Female sex workers are at high risk for infection with HIV, and their clients may act as a bridging population by spreading HIV to the general population. Comprehensive HIV surveillance among sex workers includes surveillance of HIV infection, of sexually transmitted infections and of risk behavior. Surveillance of HIV infection among sex workers is critical for countries with low-level or concentrated HIV epidemics, and can help in monitoring the response to the HIV epidemic in countries with a generalized epidemic. Sex workers are a vulnerable population, and particular attention needs to be paid to human rights issues including consent, confidentiality and stigma. Collaborations with key players in the local sex work scene - sex workers themselves in the first place - and alliances with salient institutions and groups are key to the success of surveillance among sex workers. Surveillance activities should have a strong link to interventions targeted at sex workers. Surveillance for HIV infection among sex workers can be institution- or community-based. Institutional settings include screening programs for registered sex workers, of sexually transmitted diseases clinics, and re-education camps. Specific sources of bias need to be considered in different settings, and must be measured - through the collection of socio-demographic and behavioral data - to allow a correct interpretation of prevalence data and time trends. Community-based HIV infection surveillance can be conducted in a probability sample of the sex worker population, thereby reducing selection bias. This requires the mapping of sex workers' contact venues, and drawing a random sample from the resulting sampling frame.
Sex workers and their partners have played an important role in the HIV epidemics of many countries across the world where heterosexual transmission is the main mode of transmission. In these countries they are considered a core group, in that they have a high prevalence of HIV infection, and contribute disproportionately to the transmission of HIV due to their large number of sexual partners. In these countries sex workers form an identifiable population that has high HIV prevalence rates, before HIV is established among the general population. In Abidjan, Côte d'Ivoire, HIV prevalence rates reached high levels among female sex workers at 38% in 1986 [1], 68% in 1990 [2] and 80% in 1992-1994 [3], well before HIV prevalence rates increased among antenatal clinic attenders with 3% in 1986 and 13.8% in 1999 [1,4]. Testing of stored sera from Nairobi, Kenya, gives an even better picture of how HIV infection affected female sex workers before spreading among the general population. HIV prevalence increased among female sex workers from 4% in 1981 up to 82% in 1983, well before the increase among pregnant women from 0% in 1981 to 2% in 1985 [5] and up to 25% in 1995 [6]. In Bangkok, Thailand, HIV prevalence rates among brothel-based female sex workers increased from 0% in 1985-1986 [7] to 3.1% in 1989 [8], to 13.6% in 1991, and 33.3% in 1992 [9]. HIV prevalence rates among antenatal clinic attenders increased later with 0% in 1989, 0.6% in 1991, and 1.1% in 1992 [9].
In regions where HIV infection is rare among female sex workers, surveillance of risk behavior and sexually transmitted infections (STI) will indicate the potential for spread of HIV infection [10].
The objectives of conducting surveillance for HIV among sex workers are to monitor the extent to which sex workers are already infected with HIV, to serve as an early warning system for the potential for spread of HIV to the general population, to aid in planning and designing appropriate interventions for sex workers and their sexual partners, to monitor the impact on sex workers of the national response to the HIV epidemic including interventions targeted at sex workers, and to provide information to help mobilize an increased response to HIV. Declining trends in HIV and STI prevalence and rising trends in safer sex behavior can help increase public and political support for HIV prevention activities, whereas trends in the opposite direction argue for renewed efforts to promote safe behavior among sex workers and their sex partners.
This paper will review how surveillance for HIV infection has been conducted among sex workers. General principles for conducting HIV surveillance among sex workers will be discussed as well as the differing data needs according to the level of the HIV epidemic. Methods used in institution-based surveillance in different settings and community-based surveillance will be considered with their strengths and weaknesses. HIV testing issues for HIV infection surveillance in sex workers and surveillance needs in clients of sex workers will be briefly mentioned.
Female sex workers can be defined as women who exchange sex for money or goods. However, some female sex workers are more visible and identifiable than others. For example, in Burkina Faso in addition to highly visible professional sex workers, women serving in bars and cabarets, women selling fruit and vegetables, and students were included in a peer education program as occasional sex workers [11]. Similarly, in many Asian countries, direct female sex workers, who work in brothels and have no other occupation than sex work, are distinguished from indirect sex workers, who work in massage parlours, bars, and other establishments and may sell sex to supplement their income [12]. HIV surveillance among sex workers needs to take these differences into account, by precisely characterizing which groups of sex workers are included in the surveillance. Different types of sex workers will need to be accessed with different methods, as discussed below. A correct interpretation of trends over time will also need to take these differences into account.
The sex work scene differs greatly from place to place. There are differences in legal status, in work settings, in sexual practices, and in social acceptance. HIV surveillance methodologies and approaches therefore need to be adapted to the local situation. To facilitate access to sex workers and to ensure proper use of the results of HIV surveillance, alliances need to be forged. Such alliances need to be built with sex workers' groups, with authorities (police, health authorities, social services), with gatekeepers (brothel owners and pimps), with groups and institutions representing women's interests [non-governmental organizations (NGOs), the Ministry of Women's Affairs, family planning associations, human rights groups], and with specific intervention programs. These organizations and groups may facilitate access to the sex worker populations either because of their legal mandate (police, health authorities) or because they have already established a working or social relationship with sex workers. For example, in a study in the Gambia the co-operation of the bar owner was a decisive factor in selecting the study sites [13]. Working with established peer educators is particularly valuable as it may contribute to reducing suspicion about potential harm to sex workers. Meeting with the sex workers and explaining the purpose of surveillance, incorporating their expertise and utilizing them as adjunct researchers is highly recommended. In a study in Italy, access to sex workers for behavioral surveillance was facilitated through the Italian Committee for Civil Rights of Prostitutes, an organization of sex workers [14]. Providing stakeholder organizations with feedback on the results of the surveillance not only engages their interest and co-operation but contributes to a heightened sense of the need for prevention. One objective of HIV surveillance is to aid in planning specific interventions. Therefore no surveillance should be conducted without collaboration with existing interventions, or if no interventions exist yet, without the intention to create services based upon the results of the first surveillance results. The process of initiating surveillance activities and consolidating the HIV prevention services that go with them may in itself strengthen the partnerships and build up the capacities needed to expand surveillance activities in the future. The stable partners of sex workers may be at particularly high risk for HIV infection, but they also risk infecting their partners sex workers, as condom use in these couples is typically low [13,15]. Surveillance and services for sex workers should include their stable partners, whenever possible.
Sex work is either illegal or socially undesirable in most countries. Conducting surveillance among sex workers may entail dangers, including the breach of confidentiality of individual results, and negative reactions to publicized results by site, nationality, or other characteristics that may identify a community. Several examples of the negative impact of both official and unofficial surveillance among sex workers have occurred in Bangladesh. In one, unofficial testing led to the incarceration of HIV-seropositive sex workers through a breach of confidentiality. In another, after the official dissemination of surveillance results, social welfare agencies and police forcefully evicted women from the surveyed brothels, spreading them through the city and compromising existing HIV prevention efforts. These dangers need to be carefully considered in the planning and design of surveillance activities.
Several types of information can be collected as part of an HIV surveillance system. These include AIDS and HIV case reporting, HIV infection surveillance, and surveillance for behaviors and STI.
Reporting of AIDS and HIV cases is not a good HIV surveillance method for sex workers. In the general population there is already considerable under-reporting of AIDS cases, especially in developing countries [16]. This is largely due to weak health information systems. The limited access to antiretroviral therapy in most developing countries may further contribute to under-reporting, as access to antiretroviral drugs can constitute an incentive to come forward for testing. Moreover, health service providers who diagnose AIDS or HIV cases are unlikely to report that the individual is a sex worker - except for those providing specialized services for sex workers, either because sex workers do not self-identify, or because of the health workers' fear of negative reactions. However, some countries collect sex worker status for AIDS case reports, as for example in Honduras, where nearly one-third of AIDS cases reported up to November 1997 were among female sex workers [17].
The information gained from surveillance of HIV infection, STI and behavior will serve different objectives depending on the level of the HIV epidemic [10]. In low-level epidemics HIV infection has never spread to significant levels (< 5%) in any sub-population, although HIV may have existed for many years. In this situation it is important to monitor the levels of HIV infection among sex workers, as this group is likely to be one of the first groups where HIV infections will appear. For example, in Mexico the sentinel surveillance program among female sex workers has registered HIV prevalence rates ranging from 0.4 to 1% between 1990 and 1996 [18]. In addition to HIV sero-prevalence, data on behavior and STI will indicate the potential for spread to sex workers. In the Philippines, where HIV prevalence exceeded 1% among registered sex workers in seven of ten sentinel surveillance sites in 1997, consistent condom use was reported by 48% of registered female sex workers and by 28% of freelance female sex workers [19]. A study in a brothel in Bangladesh found a prevalence of 28% of infection with either N. gonorrhoeae or C. trachomatis, whereas the HIV prevalence was 0% [20]. In Bolivia, although HIV sero-prevalence among female sex workers was just 0.1% in 1995, the prevalence of gonorrhoea had decreased from 25.8% in 1992 to 9.9% in 1995 [21]. These examples indicate that, while HIV prevalence among sex workers is low in these countries, there is a great need to strengthen preventive programs targeting sex workers, to keep HIV prevalence low.
In concentrated epidemics HIV infection has already reached a significant level (> 5%) in a sub-population, but remains at less than 1% in pregnant women in urban areas, indicating that it has not spread widely to the general female population. In concentrated epidemics, the surveillance for HIV infection among sex workers allows for the monitoring of any additional spread of HIV to the sex workers group, while data on STI and behavior will indicate both the potential for further spread to sex workers, as well as the potential for spread of HIV from the sex workers to their clients, who may in turn serve as a bridging population to the general population. For example in Dakar, Senegal, HIV prevalence among sex workers increased from 9.8% to 17.3% between 1989 and 1996, whereas the prevalence of N. gonorrhoeae decreased from 17.4% in 1991 to 4.7% in 1996, and 94% of sex workers reported condom use with their last client in the four largest cities of Senegal, including Dakar, in 1997. The increased HIV prevalence among sex workers indicates potential for spread of HIV to the general population, although the behavioral and STI data suggest that important spread of HIV to the general population is unlikely. Indeed, the HIV prevalence rate among pregnant women remained stable at 0.7-1.2% between 1989 and 1996 [22]. Similarly, in Thailand, the proportion of sex acts where condoms were used increased from 25% in June 1989 to 92% in June 1994, indicating the success of the government's 100% condom use campaign [23].
In generalized epidemics where HIV has surpassed > 1% in pregnant women, surveillance of HIV infection among sex workers is not as critical as it is in low-level and concentrated epidemics, because HIV is already firmly established in the general population. However surveillance for behavior, STI and HIV among sex workers is important as the results may indicate success or otherwise of the national response to the epidemic, including targeted programs to encourage safe sex between sex workers and clients. For example in Abidjan where HIV prevalence was 9.1% among pregnant women in 1997 [4] the proportion of sex workers reporting condom use during their most recent sex act increased from 63% in 1991 to 91% in 1997 [24].
Both behavioral and serologic information on HIV can be collected through different designs, i.e. cross-sectional surveys, repeat sentinel sero-surveys, and prospective studies. Surveillance among sex workers should be conducted on a regular basis every year. Repeat cross-sectional studies represent the preferred design both for HIV infection surveillance and for behavioral surveillance. HIV infection surveillance can be conducted at institutions or in the community, as discussed below.
The interpretation of time trends of HIV infection is difficult, as, in all countries, sex workers are known to be notoriously mobile. It is essential that the changing composition of the group is tracked as the interpretation of time trends can easily be confounded by changes in the socio-demographic composition of the group. It is therefore necessary to collect socio-demographic and behavioral data when conducting HIV surveillance.
Two approaches are possible. The first is to collect comprehensive socio-demographic and behavioral information from the same individuals who are providing specimens for HIV testing. This can be carried out at clinics that provide voluntary confidential HIV testing, where results are linked to the sex worker and shared with her. However most clinical settings produce poor results regarding behavioral data collection. Furthermore HIV infection surveillance conducted at clinics that provide voluntary confidential HIV testing is subject to both selection bias and refusal bias.
The second approach is to conduct behavioral surveillance separately from HIV infection surveillance. Separate behavioral surveillance among sex workers has been successfully conducted in several countries in Asia [25,26]. Where sample frames are carefully constructed to include sex workers of all categories, the behavioral data that is collected should be fairly representative of all sex workers. Behavioral surveillance may include the following socio-demographic and behavioral indicators: age, education, type of work site, length of time at present site, length of time in sex work, number of clients per week or in last 24 h, country/region of origin, language, condom use with last client, condom use with last non-client, type of sexual practice, substance abuse, other occupation, income, exposure to interventions, and sexually transmitted diseases (STD) treatment-seeking behavior. Some behavioral information may measure risk for HIV infection, e.g. sharing equipment to inject drugs or inconsistent condom use. Indeed, some sex workers may have additional risk as they may also inject drugs, especially in Europe and North America [27,28]. The behavioral surveillance data needs have been discussed in greater detail elsewhere [29]. Data on ethnic identity should be handled carefully as they can lead to stigmatization [15]. In situations where unlinked anonymous HIV testing is conducted, it is not possible to collect comprehensive socio-demographic and behavioral information linked to the samples, because the collection of these data could compromise the anonymous nature of the surveillance. A bare minimum of socio-demographic variables, such as age, length of time in sex work, education, and country/region of origin may be linked to the samples. Comparison between the distribution of these socio-demographic characteristics in the HIV infection surveillance data set and the separate behavioral surveillance data set may make it possible to determine whether the sample of women included in the HIV infection surveillance is representative of the wider sex worker population.
In recently published guidelines for STI surveillance WHO and UNAIDS recommend periodic prevalence assessments among sex workers of syphilis, gonorrhoea, C. trachomatis infection, trichomoniasis and genital ulcers [30].
HIV infection surveillance for sex workers can either be conducted at institutional services and settings, or in community-based surveys. In comparison with surveillance using a community-based sample, surveillance in most institutional settings is cheaper and easier. Surveillance in institutional settings is less subject to refusal bias (especially where blood or other specimens are available for unlinked anonymous surveillance), but more subject to participation bias, compared with community-based surveillance.
The institutional services and settings in which sex workers can be reached can be placed in three categories. These are screening programs for registered sex workers, STI clinics (including mobile clinics), and reeducation camps. In all of these settings incentives should be provided including sexual health information, condom skills demonstrations, and condoms, and, if not already provided, referral for voluntary HIV counseling and testing as well as STI diagnosis and treatment.
Systematic sampling may be conducted during a specific time period when all sex workers in the institution are included in the surveillance exercise. For reasons of comparability it is important that the surveillance is conducted during the same period every year. The preferred option is surveillance based on voluntary HIV counseling and testing, although this may introduce refusal bias.
Alternatively unlinked anonymous HIV testing may be used where left-over serum is available after testing for other conditions, especially for syphilis. Even where HIV infection surveillance is based on unlinked anonymous testing, it is important that voluntary HIV counseling and testing is also made available for women who want to know their HIV status.
Screening of registered sex workers on a regular and regulated basis usually takes place at a dedicated STD clinic [22]. The register may serve as the sampling base for surveillance, or sex workers presenting to the screening facility may be systematically, e.g. consecutively, sampled. In principle all sex workers are registered; however in practice many are not [22] and the ones on the register may have substantially different characteristics in comparison with those not on the register. Non-registered sex workers may be more likely to be young, work outside establishments, be foreigners, and generally be at higher risk than the registered sex workers. In some Asian and Latin American countries HIV-positive sex workers may be excluded from the registration system. This may lead to additional bias as these women may continue to work outside the registration system.
Sexually transmitted diseases clinics may provide services to sex workers, either because sex workers perceive them as providing good quality services [14], because of their location in or near red light districts, or because they are dedicated clinics for sex workers [3]. In some settings, private physicians who provide services in sex work settings may be approached for collaboration. The universe of sex workers that can be reached through STD clinics corresponds to the catchment area and population of these clinics. However, symptomatic women may be over-represented at STD clinics. Depending on the quality of health services and whether or not they are free, relatively poor or relatively wealthy women may be over-represented. At satellite or mobile clinics and fixed facility brothel clinics there would be no bias linked to the distance to the clinic. Furthermore, illegal sex workers may avoid STD clinics for fear of coming into contact with the legal or other government authorities. Sex workers may also avoid STD clinics because of the stigma that may be attached to them and because they do not wish clients or other sex workers to know that they have an infection.
In addition to the above-mentioned key measures, the reason for the visit may be collected, for example, whether it is because of symptoms, partner referral, or a routine screening visit. It is indeed important to know whether the woman has STI symptoms because the presence of an STI indicates inconsistent condom usage.
Re-education camps were included in this discussion because they are currently used in some countries for surveillance among sex workers, e.g. in China and Vietnam [31]. In principle any sex worker may pass through these education camps; however in practice the poorer ones are more likely to be arrested by the police and taken to these camps. Surveillance based in these camps may reach only a small fraction of sex workers. It may be biased through an over-representation of women of low socio-economic status, of rural origin, or of low education status and this selection bias may change over time.
Data from re-education camps are difficult to interpret, and it is recommended that use of this institutional setting for surveillance be phased out. A major danger of conducting surveillance at rehabilitation centers is the potential for breach of confidentiality [31]. Other ethical issues are of importance as re-education camps are coercive environments just like prisons, and individuals may not be free to willingly participate in surveillance. Another danger of continuing to conduct surveillance at re-education camps is that the opportunity to develop a better surveillance system is missed.
Community-based surveillance may initially be conducted where no services are available. However, if this is the case, then services need to be established, informed by the results of the initial surveillance exercise. Community-based HIV infection surveillance among sex workers is made easier where an intervention has successfully motivated the women to become active preventionists. In Bangladesh, street-based sex workers have participated in HIV surveillance for several years through clinics associated with their intervention. Organizations that are run by sex workers have spearheaded this involvement which takes place with informed consent, although the HIV testing is unlinked and anonymous. In this way feedback can be given to the entire group when results are available, but there is no chance that individual confidentiality can be breached.
As the aim of community-based HIV surveillance is to be able to reproduce comparable samples in repeated surveys over time, the most robust method is to select the survey population randomly from a sampling frame [32]. Where the community is relatively small, systematic sampling can be used after mapping [33].
Incentives should be provided to sex workers who participate in community-based surveillance. The provision of incentives can increase the participation rate and is part of an effective intervention. They may include sexual health education, condoms, referral for voluntary HIV counseling and testing, referral for legal services, STD diagnosis and treatment for symptomatic women, syphilis screening, and, perhaps in some settings, financial compensation for the time devoted to participation in the survey.
A random selection of sites will be made from the sampling frame based on mapping. Sampling procedures should be carefully documented to ensure comparability across rounds of surveillance. If there is an interest in how behavior and/or HIV infection rates differ between categories of sex workers, a stratified sample must be drawn, e.g. a separate sub-sample for each type of sex worker. The actual survey should then be conducted at the selected sites. No matter which type of sampling is used, when survey personnel return to a previously mapped and selected site, they must record how many women were actually seen at that time, how many left before providing a biological sample, how many refused and how many were actually surveyed. In this way, if self-weighting is not possible, weighting can be accomplished during the analysis. In most instances it is preferable for the purpose of HIV infection surveillance to accompany sex workers to a nearby clinic where a biological specimen can be collected. Experience in many places shows that simply referring them to a clinic is not sufficient as some may come who were not selected and others who were selected may refuse to come. As discussed below, biological samples that require less invasive specimen collection methods, such as fingerprick bloodspots, saliva, and urine, may be collected at the places where the sex workers are contacted, - on the street or in hotels and bars [13,33].
The dangers of conducting community-based HIV infection surveillance include the breach of confidentiality of individual HIV test results, negative reactions of the government or the public, disclosing the location of 'illegal activity' to the authorities, all of which constitute unethical acts against sex workers and must be avoided. For the purpose of community-based studies, all settings in which sex work occurs or at which sex workers meet their clients need to be taken into account and the sampling universe defined. In terms of access, three broad categories of sex work sites can be considered, namely brothels, other establishments (including hotels, dance halls, bars, video houses, karaoke bars, massage parlours), and non-establishment-based or 'floating' sex workers (those working in the street, freelancers). Although it is useful to categorize sex work sites in this way, it is recognized that there may be some overlap between these categories. Floating sex workers, in particular, may be the same women who work in hotels or bars. Formative qualitative research is required before beginning the mapping in order to sort this out. If a large proportion of street-based sex workers who access clients on the streets in fact go into hotels to perform sex, then a serious level of overlap is likely. All aspects of working with illegal sex workers requires their co-operation and this is best accomplished where NGOs or other types of organizations are already offering services. In such cases, peer educators or others can explain to the women both the purpose of the survey and how it will be carried out. They can request their co-operation. If services are not yet available, then guides must be selected from among the sex workers who can introduce the survey personnel and explain the purpose of the survey. As mentioned earlier, this is only ethical if services, both for behavior change and STI treatment, will be made available immediately following the survey.
The dangers associated with mapping and surveying illegal sex workers requires some degree of protection for staff. This may be offered through alerting the local police, local gang leaders, or other local authorities. Sex workers should always be involved directly as guides or mappers or even as interviewers in behavioral surveys, if their literacy is adequate.
The collaboration of the brothel-owner or manager may be useful to gain access to brothel-based sex workers. Sampling at brothels may be biased if the owner/manager refuses access, or if individual sex workers refuse to participate. There is a danger of a negative reaction from the owner or manager if sex workers from a brothel are recruited for surveillance without his/her knowledge and the owner/manager later finds out.
The collaboration of the owner/manager may be sought, or sex workers and/or ex-sex workers may provide access to sex workers in other establishments. Potential sources of bias include the exclusion of women who sell sex at these sites but do not think of themselves as sex workers, and changes in the population over time; for example following the expulsion of foreign women. In addition, where peer educators contribute to the mapping and sampling process, it is possible that women who have already had contact with peer educators and had access to prevention information and/or condoms will be over-represented in the sample.
It is more difficult to gain access to floating sex workers than to sex workers in the above two categories. NGO or other community-based organizations with good rapport and who offer services to these women can often be helpful. Again, sex workers serving as key informants are essential collaborators. Potential sources for bias are similar to those of other establishments, but as the sex work scene is more fluid, the biases are more difficult to measure and characterize.
The cost of community-based studies is high in comparison with institution-based studies. The cost is higher in other establishments compared to brothels, and is likely to be highest for floating sex workers.
To allow for comparability of data collected in different years of surveys, there is a need to avoid major events/manifestations that may influence the sex workers' scene, such as major holidays or conventions.
Collecting blood for HIV testing has the advantage of allowing an additional test for syphilis, which is an infection of particular concern for sex workers. Treatment for syphilis represents both a significant health benefit and an effective intervention to reduce the further spread of HIV. Moreover, in the case of anonymous unlinked HIV testing, the blood drawn for syphilis testing provides the portion needed for HIV testing.
Dried blood spots on filter paper may be collected through a fingerprick. This method of collecting a blood sample is easier and more acceptable than collecting venous blood. However a much smaller volume is collected, limiting the opportunity for confirmatory tests, additional testing for HIV subtype or for other infections, e.g. syphilis, hepatitis B and C.
Saliva is easy to collect and non-invasive. Saliva samples are therefore convenient for collection in population-based studies [13,33,34] but also limit opportunities for additional types of testing.
Urine has not been used much yet for HIV infection surveillance. However urine is easy to collect and has the added advantage of allowing testing for other STI, i.e. N. gonorrhoeae, C. trachomatis, T. vaginalis, and, in the case of anonymous unlinked testing for HIV, the urine collected for STI testing or for pregnancy tests allows testing for HIV.
Saliva and urine samples are safer to collect than blood spots or venous blood, especially where HIV sero-prevalence is likely to be high and the risk of needle-stick injury substantial. Non-invasive specimen collection is also likely to be more acceptable to the individuals among whom surveillance is carried out.
Clients of sex workers are a potential bridging population. Their contact with sex workers may put them at risk for HIV infection, but they may also transmit HIV to sex workers and their other sex partners who are not sex workers. Clients of sex workers should therefore also be the subject of surveillance. Surveillance of clients of sex workers has usually been indirect, by including population groups that are a priori believed to have sexual interactions with sex workers, including truckers, miners, the military, and STI clinic clients [35]. Few studies have attempted specifically to study behavior and/or HIV infection levels among clients of sex workers, identified at sex work sites. These special studies can better define who makes use of the sex workers' services and allow for a more direct measure of risk behavior and HIV and STI prevalence in this bridging population [13,36,37].
Surveillance of HIV infection among sex workers is critical for countries with low-level or concentrated HIV epidemics, and can help in monitoring the response to the HIV epidemic in countries with a generalized epidemic. Surveillance of STI and of risk behavior complement surveillance of HIV infection. Surveillance for HIV infection among sex workers can be institution-based or community-based. Specific sources of bias need to be considered and measured to allow a correct interpretation of prevalence data and time trends. Collaborations with key players in the local sex work scene - sex workers themselves in the first place - and alliances with salient institutions and groups are key to the success of surveillance and interventions for sex workers.
We acknowledge the contributions of the members of the working group on HIV surveillance among sex workers Michel Caraël, Rebecca Martin, Georg Pauli, Ly Penh Sun, Tobi Saidel, and Bernhard Schwartländer. We thank Jesus-Maria Garcia Calleja and Neff Walker for their review of the paper.
Mapping is carried out in order to produce a sampling frame, but also to prepare for future interventions or update information for ongoing interventions. The entire universe of all types of sex worker must be mapped in a systematic way. Sex work takes place in both public and private places and has both contact and action venues. These require different mapping and sampling strategies. In brothels, all used rooms may be counted and the number of women living in or operating from each room must be recorded. In clubs or massage parlours, the roster of women hired can be accessed from the managers. For street-based sex work, where only the contact venue rather than the action venue is accessible, the approach to mapping must be observational. If sex workers are scattered along a road, then a section of the road may be considered a sampling unit. The location of each sampling unit, the number of women seen at the time of mapping, the time and day of the week they are seen and any other pertinent information relating to their accessibility should be recorded. Ordinarily it is very time-consuming to return to each site to sample differences in time of day or night and day of the week. It is possible to use key informants who are always present at those sites to obtain this information, such as a local businessman or street merchant, and the obtained information should be triangulated, i.e. checked for consistency between different sources. It is best to conduct the mapping close to the time the surveillance will actually take place.
Caution must be exercised regarding the use of mapping to count the total number of sex workers. In brothels or contract sex -work establishments, as in some clubs or massage parlours, the numbers may be fairly consistent and predictable. However, for street-based sex workers mobility presents serious problems. If the mapping could take place on a single day with a very large team of people, then counts of available sex workers may be more accurate. Otherwise, the high mobility of sex workers in most cities means there will be a lot of double-counting. Double-counting can be avoided either by recording a unique identifier for every sex worker or by giving sex workers a card or voucher for some sort of service to be shown later if re-captured during the mapping exercise. Capture-recapture estimation procedures can then be used to estimate the size of the street-based sex worker population [33].
This supplement is sponsored by the Ministry of Health, Federal Republic of Germany, and the joint United Nations Programme on HIV/AIDS (UNAIDS)
The papers in this supplement are based on a workshop organized by the Robert Koch Institute, Berlin, Germany, sponsored by the Ministry of Health of the Republic of Germany, the Robert Koch Institute, the Joint United Nations Programme on HIV/AIDS (UNAIDS) and the World Health Organization, and held in Berlin in November 1999
Female sex workers; surveillance; sexually transmitted infections; behavior; mapping