UNAIDS and WHO have recently classified HIV epidemics into three broad categories, namely low-level, concentrated, and generalized epidemics . In a low-level epidemic, HIV infection has never spread to significant levels in any sub-population, although it may have existed for many years. Infections are largely confined to individuals with higher risk behavior, often among groups such as sex workers, drug injectors and men having sex with men. This epidemic state suggests that networks of risk are rather diffuse and only low levels of partner change or sharing of drug injecting equipment exist, or that the virus has been introduced very recently. A numerical index that has been adopted for low-level epidemics is that HIV prevalence has not consistently exceeded 5% in any defined sub-population. In a concentrated epidemic, HIV has spread substantially in one or more sub-populations, but is not well-established in the general population. This epidemic state suggests active networks of risk within these sub-populations. The future course of the epidemic is determined by the frequency and nature of links between highly infected sub-populations and the general population. The numerical index for concentrated epidemics is that HIV prevalence is consistently over 5% in at least one defined sub-population, although it is below 1% in pregnant women in urban areas. In a generalized epidemic, HIV is firmly established in the general population. Although sub-populations at high risk may continue to contribute disproportionately to the spread of HIV, sexual networking in the general population is sufficient to sustain an epidemic independent of sub-populations at higher risk of infection. The numerical index for generalized epidemics is that HIV prevalence is consistently over 1% in pregnant women.
In many countries HIV infection is concentrated in sub-groups of the population whose behavior exposes them to especially high risks of acquiring or passing on HIV. Therefore, HIV surveillance should focus on these sub-populations. They include injecting drug users, men who have sex with men, sex workers and their clients, and highly mobile populations. They are people who may be avoiding contact with government authorities and service providers because they are engaged in illegal activities, or because they are on the move for various reasons, or because they do not choose to disclose their activities for personal reasons. They may be hard to reach for the purpose of surveillance, as the surveillance system may be perceived to be a part of the government authorities and service providers. HIV surveillance and prevention among hard-to-reach groups are no easy matter. However, in low-level and concentrated HIV epidemics where there is virtually no HIV in the general population, it may be the only surveillance worth doing. HIV surveillance in high-risk groups should be the cornerstone of the surveillance system in a majority of countries worldwide. Even in generalized HIV epidemics, surveillance in hard-to-reach populations continues to be important to monitor the success of interventions targeted at these groups. Indeed, if no impact is being made where the risk is highest, the interventions are unlikely to have a wider impact on the general population.
The extent to which HIV will spread from these sub-populations to other groups depends on patterns of sexual networking and condom use. An important function of behavioral surveillance in hard-to-reach populations is to identify sexual links that have the potential to carry HIV infection from a sub-population with identifiable high-risk behavior into a larger population with no especially high risk. Another important function is to identify the sub-groups that are vulnerable to HIV and the behaviors most likely to spread it, and to use this information to design appropriate interventions with the participation of these vulnerable groups. There is no point in doing surveillance for HIV unless it leads to improved information and services that will help people with high-risk behaviors to reduce their own and their sexual or injecting partners' exposure to the virus. Because HIV surveillance systems measure HIV infection and related behaviors over time, surveillance in marginalized groups can help to assess the success or failure of a nation's efforts to promote safer behavior among those at highest risk of contracting or passing on HIV infection. Positive trends can help increase public and political support for HIV prevention activities, whereas negative trends argue for renewed efforts to promote safe behavior in the populations in question.
Four papers in this supplement discuss HIV surveillance in specific hard-to-reach populations. The sub-populations include mobile populations with a focus on forced migrants , injecting drug users , men who have sex with men , and female sex workers . For each of these groups the authors review the strategies and approaches that have been used in HIV surveillance. They also address the major challenges of HIV surveillance in these populations, including defining the population, gaining access to the population, sampling, dealing with bias, and ethical challenges.
HIV surveillance in any population must begin with a definition of that population. Although this seems obvious, it is not always easy to do in hard-to-reach populations. This is in part because many people who engage in high-risk behavior do not consider themselves to be part of any particular sub-population. Individuals with high-risk behavior do not always belong to readily identifiable groups. Still, unless a population sub-group can be identified and approached, HIV surveillance cannot be undertaken. The difficulty of defining sub-populations at high risk for HIV transmission under-scores the importance of formative behavioral research. Formative behavioral research - which has been conducted in many countries but is frequently under-used in planning surveillance systems - is critical in helping to define the population of interest. The use of key informants, snowball recruitment, and mapping of sites where risk behavior takes place will help to include a broad spectrum of the sub-population into surveillance systems.
During the process of characterization of a hard-to-reach population, access points are usually identified that will allow surveillance activities to be conducted in these populations. Ideally, access is provided through existing services which meet the health or welfare needs of the hard-to-reach population in question. This guarantees that individuals participating in surveillance will receive services that improve their well-being, and that may reduce their risk of acquiring or passing on HIV. In some settings, outreach workers have been used to advertise the provision of health and welfare services, including screening and management of sexually transmitted infections, counseling and voluntary HIV testing. In many settings, outreach workers have been able to increase attendance at such services by hard-to-reach populations. This increases the utility of such services as sentinel sites for HIV surveillance. Outreach workers providing HIV prevention information and referring for services have also been used in recruiting volunteers for population-based HIV surveillance. Whichever access point is used for surveillance in marginalized sub-populations, careful ground work must be undertaken before surveillance can begin. Alliances must be formed with key actors - including members of these groups - who have the power to determine whether a prevention programme, and the surveillance that accompanies it, succeeds or fails.
As most hard-to-reach populations live on or beyond the limits of the law and all by definition are frowned upon by society, HIV surveillance in these groups is potentially perilous. One danger is that the very process of surveillance alerts law enforcement agents to populations, individuals or locations where illegal behavior is taking place. Another is that findings of high HIV sero-prevalence in a marginalized population engenders a public backlash, fostering repression and increasing marginalization. Any such negative effects would make it more difficult to reach these populations with prevention programmes. Publicity surrounding high HIV prevalence rates among refugees from conflict situations may, for example, seriously jeopardize their chances for resettlement. Surveillance in hard-to-reach groups may also undermine prevention efforts in the general population, as the HIV epidemic becomes associated in the public mind just with one or two specific groups. The illegal status of some groups may make it logistically difficult, as well as potentially dangerous, to undertake surveillance.
There are three major reasons why a surveillance system should attempt to estimate the overall size of a marginalized population whose members are at high risk of HIV. First, it is hard to gauge the representativeness of the sample included in surveillance (or even the sample frame used for subject selection) without an idea of the overall size of the sub-population at risk. Second, the results of HIV surveillance are often used to estimate the overall prevalence of HIV in a country, and the number of people infected. Good sentinel or population-based surveillance systems will give a prevalence rate in each sub-population at high risk for HIV, but these can only be turned into a national estimate if the size of each sub-population can be estimated with any confidence. Third, and most important, the scale of necessary interventions to reduce the spread and impact of HIV will depend on the size of the population at risk. Such estimates are therefore vital for programme-planning purposes. One method for estimating the size of the sub-population is mapping of the sites where high-risk behavior takes place, along with an estimation of the number of individuals associated with each site. A statistical method that has been used to estimate the size of various sub-populations at high risk of HIV infection is the 'capture-recapture' method. However this method is fairly labor intensive, and assumes a certain predictability of contact and independence of the samples that may not be fulfilled by the often extremely fluid and diverse sub-populations at high risk for HIV infection. A fifth paper in this supplement by Archibald et al.  describes an innovative method for estimating the size of hard-to-reach groups in Canada.
The sixth and final paper by Respess et al.  gives an update on HIV testing for surveillance in hard-to-reach groups. It discusses HIV testing strategies for use in unlinked anonymous testing used at sentinel surveillance sites, as well as strategies that can be used when HIV testing results are shared with study participants. It discusses these strategies in the light of the specific field conditions encountered when accessing hard-to-reach populations. Much weight is given to the recently increasing availability of rapid tests.
Data from a diversity of different sources contribute to a stronger surveillance system for HIV and the behaviors that spread it. It is, however, critical to note that surveillance is not an 'all-or-nothing' activity. It is not usually possible - for financial, logistic or political reasons - to introduce simultaneously a full range of data collection activities that covers a representative sample of every sub-population with high-risk behavior for HIV transmission. Compromises must be made between the ideal and the practical. Precisely because the populations of interest are hard to reach, coverage is likely to be incomplete and sampling may not be entirely representative. Despite these limitations, surveillance data from these groups are critical, especially in low-level and concentrated epidemics. The feasibility of carrying out high-quality surveillance in hard-to-reach populations will increase over time if the members of a sub-population recognize that the results of the surveillance are being used to improve the services available to them. Such services should, at a minimum, support individuals with risky behavior in reducing their own or their partners' exposure to HIV.
The papers in this supplement represent a synthesis of issues raised at a workshop held in Berlin in November 1999 on HIV surveillance in hard-to-reach populations. The workshop was organized by the Robert Koch Institute, Berlin, Germany, and sponsored by the Ministry of Health of the Federal Republic of Germany, the Robert Koch Institute, UNAIDS and the World Health Organization.
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