Kruse, Natalie MPH*; Behets, Frieda M.-T. F. PhD, MPH†‡; Vaovola, Georgine§; Burkhardt, Gillian BS∥; Barivelo, Texina RN¶; Amida, X. RN¶; Dallabetta, Gina MD*
Sex work cannot be ignored in the global HIV epidemic. Sex workers (SWs) are often among the first social groups to be affected by HIV in a given country. Limiting HIV transmission among SWs and their clients can also limit the spread of HIV to the general population. 1 Reliable estimates of the size of SW populations are needed to plan health and prevention services at the clinic and community levels. Enumeration can also be used as a way to generate more information about the context of sex work (e.g., how many, where, and at what times SWs congregate).
Diego-Suarez is a provincial capital of approximately 90,000 inhabitants located on the northern tip of Madagascar. Shipping and fishing industries, refinement of sugar, and exportation of coffee and spices are the main staples of the local economy. The primary vessels visiting the port of Diego-Suarez come from African, Indian Ocean, and European countries, and from Japan. In Diego-Suarez, some aspects of the sex trade can be quite visible, but before this study was conducted, the approximate number of SWs was unknown.
As in most places, sex work in Diego-Suarez comes in many forms. There are SWs who rely on sex work as their primary source of income, those who engage in sex work occasionally to supplement other income, and those who exchange sex for goods frequently or infrequently. Some SWs in Diego-Suarez are registered members of the association Fikambanan'ny Vehivavy Mpandeha An-Tsambo (FIVMATA), or “Women Who Go to the Boats,” and frequent the port and various nightclubs. The clients of the registered SWs are generally sailors or tourists. Registered SWs carry a health card and are required to make regular visits to the public sexually transmitted infection (STI) clinic to gain access to the port. Since 1987, approximately 1200 SWs have registered with FIVMATA. However, no attempt was ever made to “remove from the rolls” those SWs who have left Diego-Suarez, died, or ceased to be active SWs. Many registered SWs only keep up with clinic visits during the high shipping season. The number of registered SWs present in Diego-Suarez at a given time fluctuates depending on the shipping and tourist industries, which are seasonal. During the height of the shipping season, the maximum number of registered SWs who presented to the clinic in 1999 for screening was approximately 300 per month.
Nonregistered SWs in Diego-Suarez generally serve the resident men; they are often women in difficult circumstances or girls who are lured by the “easy money.” Employment possibilities for women in Diego-Suarez are slim, and many women engage in sex work to support their families or to supplement meager incomes. Because there is no system of enforcement to ensure that nonregistered SWs make visits to the STI clinic, clinic records cannot be used to estimate the total number of SWs.
Capture–recapture (CR) is a technique that was originally used to count and track animal populations. During the last several years, the CR method has been employed to estimate the size of hidden or difficult-to-reach human populations such as SWs, homeless persons, and intravenous drug users. Specific methodologies have varied across studies. Some have used variations of CR that are closer to the original applications used with animal populations, such as directly “capturing, tagging, and recapturing” human subjects. 2–4 Other studies have estimated population sizes by applying CR indirectly to lists obtained from clinics, treatment centers, or social services. 5–9 CR has been used frequently in epidemiology to estimate disease prevalence by comparing two or more sources of medical data such as disease registries, private practice records, or community-based surveys. 9–15
Although many publications have provided detailed descriptions of CR techniques used in epidemiology, few published studies have counted hard-to-reach human populations using direct CR (i.e., without the aid of existing lists). This article describes a study undertaken to estimate the size of a mobile SW population in Diego-Suarez, Madagascar, for which no estimates of the size were available. Geographic mapping of commercial sex sites was undertaken in order to better understand the context of commercial sex, assist in the planning of peer education activities, and prepare for the CR study. Both geographic mapping and CR were conducted with the active participation of local SWs. This article emphasizes the process that was used and the constraints that were encountered.
The study took place in the context of efforts to improve the quality of curative and preventive services for STIs offered to SWs in Diego-Suarez, in partnership with SWs. The project aimed to empower SWs; FIVMATA received funding and technical assistance to carry out community-based prevention activities through peer education. The Committee on the Protection of the Rights of Human Subjects of the University of North Carolina and the ethical review board of the Malagasy Ministry of Health approved the study.
Geographic Mapping of Commercial Sex Sites
Sites where SWs recruit clients in the city of Diego-Suarez were identified during a 1-week period in August 2000. Preparation included a search for existing local maps and consultations with government agencies and health centers to obtain population estimates and fokontany (smallest geopolitical unit, which is roughly equivalent to village or neighborhood) breakdowns. As is often the case in urban communities, it is difficult for outsiders to determine how residents define their neighborhoods. 16 Thus, SWs were chosen to act as fieldworkers in order to build on their knowledge of the environment and to optimize the identification of SWs in Diego-Suarez. A total of 40 fieldworkers participated in the mapping exercise. Fieldworkers included 25 FIVMATA peer educators and 15 female SWs recruited by the peer educators to assist with the activity. The 15 SWs who were recruited to assist with the activity were selected based on their knowledge of the local environment, place of residence and work, participation in past activities, and interest in the fieldwork. Fieldworkers were not selected based on their years of schooling as most SWs in Madagascar have little formal education.
Two project coordinators, a nurse from the STI clinic, and two FIVMATA leaders facilitated the training and acted as fieldworker supervisors. A working definition of “sex worker” was established as a girl or woman who was looking or waiting for a client, or had already engaged a client at the time of the observation. A “sex-trade site” was defined as a place where SWs congregate to wait for or solicit clients. After working definitions were agreed upon, fieldworkers discussed different kinds of places where SWs operate in Diego-Suarez. Preliminary lists of possible sex-trade sites were generated for each fokontany, and fieldworkers were split into teams and assigned to fokontany based on their knowledge and experience of the terrain. Teams were instructed in basic map-making techniques. Symbols for different kinds of sites were determined and teams practiced map making. The team also identified key informants who might help fieldworkers to locate sex-trade sites. These were bar and shop owners, taxi drivers, fokontany presidents, neighborhood leaders, and other SWs.
Fieldworkers first contacted fokontany presidents to determine fokontany boundaries, introduce the project, and ask for advice in contacting other key informants. Basic maps were designed by hand. Fieldworkers traversed their fokontany, speaking with key informants and noting the location of sex-trade sites, the times they were active, and the estimated number of SWs who frequent the sites. Depending on the time of activity, sites were visited again and times of sex work verified.
Fieldworkers presented their findings to the group and made corrections as necessary. Maps were copied and a list of sex sites by fokontany and time of activity was generated. Supervisors then surveyed by car all the sites, verified the peak times of sex work activity, and talked with local key informants.
Estimating the Number of Sex Workers Using Capture–Recapture
To obtain an accurate estimate of the number of mobile and often street-based SWs in Diego-Suarez, and to test and adapt a methodology for enumerating SWs that could be replicated in other resource-poor sites, a two-sample CR method was used. In the first phase, SWs were “captured” by giving them an educational brochure in predetermined geographic areas and at predetermined times. Put otherwise, a sample of individuals (M) was captured from the population of SWs, “marked” using a brochure, and returned to the population. Two weeks later, in the same places and at the same times, a second sample was “captured” (C), which comprised a certain number of persons who were captured in the first sample (R). Under the assumption that the proportion marked in the second sample (R/C) is a reasonable estimate of the marked proportion in the unknown population, an estimate of the size of the entire population (N) can be made by equating the two proportions 7,17,18:MATH
The total population was thus estimated using the following formula:MATH
The 95% confidence intervals were determined by calculating the limits for R/C:MATH
and substituting the estimated limits for R/C in the following:MATH
FIVMATA peer educators carried out the CR fieldwork. Fieldworkers were selected from the pool of FIVMATA peer educators based on their skills, knowledge of the local environment, place of residence and work, and participation in the mapping of sex-trade sites. For each of the two rounds, fieldworkers participated in 1-day training sessions. Project coordinators, a nurse from the STI clinic, and FIVMATA leaders facilitated the training activities and acted as fieldworker supervisors.
The enumeration study used the same definition of “sex worker” that was employed during the mapping exercise. No attempt was made to distinguish between registered and nonregistered SWs; the team believed that asking a SW about her registration status would lead to suspicion and distrust, as she might associate the fieldworkers with authorities (e.g., police) trying to regulate sex work.
In order to keep track of the number of SWs encountered during both rounds of fieldwork, SWs were “tagged” using an educational brochure. The team believed that this “tag” would be acceptable to the SWs if attractive and of high quality, while not too valuable to engender dishonest reporting. A brochure that promoted services at the STI clinic was used because we expected that SWs would honestly report whether they had already received a brochure. We also wanted to take advantage of the study to deliver STI prevention messages to as many SWs as possible.
Sex-trade sites, determined during the geographic-mapping phase of the study, were for logistical reasons grouped into 19 clusters, determined by geographical proximity, time of activity, and the expected number of SWs. For example, a cluster could be a street and its bars, one large disco with two nearby bars, or a public place and the adjacent streets. Fieldworkers worked in teams of two or three persons, and each team was responsible for one or two clusters. Fieldworkers were assigned to clusters based on their knowledge and experience of the terrain. A total of 28 fieldworkers surveyed all clusters identified during the mapping.
To avoid overestimating or underestimating the number of SWs, we conducted the study during late February and in early March 2001, when the shipping industry was moderately active based on STI clinic attendance rates for registered SWs. Each round took place on a Friday, two weeks apart. During each round, clusters were surveyed when the sex sites were most active.
Teams made walking tours of their clusters, visiting the previously identified sex sites. When a SW was identified, she was asked a series of questions (Figure 1), and fieldworkers recorded the number of refusals in their notebooks. Supervisors circulated by car to watch over the field teams, verify results, and help solve problems.
The same clusters were revisited at the same times by the same teams 2 weeks later. Once again, SWs were identified and asked a series of questions (Figure 2), and teams recorded the number of refusals.
Geographic Mapping of Commercial Sex Sites
At the time of the exercise, there were no locally available, recent maps of Diego-Suarez. A comprehensive list of sex-trade sites organized by fokontany was developed. Types of sex-trade sites included certain neighborhoods, streets, corners, bars, marketplaces, fronts of pass hotels or bars, public gardens or squares, “taxi-brousse” stations, discos, and the port. While there were certain well-defined neighborhoods and discos where SWs looked for clients, they appeared to engage clients almost anywhere. Many SWs in Diego-Suarez were street-based and mobile and moved from site to site depending on the time of day, day of week, and season. Most sites were located in the central fokontany of Diego-Suarez. However, there were many small bars and cheap pass hotels in outlying neighborhoods that SWs visited in the company of clients. There were no brothels or other establishments where SWs were permanently located, nor were there pimps. Sites were most active on weekend nights, particularly the discos and large bars.
Often SWs living in outlying fokontany left their homes in the early evening in the company of female friends. Sometimes they gathered for a drink or snack at certain small bars on the way to active sites. Many SWs looked for clients at or around the old market place (tsena), around the new market place, in the public garden in the center of town, at the taxi-brousse stations, and in a neighborhood called Tanambao IV. Later in the night SWs often moved to discos, certain bars, and other streets that liven up.
Key informants described that during the high shipping season, many nonregistered SWs frequent discos or bars where foreign clients may be found. When there is a boat in port, registered SWs gather at the port in the evening and frequent larger discos at night. During the low shipping season, some registered SWs leave Diego-Suarez to travel or work in other cities. Those registered SWs that stay in Diego-Suarez settle down with husbands or boyfriends or continue to engage in sex work at the discos or in the fokontany, often with local clients.
Most peer educators and SWs were capable fieldworkers. Although some fieldworkers had difficulty drawing maps, all were able to determine fokontany boundaries, show where sex work sites were located, and describe the local system of sex work.
Capture–Recapture Estimate of the Number of Sex Workers
During the first round of fieldwork 1455 SWs were counted of whom 286 (19.7%) were observed at least at one other site (Table 1). Of 1455 SWs captured during the first round, 58 (4%), some of whom could have been identified more than once, refused to take a brochure. During the second round, 599 SWs were newly identified and 709 SWs were recaptured in the second round. Of the 709 recaptures, 292 SWs received two brochures and were thus captured in both rounds. Four hundred and seventeen SWs were captured in both rounds but refused to take a second brochure during the second round. Of the1308 SWs captured or recaptured during the second round, 136 (10.4%) refused to take a brochure and did not receive this information in either round. Again, some of these women might have been counted more than once.
Applying the formula:MATH
we estimated the population of SW at 2684 persons (95% CI: 2588–2780). The ratio of the number of recaptures to the total number of captures during the second round was 709/1308 (54.2%).
Relevance of Mapping
Geographic mapping of sex work sites was a useful way of learning more about the local context of the sex-trade and was used to organize peer education teams and outreach activities for SWs. Mapping sex work sites was essential to plan enumeration (i.e., to determine where and when to count) and to plan the number of fieldworkers for each site. The obtained information could also be used for more sophisticated applications using GPS.
Validity of Estimate
We had no data sources other than CR, and therefore it was impossible to validate the estimate we obtained. Accuracy of the population estimate depends on fieldworkers’ ability to correctly identify and approach SW. While fieldworkers who are also SWs may be best able to identify another SW by sight, relying solely on appearances or location may lead to overcounting or undercounting. Therefore, it is important to choose fieldworkers who are capable of approaching SWs and engaging them in a dialogue without offending them, and asking acceptable, noninvasive questions in order to determine whether the person engages in sex work.
There are four assumptions that must be met in order to generate reliable estimates based on the CR method 5,17,18,20:
1. The population is closed. No urban center is completely closed, so as one can expect there was some mobility among SWs in and out of the city of Diego-Suarez. However, we believe that the number of SW entering and leaving the city was comparable during the study period. In addition, the time period between the first and second sample, 2 weeks, was not sufficient for a significant number of SWs to enter or leave the city. Importantly, no large military boat had come to port during the study, which otherwise could have lead to a large influx of SWs to Diego-Suarez from other parts of the country. To make sure no large military ships were scheduled to visit, port authorities were contacted before beginning training and fieldwork.
2. The capture sources are independent. The goal of the exercise was to capture as many SWs as possible in both samples. All known active sex-trade sites were visited during both samples. There is no reason to suspect that being counted in the first sample of SWs would affect the probability of being counted in the second sample. The exercise itself should not have affected a SW's choice of venue. However, during the second round there was a higher percentage of “trap avoidance” than in the first round, and a significant proportion of SWs who were recaptured during the second round refused to take a second brochure. If subjects avoid the tag, particularly during recapturing, and are not counted, population size can be overestimated. In order to minimize the effects of trap avoidance on the estimate, fieldworkers kept track of the number of refusals and if these were made by new captures or recaptures. Most of the refusals were made by recaptures during the second round. Some of these SWs might have been recorded as refusal-recaptures more than once; if this occurred, then there would have been fewer true recaptures. If the first- and second-round new captures were accurate but the number of true recaptures was lower, the population estimate would increase.
3. All members of the population have the same probability of being captured. It is likely that we did not capture all SWs in Diego-Suarez. For example, SWs who work out of their homes, reportedly few in number, were not identified during CR. Time and resource constraints prevented a comprehensive survey of neighborhoods to include these persons. Therefore, our results only apply to SWs who are street, port, or bar based. Thus, the estimate derived from the study may be lower than the actual number of all SWs.
4. The capture history of each member is accurate. We attempted to avoid miscounting by using a neutral tag, a brochure advertising STI clinic services, that was not valuable enough to engender misreporting. During the first round, brochures were generally well received by SWs; however, during the second round, 59% of the recaptures refused to take a second brochure. Some SWs were hostile to fieldworkers and did not want to accept the brochure because it was associated with AIDS prevention or condom use. While the refusals were recorded and taken into account in the total population estimate, they were a major source of discouragement for fieldworkers. Choosing a different tag during the second round might have avoided this problem.
The ratio of the number of recaptures to the total number of captures during the second round was 54%. In a two-stage CR study, precision of the estimate declines when this ratio decreases. Having a ratio of at least 30% is a good target. Although we were not able to compare our CR estimate to other data sources, it is unlikely that we overestimated the true number of SWs because the fieldworkers directly counted 2054 SWs during the course of the two rounds.
Manual tallying and recording into a field notebook of the number of SWs encountered was a good way to verify results and gather other information on mobility and refusal rates. Most fieldworkers were capable of keeping track of several variables. However, we had to limit the amount of information being recorded to avoid making mistakes or burdening fieldworkers.
Involving Sex Workers as Fieldworkers
Sex worker peer educators were chosen to act as fieldworkers because SWs may be in the best position to identify other SWs in an environment where it is difficult for outsiders to distinguish who is engaging in sex work. In addition, involving SWs as partners in the research process can be a method of empowerment and a way to transfer skills. Peer educators carried out the fieldwork in neighborhoods in which they were accustomed to working. They were able to take advantage of the relationships they developed with SWs in these neighborhoods to identify other SWs.
We found that with adequate training and support, SWs can be capable fieldworkers, map out sex work sites, and estimate the number of other SWs using CR methodology. Involving SWs as active participants in research concerning sex work should be considered whenever possible.
Implications for Research
Geographic mapping and CR proved to be low-cost and relatively easy-to-use techniques for counting a hard-to-reach population, such as female street-based SWs. When it is not possible to physically count or register a population subgroup for which information is needed, CR should be considered, particularly in resource-poor settings where more sophisticated technology is unavailable.
Implications for STI/HIV Transmission and Control
What does our estimate of approximately 2700 SWs in Diego-Suarez mean for a population of 90,000 inhabitants? Based on the Demographic and Health Surveys conducted in 1997, this estimate suggests that about 12% of 15- to 49-year-old females in Diego-Suarez engage in sex work. 19 How this finding compares to other cities in Madagascar is not known.
Further research on sexual mixing patterns and risk behaviors among SWs is needed in order to better understand the role that SWs and their clients play in the epidemiology of STI and HIV within populations like Diego-Suarez. No data were available on the number and frequency of sexual contacts between SWs and their clients, a key factor in the rate of STI/HIV transmission. What is clear is that about 12% of reproductive-aged females are engaging in sex work and that up to 78% of asymptomatic SWs in two other cities in Madagascar have at least one curable STI, according to a study conducted in 2001. 21 We also know that SWs in Diego-Suarez frequent diverse sex-trade sites where they encounter clients from all walks of life. Many clients have wives and partners and it is not uncommon for SWs themselves to have steady partners in addition to clients. It is clear that SWs and clients are not isolated or small subpopulations that have limited contact with the general population. This may partially explain why STIs are the sixth most common reason to seek care in public clinics according to Ministry of Health statistics. While HIV prevalence is still low in Madagascar, even among SWs (2/986 SWs in Antananarivo and Tamatave tested positive for HIV in 2001), 21 there is a real potential for rapid spread of the virus and progression to a generalized epidemic.
The current study shows that there is an urgent need to strengthen and scale up of STI/HIV prevention efforts among SWs, their sexual partners, and in light of the high proportion of women engaging in sex work and the potential epidemiologic impact, the general population. Efforts were initiated to improve clinic- and community-based education in order to promote consistent condom use and STI care seeking among SWs. The study also demonstrated that the existing public STI clinic in Diego-Suarez would not be able to provide STI care to all SWs if they presented once a month as recommended previously by local public health authorities. Plans were put into place to train additional staff and increase the opening hours of the clinic. Taking into account the capacity of the public STI clinic and the large numbers of SWs in Diego-Suarez, the policy of routine monthly STI care and prevention visits among SWs was subsequently changed to promoting routine visits every three months and curative visits as needed so that the clinic can serve a greater number of SWs. More resources are needed to improve STI/HIV prevention in Diego-Suarez and in other cities in Madagascar, including the creation of income-generating options for women.
1. Celentano DD, Nelson KE, Lyles CM, et al. Decreasing incidence of HIV and sexually transmitted diseases in young Thai men: evidence for success of the HIV/AIDS control and prevention program. AIDS 1998; 12: F29–F36.
2. Ghys PD, Jenkins C, Pisani E. HIV surveillance among female sex workers. AIDS 2001; 15( Suppl 3): S33–S40.
3. McKeganey N, Barnard M, Leyland A, Coote I, Follet E. Female streetworking prostitution and HIV infection in Glasgow. BMJ 1992; 305: 801–804.
4. Status and Trends of HIV/AIDS/STD epidemics in Asia and the Pacific. A report of the official symposium at the 4th International Conference on AIDS in Asia and the Pacific held October 21–23, 1997 in Manila, Philippines. Monitoring the AIDS Pandemic Network; 1997.
5. Mastro TD, Kitayaporn D, Weniger BG, et al. Estimating the number of HIV-infected injection drug users in Bangkok: a capture–recapture method. Am J Public Health 1994; 84: 1094–1099.
6. Fisher N, Turner SW, Pugh R, Taylor C. Estimating numbers of homeless and homeless mentally ill people in north east Westminster by using capture-recapture analysis. BMJ 1994; 308: 27–30.
7. Larson A, Stevens A, Wardlaw G. Indirect estimates of ’hidden’ populations: capture-recapture methods to estimate the numbers of heroin users in the Australian Capital Territory. Soc Sci Med 1994; 39: 823–831.
8. Archibald CP, Jayaraman GC, Major C, Patrick DM, Houston SM, Sutherland D. Estimating the size of hard-to-reach populations: a novel method using HIV testing data compared to other methods. AIDS 2001; 15( Suppl 3): S41–S48.
9. Hsieh YH, Chen CW, Lee SM. Empirical Bayes approach to estimating the number of HIV-infected individuals in hidden and elusive populations. Stat Med 2000; 19: 3095–3108.
10. Stephen C. Capture-recapture methods in epidemiological studies. Infect Control Hosp Epidemiol 1996; 17: 262–266.
11. International Working Group for Disease Monitoring and Forecasting. Capture-recapture, multiple-record systems estimation II. Applications in human diseases. Am J Epidemiol 1995; 142: 1059–1068.
12. Laporte RE. Assessing the human condition: capture-recapture techniques. BMJ 1994; 308: 5–6.
13. Desenclos JC, Hubert B. Limitations to the universal use of capture-recapture methods. Int J Epidemiol 1994; 23: 1322–1323.
14. Gutteridge W, Collin C. Capture-recapture techniques. Quick and cheap. BMJ 1994; 308: 531.
15. Black JF, McLarty D, Mtasiwa DE. Capture-recapture techniques. Difficult to use in developing countries. BMJ 1994; 308: 531.
16. Emmel ND, O'Keefe P. Participatory analysis for redefining health delivery in a Bombay slum. J Public Health Med 1996; 18: 301–307.
17. Cox S, Shipley M. Counting the uncatchable? An epidemiological method for counting drug misusers. Soc Psychiatry Psychiatr Epidemiol 1997; 32: 19–23.
18. Neugebauer R, Wittes J. Voluntary and involuntary capture-recapture samples—problems in the estimation of hidden and elusive populations. Am J Public Health 1994; 84: 1068–1069.
19. Anonymous. Enquete demographique et de sante, Madagascar, 1997. Institut National de Statistique, Antananarivo, Madagascar. Calverton, MD: Macro International, Inc., 1998.
20. Waters WE. Capture-recapture techniques. More unreliable in humans than in birds. BMJ 1994; 308: 531.
21. Behets, Frieda M.T.F.et. al. Evidence-based treatment guidelines for sexually transmitted infections developed with and for female sex workers. Trop Med Int Health 2003; 8:3.: 251–258.