WHAT IS BEHAVIOURAL SURVEILLANCE?

Behavioural surveillance is the ongoing systematic collection, analysis, and interpretation of behavioural data relevant to understanding trends in the sexual transmission of infection.²⁸ This should be followed by timely dissemination of these data to those responsible for prevention and control. Knowledge of the size of the population groups at risk, and the nature and determinants of risk within those populations are necessary. Behavioural surveillance generally aims to monitor trends in two broad groups of indicators; firstly, those that allow the identification of population subgroups at increased risk—for example, age, sex, sexual orientation, and ethnicity. Secondly, those behaviours that are amenable to change—for example, number and type of sexual partnerships, condom use, unprotected anal intercourse. The validity and reliability of sensitive data on behaviour are critical as they are self reported and can’t be directly measured.²⁹ The triangulation of a small set of core measures selected from surveillance data and other complementary sources can strengthen the interpretation of these data as the relation between sexual behaviour and STI transmission is complex.

Any attempt to establish behavioural surveillance in England and Wales should therefore seek to answer the following questions: which behaviours are important determinants of current STI and HIV transmission? How are these behaviours distributed and how can they be measured over time? What key behavioural data are not currently being collected? How best can these gaps be filled?

HOW MAY IT BE ACHIEVED?

General population surveys

Behavioural surveillance is generally conducted at two levels, among the general population and within targeted risk groups. General population surveys are useful in assessing overall trends and distribution of behaviours that may be associated with STI transmission. These provide the most robust estimates of prevalence of behaviours, as they largely avoid the biases inherent in most targeted population surveys. Although regular repeated surveys are needed to measure changes in behaviours over time their expense may make this difficult. Adding additional questions to existing population social surveys is a method that has been successfully deployed in other countries³⁰ as a cost effective way of getting population based estimates. This has been suggested for collecting sexual behaviour data in the United Kingdom.²⁹ A large number of surveys are currently carried out which could be used in this way.^31,32 This kind of survey makes it possible to access a general population sample, but does limit the number of questions that can be asked.

General population surveys are usually less suitable for obtaining detailed information on population subgroups at highest risk. These groups tend to be small, more clustered, and difficult to access and small subgroups of individuals with relatively rare risk behaviours may not be captured in sufficient numbers. Groups of particular interest for HIV and STI transmission include homosexual and bisexual men, injecting drug users, commercial sex workers, and ethnic minorities, particularly those from or who have contact with countries with a high HIV/STI prevalence. These problems can be overcome through adapting study designs to include oversampling and focused enumeration.¹³

Targeted population surveys

Targeted population surveys are also a useful adjunct to these general population surveys as they give greater detail on populations at highest risk. However, the difficulty in accessing these populations makes probability sampling costly. More cost effective sampling strategies are needed; these can include advertising, snowballing, recruiting from GUM clinics, and social and commercial venues. However, these strategies may result in a sample selection bias and decreased representativeness of results. Targeted behavioural surveillance can include serial cross sectional surveys, using the same sampling strategy and using core questions to ascertain the prevalence of risk behaviours.

The disadvantage of targeted population surveys is that they are likely to be unrepresentative, given the nature of the convenience sampling. Those accessed through this mixture of social venues can only be representative of those using these sites. In addition, even among venue attenders the behaviour of study respondents may systematically differ from non-respondents. In order to overcome this problem, surveys from a range of settings are needed, in order to achieve a more representative sample. New and innovative ways of accessing these populations are needed—for example, accessing MSM through internet chatrooms.^33,34 Cross comparability of surveys done in different populations accessed through different means will allow an overview picture of the distribution of behavioural risk within the population under investigation. Questions that will allow the linking of the populations will enhance the interpretation of the individual surveys.³⁵

Behavioural surveillance in England and Wales: assessing the existing capacity

Disease surveillance

Current surveillance systems collect limited data on the behavioural determinants of STI transmission. Where they exist they are often limited to facilitate ease of completion by busy clinical staff. Most systems rely on methods more focused on disease outcome, practicality, uniformity, and rapidity rather than on obtaining full demographic and behavioural details. Generally, the additional data collected are minimal (typically age, sex, sexual orientation) (table 1). These allow the grouping of diseases by risk factors, although clearly these are not behaviours amenable to change. Some enhanced surveillance systems have been developed that include more detailed behavioural data to allow the characterisation of those with diagnosed infections^36,37 (table 1) For example, the enhanced KC60 surveillance system will not only allow more risk factor information to be collected on an individual basis, but will also allow rates of co-infection and re-infection of STI to be examined and core groups to be more accurately described.^3,38

There is comprehensive national surveillance of AIDS cases and diagnosed HIV infections.^39–41 This surveillance system has recently been enhanced, and now clinicians are also asked to report all newly diagnosed HIV infections. The new clinician HIV and AIDS report form collects more behavioural data at the time of first HIV diagnosis (table 1) and provides the most comprehensive picture of all surveillance systems.

The unlinked anonymous HIV seroprevalence surveys provide sentinel HIV prevalence data and have been ongoing since 1990.⁴² Limited demographic and behavioural data are collected with the unlinked residual specimens following clinical tests. The surveys cover both those at higher risk of infection, such as homosexual men and heterosexuals attending GUM clinics and injecting drug users attending services, and a more general population sample through monitoring HIV prevalence in over 60% of all pregnant women. The survey of injecting drug users differs in that a voluntary saliva sample is provided with a self completed questionnaire detailing demographic, sexual, and drug injecting behaviour. This survey represents some of the most detailed sexual behaviour data collected within the existing surveillance systems.⁴³

Data from the National Blood Service (NBS) provide prevalence information in a lower risk population group, as the criteria for donation excludes those at increased risk of blood borne infections, including men who have had sex with men, those who have ever injected drugs, and those who have had heterosexual contact with high risk partners⁴⁴ (table 1). Laboratory reports for confirmed acute hepatitis B are also routinely collected nationally.^45,46

BEHAVIOURAL SURVEYS

Table 2 illustrates existing ongoing behavioural surveys carried out by different academic and research groups in Britain. Two general population surveys of adults are currently carried out. The first, Natsal, a probability sample study has been carried out twice a decade apart,^13,47 remains the largest probability sample study of its kind in Britain. The 2000 survey also collected and tested urine samples for genital Chlamydia trachomatis using ligase chain reaction (LCR) techniques to provide the first national prevalence estimates.⁴⁸ The second, the Omnibus survey is a multipurpose survey of the adult population routinely carried out by the Office for National Statistics. A module on contraceptive use and general sexual health including condom use has been included annually since 1997⁴⁹ (table 2).

A national survey of young people is currently being carried out by the Teenage Pregnancy Unit, as part of an evaluation of the teenage pregnancy strategy (table 2). An individual based tracking survey will be repeated three times a year to collect information from young people aged 13–21 and parents of young people aged 10–17 over a 3 year period. It will collect information on knowledge, attitudes, and behaviours around sex and relationships.⁵⁰

A number of annual surveys of homosexual men attending social venues,^51–53 GUM clinics,⁵¹ and Gay Pride events⁵⁴ are currently carried out (table 2). These use a stable set of behavioural indicators that can be monitored repeatedly. The three surveys developed and used a common set of core behaviour questions that allow comparisons of the three populations of MSM. A number of other surveys of injecting drug users^55,56 and among ethnic minorities^2,4,57,58 have also been carried out but none have been sustained. There is clearly a need for more ongoing investment and support to continue projects once established.

HOW DO WE USE BEHAVIOURAL DATA?

Behavioural surveillance data can be used in a number of ways. They can allow the monitoring of the risk behaviours underlying HIV and STI transmission over time. UNAIDS has recommended that behavioural data collection should be a central part of HIV and STI surveillance programmes.^28,59

A range of indicators can be used to measure the effectiveness of both HIV and STI prevention interventions in England and Wales. These include the behavioural determinants of disease transmission (for example, condom use, reported sexual partnerships) as well as disease incidence and prevalence in England and Wales. These “prevention indicators” have been developed to monitor four key areas relevant to HIV transmission and disease prevention and include HIV prevalence, HIV incidence, risk behaviour, and healthcare utilisation.⁴² The indicators for monitoring the success (or failure) of HIV prevention in men who have sex with men are illustrated in table 3. Similar indicators have been used elsewhere,^25,26,60 although the use of behaviour change as a proxy marker for STI incidence has raised debate.^61,62 The disproportionate effect of some factors on the transmission dynamics of STI means that reported risk behaviour doesn’t entirely correlate with transmission. The role of sexual networks in transmission is important and behavioural surveillance cannot always measure these. Prevention indicators have been evaluated in a number of settings, however, and found to be useful for measuring the success of prevention programmes, although multiple sources of data are necessary to provide context.⁶³ This in turn facilitates more effective HIV prevention and community planning. Prevention indicators may be developed using a variety of available data within ongoing surveillance systems. This allows the interpretation of HIV and STI trends within different population groups, and through the monitoring of risk behaviours, can indicate when outbreaks of infection may occur.⁶⁴

A potential research priority highlighted in the new national strategy for sexual health and HIV was a need for better understanding of the sexual networks, health seeking behaviour, and risk behaviour of targeted groups.⁶⁵ The monitoring of behavioural indicators within different population groups would provide data on both health seeking behaviours and risk behaviours. Behavioural surveillance could also measure progress towards increased HIV testing of GUM clinic attendees through monitoring HIV testing patterns in different population groups.

Finally, behavioural surveillance data will enable us to identify priority areas for further in-depth epidemiological or socioanthropological research. Much of this research should be developed in collaboration with local academic and service partners in the most vulnerable areas or population groups.

WHAT ARE OUR OPTIONS?

Behavioural surveillance programmes have now been implemented in the United States,^26,66 Switzerland,²⁵ Australia, and Hong Kong.⁶⁷ The United States has formed a HIV/STD Behavioural Surveillance Working Group to build and maintain a behavioural surveillance system for HIV and STI. They have achieved this through developing standardised measures of risk behaviours for comparability of data across systems and used these in monitoring a combination of general population, at-risk populations, and infected populations. Modules of questions have been provided at the national level for states to use as appropriate.⁶⁶ In addition, HIV prevention indicators have been developed, which have set out specific indicators suitable for monitoring at state and local level. Collection of data for these is coordinated at local level.

Canada has similarly combined national behaviour telephone surveys with more targeted behavioural surveys in homosexual men and injecting drug users (IDU) although they have not established nationally standardised modules of questions. Australia has used a combination of targeted behavioural surveys in MSM and IDU, from which key indicators are coordinated nationally with HIV surveillance and incidence data. They are currently moving towards national coordination of STI surveillance,⁶⁸ and the development of a coordinated national approach to collection of behavioural risk factor data. The first national survey of sexual health and sexual behaviour and attitudes administered through telephone interview is currently being carried out. Hong Kong has established a behavioural surveillance system, carrying out an annual general population survey of sexual behaviour in men aged 18–60 using a combination of personal interview and a prerecorded telephone interview using a mobile phone.²³

A combination of approaches could be used in England and Wales. A behavioural surveillance unit (BSU) within the HIV and STI Division has now been established at the Communicable Disease Surveillance Centre (CDSC). In association with key external partners the unit aims to collate data derived from ongoing local and national sexual behavioural surveillance and research programmes within CDSC and outside.

The BSU will streamline current behavioural data collection through existing surveillance systems. Collaborative partnerships with academic and research institutions involved in behavioural research will be established to define and collate key behavioural indicators relevant to HIV and other STI transmission. These indicators will include sexual behaviours such as number of sexual partners, types of sexual intercourse (vaginal, anal, and oral), and potentially preventative behaviours such as condom use and health service use for HIV and other STI screening. This would give an overview of behaviours at the population level in both the general population and in those with disease. A surveillance system, which will allow the prospective monitoring of the important risk indicators, could then be established.

A set of core questions will be established, which will draw on existing validated questions used in a variety of studies. This will enable improved comparability of data from diverse sources, at both national and local level. It will provide a comprehensive picture of sexual health, which can be monitored over time.

As a secondary, longer term objective, the BSU will work towards developing new behavioural surveillance systems for monitoring groups where there are currently inadequate data. Specially designed studies will be developed to complete the knowledge gaps—for example, in primary care and in ethnic minorities, where data cannot be obtained through enhancing existing systems. Again this is likely to be best achieved in partnership with external collaborators.