gains in protecting commercial encounters [40–43], while all along acknowledging the potential for desirability bias [36,42–45]. These levels are quite contrary to those reported with other partners [36,44,46,47]. . This has lead to an implicit assumption that the main problem now lies in unprotected sex with intimate and regular partners [48]. The second emerging concern is that the sex trade is becoming increasingly hidden, with larger proportions of sex workers operating in less visible ways including via mobile phones [11,17,18]. These sex workers are less easy to reach with interventions as well as research. By definition clients of sex workers are difficult to identify and thus difficult to reach with interventions and surveys. Thus an encouraging finding is that 70% of men soliciting these less visible sex workers (home-based and the remaining ‘other’ category) had been reached with messages on STIs. Presumably this exposure was mainly through mass-media. At brothels, where the potential for more interactive communication channels is highest, exposure to STI messages was lowest (58%; even though 98% had received messages on condoms). After matching clients on education, occupation and other important covariates, the impact of having received messages on STIs on consistent condom use was lower among clients of brothel-based sex workers than among men frequenting the more hidden home-based sex workers and those soliciting via phone, at lodges or in bars. The impact among clients soliciting at public places in Andhra Pradesh was also high in contrast to Tamil Nadu where campaigns seemed ineffectual. While clients reported highest exposure to messages, those exposed were least likely to use condoms consistently. This is difficult to explain and our PSM analysis confirmed a bias caused by unmeasured factors in Tamil Nadu. These may include more prevention fatigue compared to other states, and a potential adverse effect of propaganda about the free availability of antiretroviral treatment in Tamil Nadu.

Several study limitations need consideration. First, caution is needed in comparing the risk profiles of clients at various solicitation sites, as this may be biased by over and under-sampling of certain groups in particular states. While Tamil Nadu has very little brothel-based sex work [7,17], the vast majority of brothel clients were sampled in Maharashtra where street-based sex workers were under-represented. Most clients of home-based sex workers were in Andhra Pradesh as this was the only state where clients were sampled at solicitation sites for home-based sex work [30]. Indeed, all state-specific samples under-represent clients of hidden sex workers. Secondly, we cannot claim to have a representative sample of clients of the home-based and the ‘other’ sex workers. Men who reported usually soliciting these less visible FSWs were in fact recruited at hotspots or near brothels and may have a different risk profile from those who exclusively solicit hidden sex workers. A third limitation that potentially affects the generalisability of our findings is that nearly half of the men approached and eligible for interview refused to participate. Response rates were higher than in IBBA–I, ranging from 24% to 53% between districts [7] and 37% in the PSI endline survey [28]. While inevitable, and the bias introduced unknown, it seems unlikely that non-responders would have either much lower or much higher risk profiles. Fourthly, the survey has poor measures of exposures with no detail on intensity or communication channel and no ability to attribute exposure to any specific campaign. Finally, social desirability bias in reporting desirable behaviour cannot be ruled out, though as discussed above the extent is probably minor. This is generally a problem in interviewer administered face-to-face surveys related to sensitive issues which can be mitigated using audio computer-assisted self interviews, as successfully applied in Indian settings [49].

Our analysis has important implications for HIV prevention programmes in India. Although it has been strongly argued that it is more effective to focus HIV prevention efforts on sex workers rather than on their commercial partners [24], the balance has probably swung too far. In rapidly changing context where sex workers operate in more covert ways and they themselves become more difficult to reach, recommendations still seem to exclusively focus on targeting FSWs by extending interventions to lodges for example 17. We challenge this neglect of targeting men who buy sex and who have extensive networks with other male and female partners. In this study, men reached with intervention messages are more likely to use condoms (with the important exception in Tamil Nadu) and clients who solicit the more hidden sex workers had been exposed to messages. Considering that clients are such a diffuse group and they do not restrict their partners to sex workers, mass-media campaigns might be the best option to further normalise condom use in all sexual relationships. The high levels of unprotected sex reported by clients show that HIV prevention is still an unfinished business in India and that sex workers too need continued support to increase condom use with all partners, though all responsibility should not rest on them.