India has an estimated 2.5 million HIV infections, making it the country with the second largest number of people living with HIV in the world . Within India, the northeastern states of Manipur, Mizoram and Nagaland, and the southern states of Andhra Pradesh, Karnataka, Maharashtra and Tamil Nadu have the highest HIV prevalence. There are large district variations in HIV prevalence within the 111 districts in these states, ranging from less than 1% among antenatal clinic attendees in 68 districts, to 3% and above in 10 districts . Whereas the majority of HIV infections in the northeastern states are believed to occur through injection drug use, the epidemic in southern India has been attributed largely to heterosexual transmission, with unprotected paid sex the main transmission route .
In 2003, the India AIDS Initiative (Avahan) of the Bill and Melinda Gates Foundation initiated a focused HIV prevention programme among female sex workers (FSW), high-risk men who have sex with men (including transgenders), clients of sex workers and injection drug users in these states, and along the national highways . Three main strategies were employed to address HIV prevention among sex workers and their clients: promotion of safer sex behaviour through a peer-mediated communications strategy, with particular focus on promotion of condom use; enhanced sexually transmitted infection management, with distribution and social marketing of condoms; and enhancing the enabling environment for the adoption of safer sex practices. Underlying these programme strategies were the principles of community mobilization and participation in programme design and execution. A series of district level cross-sectional surveys assessing behavioural and biological endpoints, termed the integrated behavioural and biological assessments (IBBA), forms an essential part of the Avahan evaluation strategy .
Using data collected from the first round of IBBA among FSW in 23 districts of the four high-prevalence southern Indian states, this paper describes sociodemographic and sex work characteristics, explores risk factors for HIV infection, and examines factors that may explain interdistrict variations in HIV prevalence.
Drawing upon the proximate determinants conceptual framework proposed by Ties and Weir , the factors associated with HIV risk among FSW were grouped into the following four broad categories: (1) sociodemographic characteristics, including age, age at sexual debut, marital status, literacy, sources of income other than sex work, local residential status, and migration for sex work; (2) sex work characteristics, including age at starting sex work, duration in sex work, usual place of solicitation, and sex client volume; (3) ‘agency’, including membership in a collective (sex worker association or organization), and exposure to HIV prevention programmes; and (4) condom use. The variables in the latter two categories were not included in this analysis, as they will be explored in the context of a follow-up IBBA, planned to occur 2–4 years after the initial IBBA.
Study design and sampling
The 23 districts for IBBA were chosen purposively, based on each state's sociocultural regions and size of the high-risk population. In each state, the district with the capital city was also included. The target sample size per district was fixed at 400 completed interviews plus blood samples, and the sample sizes in Mumbai, Pune, Thane and Bangalore cities were enhanced to 800 each, to represent the different sex work typologies in these very large cities. Full details of the survey design can be found in Saidel et al. .
A probability sampling method was used. Four different sampling approaches were adopted: (1) conventional cluster sampling was used for FSW practising sex work at homes, brothels, lodges and dabhas (road-side eating establishments), where the population of FSW was relatively stable; (2) conventional time-location cluster (TLC) sampling  (dividing a site into several TLC and selecting the required number of TLC randomly) was used for street-based FSW; (3) a variation of the TLC sampling was used in Mysore city, wherein the three sex work zones were divided into TLC, and FSW were recruited proportionately from all of the TLC; and (4) respondent-driven sampling was used for populations in which a substantial proportion of FSW did not congregate in identifiable locations and would have been missed if venue-based or TLC sampling methods were used. The data from locations where respondent-driven sampling was used were excluded from the current analysis, as because of the sampling method used, it was not possible to integrate these with the rest of the data, using appropriate weightings. This comprised bar-based FSW and certain other FSW populations in Mumbai and Parbhani districts in Maharashtra district only, totalling 1108 women. For the calculation of proportions and averages, appropriate district, state and global weights were used to account for differential recruitment of FSW by typology within districts, differential non-response rates, and differential probabilities of selection across districts.
Survey organization and methods
Statutory approval for the conduct of the IBBA and its protocols was obtained by the Government of India's Health Ministry Screening Committee. The National AIDS Research Institute, Pune, National Institute of Epidemiology, Chennai, National Institute of Nutrition, Hyderabad, and Karnataka Health Promotion Trust, Bangalore, implemented the IBBA in Maharashtra, Tamil Nadu, Andhra Pradesh and Karnataka states, respectively. The survey was conducted anonymously, with no names or personal identifiers recorded. A detailed and standardized consent process was implemented for each respondent, and consent was obtained separately for the interview and for giving biological samples.
Blood (and urine, in Karnataka only) samples were taken for HIV testing. HIV serological testing was conducted using Microelisa (J. Mitra and Company, India), and positive tests were confirmed using Genedia HIV 1/2 ELISA 3.0 (Green Cross Life Science Corporation, South Korea). The Karnataka survey used the Detect HIV 1/2 system (BioChem ImmunoSystems, Montreal, Canada) as the screening test. In Karnataka, where serum samples were not provided, a dried blood spot was performed on finger prick blood using the same serological tests. Also in Karnataka, urine samples were tested for HIV by Calypte (Biomedical Corporation, Berkeley, California, USA), and confirmed by Western blot, when neither serum nor finger prick samples were provided.
The district FSW IBBA data were merged to form one database to allow analysis across districts and states. Analysis was restricted to standard data points that were available across all of the 23 districts. Bivariate and multivariate analysis of associations between independent variables and HIV infection was undertaken using STATA version 9.0. Factors were then explored that were associated with interdistrict variations in HIV prevalence by fitting a two-level model (individuals at level 1 nested within district at level 2), using multilevel regression procedures. The principles of multilevel modelling procedures have been discussed elsewhere [9,10]. As the response consisted of a binary variable classifying individuals as either HIV positive or not, we used a random intercept logistic model, based on a logit link function for estimates of the log odds of being HIV positive.
Each of the districts and the state that they belong to is shown in Table 1. Overall, 73% of the FSW invited for the IBBA gave both the interview and a biological sample, the response rate ranging from 45% in Kolhapur to 89% in Shimoga. The achieved sample was 10 096 and ranged from less than 200 FSW in Kolhapur (115) and Yevatmal (153) to over 600 in Mumbai (801), Thane (795), Bangalore Urban (669) and Pune (661).
As shown in Table 1, the median age of the respondents was 30 years, with 18% under the age of 25 years. The majority of the respondents were illiterate, most being currently married (46%) or widowed, divorced or separated (44%). Just under half of the respondents did not have any source of income other than sex work. Over 90% of the FSW interviewed were local to the cities/towns where the IBBA were conducted, and only approximately one-third had ever gone out of their local city/town for sex work in the year before the survey. One in four FSW reported sexual debut before the age of 15 years.
Overall, 19% of the FSW had started sex work when they were under age 20 years (median 25 years), and over half (53%) were in sex work for a duration of less than 5 years (median 4 years). Over half (54%) of the respondents usually solicited in public places such as parks, streets, markets, open spaces, etc.; another 18% operated from homes (rented or their own), and only 16% worked at brothels, lodges or dabhas. One-third of the FSW entertained 10 or more sex clients per week. The profile of FSW varied greatly by state and district. The median age of the respondents ranged from 25 years in Yevatmal to 34 years in Coimbatore. The proportion of FSW under age 25 years was substantially lower in Tamil Nadu (14%) compared with other states (approximately 21%). Yevatmal had the highest proportion (35%) of FSW under age 25 years, the lowest being in Salem (8%).
In Maharashtra, the majority of FSW were either widowed, divorced or separated. On the other hand, the majority in Tamil Nadu were currently married, with only 37% being widowed, divorced or separated, and another 3% unmarried. Andhra Pradesh and Karnataka were somewhat similar with respect to the distribution of FSW by current marital status. Across districts, the proportion currently married ranged from less than 20% in Yevatmal, Belgaum, Pune and Kolhapur to more than 50% in Hyderabad, Coimbatore, Madurai, Chennai and Prakasam districts. The percentage of FSW never married ranged from a low of less than 5% in several districts to a high of 25% or more in Bellary, Belgaum, Thane and Pune. Women from the devadasi tradition were reported only in Belgaum (23%) and Bellary districts (11%) in Karnataka. The devadasi tradition involves a ritual in which a young girl is dedicated to a religious deity, and is a form of traditional sex work in northern Karnataka and in certain other states in India.
Over 70% of respondent FSW in Andhra Pradesh and Maharashtra were illiterate, compared with approximately 57% each in Tamil Nadu and Karnataka. Literacy was lowest in Hyderabad (14%) closely followed by Belgaum (18%). Over half of the respondents were literate in Coimbatore and Bangalore urban districts. Over 90% of the respondents in Maharashtra were exclusively engaged in sex work, with no other source of income. This proportion was approximately 40% in the other three states. A relatively greater proportion of FSW in Shimoga (80%), Madurai (78%), Karimnagar (73%) and Bellary (71%) had sources of income other than sex work.
The median age at sexual debut varied from 15 years among FSW in Andhra Pradesh to 18 years in Tamil Nadu (Table 2). Across districts, the median age at sexual debut ranged from 14 years in Belgaum and Hyderabad to 18 years in Chennai, Coimbatore and Madurai. More than half of the FSW in Hyderabad and Belgaum had their first sexual intercourse when they were under age 15 years, compared with less than 10% in Chennai and Coimbatore. Only 2% of the FSW in Maharashtra reported that they usually reside in the cities/towns where they were interviewed for the IBBA, and this proportion in the other states ranged between 7% and 10%. In Visakhapatnam, Yevatmal, Chennai and Salem, all of the interviewed FSW were residents of the cities/towns where the IBBA was conducted. In comparison, over one-quarter of the respondent FSW were non-local in Warangal, Dharmapuri, Mysore and Hyderabad.
Almost three-quarters of the FSW in Andhra Pradesh reported that they had sold sex outside their usual place of residence, and this proportion was 29% in Tamil Nadu, 13% in Karnataka and 5% in Maharashtra. Reported movement for sex work outside the usual place of residence was the highest at 94% in Karimnagar, and the lowest at 2% in Mumbai. Large variation across districts was observed in the organization of sex work. Whereas over half the FSW in Hyderabad, Karimnagar, Prakasam, Kolhapur, Mumbai, Bangalore, Mysore and in all the five Tamil Nadu districts solicited in public places, 40% or more of the FSW in Mumbai, Pune, Thane, Yevatmal, and Belgaum worked in brothels/lodges/dabhas. The proportion of FSW working at home was 40% or more in Chitoor, Guntur, Prakasam, Belgaum, Bellary and Shimoga districts.
The median number of clients entertained by FSW per week ranged from less than five in Shimoga, Hyderabad, Karimnagar, Chennai, Coimbatore, and Madurai, to 10 or more in Yevatmal, Belgaum, Guntur, Prakasam, Visakhapatnam and Pune. The proportion of FSW with a weekly sex client volume of 10 or more ranged from 11% in Chennai to 77% in Yevatmal. FSW in Bellary and Belgaum were in sex work for a relatively longer duration (median of 7 years) compared with a low in Bangalore urban (median of 2 years). In Andhra Pradesh and Maharashtra, 55% and 58% of FSW, respectively, were in sex work for 5 years or longer. Correspondingly, the median age at starting sex work ranged from 20 years in Belgaum and Bellary (where the proportion of devadasi was relatively higher) and Warangal, to 28 years in Chennai, Coimbatore and Salem districts.
Table 3 presents HIV prevalence as well as crude and adjusted odds ratios, by selected background characteristics of FSW. The overall weighted HIV prevalence in the 23 districts surveyed was 14.5% [95% confidence interval (CI) 14.0–15.4]. In multivariate analysis, women under 25 years had a lower HIV prevalence [adjusted odds ratio (AOR) 0.73, 95% CI 0.61–0.87]. There was no difference in HIV prevalence in multivariate analysis between local and non-local dwelling FSW, and between those who had ever migrated for sex work and those who had not. Unmarried FSW and FSW who were either widowed/divorced/separated or devadasi had a significantly higher odds of being HIV positive (AOR 1.97, 95% CI 1.74–2.22 and AOR 1.75, 95% CI 1.44–2.11, respectively), compared with currently married FSW. The probability of HIV infection was significantly lower for literate (AOR 0.73, 95% CI 0.65–0.83) than illiterate FSW. The probability of being HIV infected was approximately 0.75 times lower for FSW who had their sexual debut when they were older than 15 years of age compared with under 15 years.
Compared with home-based and public place-based FSW, the odds of being HIV positive was twice as high for those operating from brothels/lodges/dabhas (AOR 1.95, 95% CI 1.64–2.32). FSW who entertained 10 or more sex clients per week had 1.25 times higher odds of being positive than those with a lower weekly client volume. The probability of HIV infection was approximately 20% lower for FSW who started sex work when they were older than 20 years of age, and 16% higher for those who were in sex work for 5 or more years.
District variations in HIV prevalence
As seen in Table 2, large variations in HIV prevalence were observed across the IBBA districts, ranging from less than 10% in five districts (Chennai 2%, Madurai 4%, Coimbatore 6%, Chitoor 8%, and Shimoga 9%) to more than 30% in four districts (Pune 38%, Yevatmal 37%, Belgaum 34%, and Kolhapur 33%). Most of the districts with a relatively low HIV prevalence (<10%) were in Tamil Nadu state, and most of the districts with a relatively high HIV prevalence were in Maharashtra state.
We examined district-level effects on HIV prevalence by adjusting for the effects of differences in the distribution of individual-level risks between districts. Using multilevel logistic regression models, district-level variance was evaluated for the different characteristics of FSW (Table 4). A comparison between the null model and random intercept model indicated that almost two-thirds of the district-level variation was accounted for by the individual characteristics considered in the model. The effects of most individual characteristics remained unchanged after allowing for district-level variation, although the effects of three variables (source of income other than sex work, sex client volume, and age at starting sex work) lost their statistical significance. Allowing for district-level variation, women who were widowed, divorced, separated or devadasi (β = 0.589, SE 0.065), brothel-based (β = 0.486, SE 0.101), or unmarried (β = 0.468, SE 0.100), remained more likely to be HIV positive.
Explanatory variables with statistically significant β coefficients were then allowed, one by one, to vary across districts (data not shown). The effects of current marital status and sex work typology on HIV prevalence varied significantly across districts, and the estimated district-level variance in HIV prevalence was the lowest for brothel-based unmarried FSW (0.152), followed by brothel-based widowed, divorced, separated or devadasi FSW (0.192). FSW living in districts with a higher proportion of currently married FSW were significantly less likely to be infected with HIV. Similarly, HIV prevalence among FSW tended to be higher in districts with a higher proportion of brothel-based FSW who were not currently married. The observed differences across districts in HIV prevalence among FSW could thus be largely attributed to the district variations in the distribution of FSW by current marital status and usual place of solicitation.
In this paper, we have described sociodemographic and sex work characteristics of FSW across 23 districts in four southern Indian states. In addition, we have explored risk factors for HIV infection among FSW, and the characteristics of individual FSW that may help explain interdistrict variations in HIV prevalence. The importance of understanding the demographic and sex work characteristics of FSW for improving the reach and effectiveness of prevention programmes has been previously observed .
Marital status was confirmed as an important indicator of HIV risk. A study of correlates of HIV prevalence among FSW attending STI clinics in Pune, Maharashtra, also indicated a higher probability of infection for widowed and unmarried FSW . In this IBBA, married FSW who resided with their husbands started sex work relatively later in life and had a lower sex client volume. FSW who were widowed, divorced or separated also tended to start sex work relatively later in life (mostly after separation from their husbands), but depended exclusively on sex work for income. Unmarried FSW, on the other hand, were younger and reported a higher client volume. Unmarried FSW were more likely to work in brothels/lodges/dabhas, and to depend largely on sex work as their main source of income. The variability in the current marital status of FSW across the districts and states represents the differential patterns of entry into sex work, as well as the structure and organization of sex work, and has an important influence on the risk of HIV. This has important programmatic implications.
Vulnerability to HIV infection was also strongly associated with the organization of sex work, including the usual place where an FSW solicits her clients . In this IBBA, brothel-based FSW were younger, more likely to be unmarried, and had more clients compared with home-based or street-based FSW. Most sex work in Maharashtra districts is organized around brothels, whereas sex work in Tamil Nadu and Karnataka is largely street-based, providing an important contrast in risks for HIV. However, there were important differences between Tamil Nadu and Karnataka in the profile of street-based FSW: a greater proportion of street-based FSW in Karnataka than in Tamil Nadu were under age 25 years, and reported 10 or more sex clients a week, indicating a higher risk profile.
Multivariate analysis indicated that several attributes of FSW help to explain variability in HIV prevalence: age, marital status, literacy, principal income source, age at sexual debut, age at starting sex work, duration in sex work, usual place of solicitation, and sex client volume. Interestingly, migration for sex work and being a local resident or not, were not associated with HIV risk. Furthermore, even in districts that are thought to receive large numbers of migrant FSW, a large majority of the respondents in the IBBA reported that they usually reside at the place of interview. According to a mapping of sex access points and relevant service outlets conducted in selected areas of Maharashtra, more than half of the brothels in Pune (72%), Thane (59%), Sangli (57%) and Mumbai (51%) reported at least one FSW from Karnataka (Blackstone Market Facts, 2000, unpublished report). This mapping did not quantify the number of FSW who were from Karnataka, and information was collected from the brothels and not from individual FSW. In addition, the individual FSW in the IBBA might have hesitated to report that they were not from the place where they worked.
Apart from confirming the greater risks for HIV associated with working in brothels and being unmarried, widowed, divorced, separated or devadasi, the multilevel analysis suggested that the distribution of FSW according to current marital status and place of solicitation in a given district are important predictors of HIV prevalence among FSW in that district. The district where the sex worker operates adds little beyond these individual risk factors. The demographic and sex work characteristics of FSW seem to reflect a host of other risk factors, such as current age, duration in sex work, sex client volume, age at sexual debut, as well as age at starting sex work. These variables have been found in other studies to be important predictors of HIV status among FSW . There are large variations across IBBA districts in the stage of the life-cycle at which a woman enters sex work (e.g. at a young age as a devadasi, or much older after becoming a widow, or being separated or divorced), and the way that sex work is organized (whether the FSW operates from an organized brothel/lodge/dabha, or whether she works from her residence, or solicits in public places). These variations in the structure of sex work clearly have implications for the way that HIV prevention programmes are organized, and how they access sex worker communities [11,15,16].
Overall, only 18% of the FSW in the IBBA sample were under the age of 25 years. In comparison, 36% and 30% of the respondents in two independent surveys of FSW in Andhra Pradesh  and Karnataka , respectively, were under the age of 25 years. In the IBBA, the proportion under age 25 years was particularly low in Tamil Nadu districts (14%). Not only were the respondent FSW in Tamil Nadu older, they started sex work at a relatively later age, and a greater proportion were currently married and living with their husbands. A similar profile of FSW was observed by another study in Tamil Nadu . This point is underscored in Belgaum and Bellary districts in Karnataka, where only 18–20% of the FSW respondents were under age 25 years, yet approximately half of those interviewed reported that they had started sex work before the age of 20 years. However, in many parts of northern Karnataka, such as the Belgaum district, FSW tend to start sex work outside their home district (mostly in Maharashtra) at a young age, and return to continue sex work in their home district after spending a few years away [11,19]. At any given point in time, the available FSW in such districts could thus be somewhat older, even if they had started sex work at an early age.
One of the potential limitations of this analysis is the difficulty in determining the temporal relationship between the predictors and outcomes, and not knowing for how long women with HIV have been infected. Therefore the value at the time of survey for some of the independent variables included in this analysis may differ from what it was at the time that HIV infection occurred. Another limitation of the study is that the profile of the non-respondents is not known, and to that extent the biases reflected in the low response rates in some of the districts cannot be ascertained. In addition, as suggested above, the representativeness of the sample, in spite of the rigorous sampling strategy, may not have been optimal.
In conclusion, this study has highlighted the heterogeneity in the structure and organization of female sex work, and how this is reflected in HIV prevalence among FSW in India. District-level variations in HIV prevalence among FSW are largely determined by the composition of FSW in terms of their current marital status and usual place of sex work solicitation, which in turn seem to reflect a variety of other characteristics associated with the risk of HIV. The district or the region does not necessarily drive HIV prevalence among FSW, but there are clusters of different cultures of sex work that operate differently across India. These understandings should form an important consideration for designing and enhancing HIV preventive interventions for FSW.
IBBA study team
National AIDS Research Institute (NARI), Pune: Abhijit Deshpande, Amey S., Amol Salagare, Arun Risbud, B. Kishorekumar, Bhagyashri, Deepak More, Dilip Pardeshi, Geetanjali Mehetre, Jagnnath Navale, Jayesh Dale, Mandar Mainkar, Milind Pore, Narayan Panchal, Rahul Gupta, Raman Gangakhedkar, Sachin Kale, Sachin P., Shailaja Aralkar, Shashikant Vetal, Shirin Kazi, Shradha Gaikwad, Shradha Jadhav, Sucheta Deshpande, Sujata Zankar, Tanuja Khatavkar, Trupti Joshi, Uma Mahajan.
National Institute of Nutrition (NIN), Hyderabad: B. Narayana Goud, B. Sesikeran, Ch. Hanumatha Reddy, G. Krishna Reddy, G.N.V. Brahmam, K. Venkaiah, L.A. Rama Raju, M. Chandra Sekhara Rao, M. Shamsuddin, R. Harikumar, R. Hemalatha, S.P.V. Prasad, V.V. Annapurna.
National Institute of Epidemiology (NIE), Chennai: A. Bhubneswari, A. Manjula, A. Pauline Priscilla, A. Sivaraman, A.K. Mathai, Beena Thomas, C. Femina, C. Kalpana Devi, C. Selvendran, C.P. Girish Kumar, D. Prabhu, J. Rajkumar, Jagan, Jeyasingh, Joseph David, K. Boopathi, K.J. Dhananjeyan, K.J. Kalyanam, L. Palani, M. Amulu, M. Stabri Dhanabakyam, Michael, Muniraja, Paul Tambi, R. Muthu, S. Karthikeyan, S. Periasamy, M.D. Gupte, S. Velan, Stephen Raja, T. Karunakaran, T. Rabinson, T. Venkata Rao, V. Selvaraj, Vasana Joshua.
Regional Medical Research Council (RMRC), Dibrugarh: Ashim Das, Basumati Apum, D. Borasaikia, Dulon Chetia, G.K. Medhi, Gajendra Singh Golap, Ch. Barua, Gunavi Sonowal, H.K. Das, J. Mahanta, Jogeswar Barman, Manas Barman, Mintu Gogoi, Nabajyoti Laskar, Purnima Barua, R. Gogoi, S.Z. Hussain, T. Rahman, Utpal Saikia, Wahid Bora.
National Institute of Medical Statistics (NIMS), New Delhi: Arvind Pandey, B.S. Sharma, D. Sahu, D.K. Joshi, G.P. Jena, M. Thomas, Nandini Roy, P. Mahato, R.P. Sharma, R.S. Chadha, S.K. Benara, U. Sengupta.
Karnataka Health Promotion Trust (KHPT), Bangalore: Anitha, Christina, Kaveri Gurav, Raghavendra, Shenvi, Sudheer, Sudipto Mondal, Usha Rani.
Family Health International (FHI): Ajay Prakash, Anjalee Kohli, Bitra George, Gay Thongamba, Kathleen Kay (deceased), Lakshmi Ramakrishnan, Motiur Rahman, Nandan Roy, Prabuddhagopal Goswami, Sharad Malhotra, Shubra Rehman, Srinivasan Kallam, Umesh Chawla.
Sponsorship: Support for this study was provided to Family Health International and the Karnataka Health Promotion Trust by the Bill and Melinda Gates Foundation.
The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Bill and Melinda Gates Foundation.
Conflicts of interest: James F. Blanchard receives funding from Canada Research Chairs, Health Canada.
All other authors declare no conflict of interest.