This chapter examines two areas critical to evaluating behavior change interventions: identifying indicators concerning HIV-related risk behavior and developing questionnaires to measure those indicators.

Since the earliest recognition of the HIV/AIDS epidemic, attempts have been made to identify and define indicators of HIV-related risk behavior that can be used to measure individual and community levels of risk, as well as to track changes in vulnerability over time. At the onset of the epidemic, a major concern was simply assessing general knowledge and awareness of the epidemic, as well as measuring general patterns of sexual behavior. Over time, and with the development of comprehensive HIV prevention and AIDS care programs, the need has been increasingly felt for more detailed indicators that can assess the subtle aspects of behavioral responses to the epidemic as well as help identify obstacles to behavior change.

As the understanding of the HIV epidemic has deepened and the sophistication and orientation of HIV-related programming has developed, it has become increasingly clear that the epidemic cannot be considered from a static perspective in which all epidemics in all settings are considered to involve the same elements and dynamics. Rather, a more targeted approach has been developed, which considers the epidemic in three stages:

• Low-level–HIV prevalence has never risen above 5 percent in groups of people with known high-risk behaviors, including sex workers, injecting drug users, and men who have sex with men.

• Concentrated–HIV prevalence has risen above 5 percent in at least one of the groups known to have high risk behaviors, but remains below 1 percent among sexually active adults in the "general population" (represented by antenatal clinic clients).

• Generalized–HIV prevalence among sexually active adults in the general population has surpassed 1 percent.

In considering these stages, considerable effort has gone into better defining the critical groups whose behaviors should be tracked over time as well as the indicators used to measure these behaviors. For example, in a concentrated epidemic, priority would be assigned to tracking the behaviors of high-risk sub-populations (such as sex workers and migrant men), whereas in a generalized epidemic, more emphasis would be placed on identifying and tracking risk behaviors in the general population, along with those of high-risk sub-populations.

Consequently, the information provided in this chapter is presented in the context of these developments. In addition, the many lessons learned about how best to ask questions and which indicators provide the most relevant information for measuring change are also presented. Accordingly, this chapter conveys these lessons and proposes standardized indicators and questionnaires appropriate to the stages of the epidemic as well as to the potential sub-populations most important in the epidemic's dynamic. These recommended indicators and questionnaires have been refined over time in collaboration with a variety of national and international partners, in a global initiative to determine a minimum set of key indicators for use in evaluating national HIV/AIDS programs. This key group of indicators is presented here, as well as other indicators deemed important for collecting more detailed data concerning HIV/AIDS-related behaviors.

Developing And Identifying Indicators For HIV/AIDS Behavioral Surveys

Because of the difficulty in measuring HIV incidence directly, as well as the difficulty in using HIV prevalence as a proxy for incidence, behavioral indicators have been heavily relied upon to help predict and monitor the course of the HIV epidemic. Monitoring risk behaviors (such as the frequency of unprotected sex with non-regular partners) over the short and medium term provides a means of assessing changes in behavior that might influence the course of the epidemic. The adoption and use of a limited set of behavioral indicators that are sensitive to the dynamic aspects of the HIV epidemic over time is therefore critical to the work of program planners, managers, and evaluators faced with few other interpretable tools.

For an indicator to be useful in a national, regional or local program, it must meet the following criteria:

• An indicator must measure a concept that is relevant to program effort.
• A program must be undertaken that proposes to effect change in the concept being measured (there is no point in measuring something that is not expected to change over time).
• The indicator must be able to measure trends over time.
• It must be feasible to collect data for reporting the indicator.
• The indicator must measure only one concept at a time and be easy to interpret.

Key behavioral indicators recommended for use in HIV and AIDS programs are presented in Table 10-1. These indicators are mostly related to sexual behavior, focusing on number and types of partners and condom use. But in many parts of the world, in addition to sexual transmission, transmission through sharing needles is also responsible for a considerable proportion of new HIV infections. Therefore, indicators relating to sharing drug injecting equipment are included as well. While it may be desirable to monitor indicators concerning knowledge and attitudes (especially in the areas of stigma and discrimination), including questions related to these areas will be a matter of choice depending upon the focus of a particular program. Examples of indicators of this type are included in Table 10-1 as well.

Indicators presented in this chapter are divided according to sub-population group of interest, including adults ages 15-49, unmarried youth, female sex workers, men who have sex with men, and injecting drug users. All indicators should be measured and reported by gender, as it is widely recognized that risk behavior characteristics tend to vary greatly between men and women.

Key Indicators In Assessing Behavioral Risk

Indicators most relevant to programming for HIV/AIDS and sexually transmitted infections (STIs) are those related to sexual behavior, partner networking, and drug injecting. Information on partner types and multi-partner behavior (including multiple regular partners, non-regular partners, and commercial partners), the frequency of sex, partners' sexual behaviors, and the timing of multiple partnerships (concurrent or serial) are important factors to consider when attempting to understand the level of risk and vulnerability of different groups. Unfortunately, however, the choice of key indicators measuring sexual behavior can be very complex. The ideal scenario, in which information on every sexual partner and every sexual act over an extended period of time is available, is unrealistic both in terms of the capacity of human memory and the logistical limitations of quantitative surveys.

Thus, the indicators related to sexual behavior and networking presented here represent a compromise between the ideal and the practical. Indicators related to the percentage of the sub-population who are sexually active, who have more than one partner, who have one or more non-regular partners, and who have engaged in commercial sex, provide information important to understanding sexual norms and practices within a community. The recall period for these indicators reflects practical considerations as well, defining a period long enough to allow for potential change and short enough to allow for reasonably accurate recall. This recall period varies, depending on the sub-population.

Condom Use
Following basic information on sexual partnering and networking, information on condom use is critical to appropriate STI/HIV/AIDS programming. Aside from abstinence, consistent condom use represents the only certain means of preventing the sexual transmission of HIV. Thus, tracking indicators of condom availability and use represents an important means of assessing levels of risk among individuals and communities. (See Chapter 5, "Evaluating Condom Programming," for more on this issue.) Questions concerning condom use can be posed in terms of ever use, frequency of use over varied recall periods (week, month, 6 months, year) or frequency of use during most recent sex act with particular partner types.

Each of these methods of examining condom use has strengths and weaknesses. First, while the question of ever use of condoms is very specific, it may not relate to recent changes in HIV-related behavior, and a one-time ever user of condoms is not guaranteed any assurance of reduced risk of HIV infection. In addition, the use of this indicator is becoming increasingly arcane, as availability of and familiarity with condoms increases globally. Nonetheless, it can be used as a gross measure of population behavior related to the introduction of condoms in a community where they were previously unavailable. In addition, increases in ever use of condoms without increases in their regular use may point to specific barriers to or dissatisfaction with their use.

Indicators measuring frequency of condom use over a given time period with different types of partners provide additional information regarding risk behavior. For example, identifying the percent of populations inconsistently or never using condoms can highlight vulnerable populations that may require targeted interventions. Distinct problems are associated with the use of this type of indicator, however. First, categorizing condom use according to frequency is imprecise, and absolute categories such as every time (100 percent of the time) and never (0 percent of the time) may have fewer responses than such categories as almost all the time or sometimes. In an effort to provide adequate (sensitive) response categories for the reporting of condom use, four response categories are therefore suggested: every time, almost every time, sometimes, and never. The use of four categories helps to achieve an improved distribution of responses across categories.

When interpreting condom use data, it is important to also consider partner type and frequency of sex, in addition to frequency of condom use. For example, men with many non-regular partners who report inconsistent condom use are likely to be at greater risk of infection and transmission than men with only one non-regular partner who report inconsistent condom use. At the same time, however, evidence suggests that individuals either adopt consistent condom use or do not, and therefore, knowing the proportion of individuals who have adopted consistent condom use is an important measure for assessing the impact of prevention programs on communities and predicting future trends in the epidemic, regardless of partner change and sex act frequency.

Measuring the frequency of condom use during the most recent sex act is often used as an internal validity check to the reported frequency of condom use over a long time period. While consistent condom use may be the ultimate goal, measuring condom use for a randomly selected sexual act (most recent) may provide an indication of the general frequency of condom use across all sex acts, and be a more sensitive measure of intermediate change.

When measuring condom use, however, it is essential to measure condom use related to specific categories of partners. Research has shown that condom use varies with the perception of partner, and asking about frequency of condom use over a given time period without specifying the type of partner will increase reporting of inconsistent use. For example, a person who consistently uses condoms with commercial partners but never uses them with regular partners will report sometimes use, although the risk of unprotected sex with these different types of partners varies greatly. Also, if some individuals are responding to last time use with a regular partner while others are responding to last time use with a non-regular partner, the understanding of the level of risk involved is obscured.

Onset of Sexual Intercourse
Another key behavioral indicator, median age at first sexual intercourse, is often chosen to reflect changes in broad social norms. When considering use of this indicator, it is important to specify the appropriate denominator. For the indicator to accurately reflect changes in the age at first sex among a population, all members of the population must be considered. Thus, the denominator must include those who have not yet initiated sexual intercourse as well as those who have. If only sexually active persons are included in the denominator, and a significant percentage of a specific age group (for example, girls aged 15-19) have not yet initiated sexual activity, the indicator will report a skewed figure of the median age at first intercourse among those sexually active. This figure will necessarily be a lower figure than the eventual median for the age group once all members have initiated sexual intercourse. This problem can be avoided by including all members of the population in question in the denominator.

Injecting Equipment Sharing Behaviors
In epidemics where there is a concentration of HIV infection among injecting drug users, it is important to include indicators that can measure awareness of the risk of unsafe injecting, as well as the frequency of unsafe injecting behaviors. Reduced sharing of injecting equipment and access to sterile injecting equipment are important IDU indicators. Sharing injecting equipment is both the biggest factor for HIV transmission among drug injectors and the most common focus of interventions, especially in non-industrialized countries where there is not a long history of prevention interventions among drug injectors. Measuring levels of sharing will not only serve to alert program planners about the need for interventions, but can also be used to advocate with policymakers about the need for harm reduction programs. Changes in sharing behavior will be especially valuable for tracking trends over time for programs that support needle exchange initiatives, or that work to improve easy access to safe injecting equipment.

When measuring sharing indicators, it is very important to operationalize what is meant by "sharing," because sharing can either be active (loaning injecting equipment) or receptive (borrowing injecting equipment). Being injected by a professional injector or a dealer should also be considered as sharing.

Just as with condom use, sharing can be measured either at last injection, or in terms of frequency over time. Because injectors who are addicted tend to inject daily, the time frame for sharing indicators needs to be relatively short–no more than 1 month. Beyond that time period, recall bias becomes an obstacle to valid data. Equally important for IDU populations are those indicators that measure sexual risk-taking behaviors, because sexual partners of IDU are at increased risk of being infected. Also, drug-taking itself may lead to increased likelihood of high-risk sexual behavior. Reducing sexual transmission should be a primary objective of IDU interventions.

Knowledge
Knowledge indicators have frequently been included in HIV/AIDS-related surveys, but while these indicators are generally useful in measuring overall awareness of HIV/AIDS information in a community, they often provide only a weak indication of risk levels within a community. An established gap exists between correct knowledge of prevention methods and the use of these prevention methods. For example, even though people know that condoms can prevent the transmission of HIV, it does not mean they correctly or consistently use condoms.

In addition, increasing levels of knowledge over time often means that knowledge indicators are not useful in interpreting risk-related change. For example, early on in the epidemic, the proportion of respondents reporting having ever heard of AIDS was often measured. Now, however, awareness is increasingly widespread and subsequently little change in this indicator can be expected over time in most populations. As a result, knowledge of the existence of AIDS is less and less frequently measured, and has been replaced in many settings by the measurement of accurate knowledge of means of transmission and prevention and attitudes concerning stigma and discrimination. These indicators have been proven to better indicate levels of risk and acceptance of people living with AIDS within a community and, thus, more useful to program evaluation and planning in the current global context of HIV/AIDS.

Another knowledge area that is often assessed is knowledge of STI symptoms because of the established link between STIs and the increased potential of HIV infection and transmission. The higher the prevalence of unrecognized and/or untreated STIs, the higher the risk of increased transmission and infection due to the presence of STIs. For a population to seek appropriate and timely care for STIs, they must first be able to recognize that they are infected. The behavioral indicator that correlates with this knowledge area is seeking appropriate and timely treatment for STIs. While measuring this indicator is desirable, it is also problematic because the number of respondents reporting having had STI symptoms and having sought treatment is often insufficient to accurately report findings or follow trends over time.

Risk Perception
While program planners frequently wish to assess risk perception, undertaking this estimate accurately is extremely difficult. First, risk perception must be examined in the context of behavior in order to be meaningful–without an accurate idea of past and current behavior, self-perceived risk is difficult to interpret, if not meaningless. But even when assessed in the context of self-reported risk behaviors, "correct" risk perception is difficult to ascertain. For example, a survey among sex workers in Jamaica found that an equal proportion felt they were at risk and not at risk. Subsequent questions determined that of the women who reported nearly 100 percent rates of condom use, half reported that they were at high risk and therefore used condoms, while the other half reported that they were not at risk because they used condoms. In addition, risk perception depends a lot on the stage of the epidemic. In low-prevalence settings, even people with high-risk behavior may justifiably perceive that their risk is low, whereas in high-prevalence settings, people may already know they are infected (making risk perception for becoming infected irrelevant). In high-prevalence settings, despite having adopted safe sex practices at present, people may perceive that they are at high risk of being infected because of their past behaviors. Finally, it is unclear whether the desirable outcome for this indicator is that it should be increasing or decreasing. It may increase initially, as people become more aware of the risk of HIV infection, but then may decrease again as people adopt safe-sex behaviors. Because of all these ambiguities, assessing risk perception is a complex issue that is more appropriately explored through qualitative research, including focus group discussions and in-depth interviews. (See Chapter 12, "The Role of Qualitative Data in Evaluating HIV Programs," for more on these methodologies.)

Additional Indicators

A range of contextual factors, such as discussions between regular partners about HIV/STIs, knowledge of someone infected with HIV or who has died of AIDS, regular use of alcohol or drugs, or having been tested for HIV infection, are understood to have a correlation to risk behavior change. These indicators may be included as additional indicators to allow for a multifaceted examination of determinants of and obstacles to behavior change. These factors, which relate to the context within which risk behavior decisions are made, contribute to the improved understanding of the sometimes unclear relationships between knowledge levels and change in risk behavior. Other indicators that can contribute to the understanding of HIV spread are frequent travel away from home for extended periods of time and age difference between sexual partners. These are not necessarily indicators that one expects to see a change in over time. They merely play a diagnostic role in helping explain the dynamics of the epidemic. Yet another category of indicators are those known as "overlap" indicators. An example of such an indicator would be injecting drug use among sex workers. This type of indicator is meant to raise awareness about the synergistic effects of multiple risk behaviors, and the potential for increased spread of the virus between the IDU network and commercial sex networks. In a similar fashion, it would also be important to measure the extent of sexual risk and selling of sex by a population of injectors.

The indicators recommended here, both the primary and additional ones, are limited in their ability to act as stand-alone measures of the course of the epidemic. However, in conjunction with one another, they provide a comprehensive framework that is useful for analyzing risk within a community and comparing risk levels between communities. In the past, quantitative surveys have often been limited in generalizability and replicability because of non-standardized instruments and indicators that restrict comparisons with other data sets. In an effort to diminish the problems associated with variations among and across questionnaires, it is strongly recommended that indicators be selected and constructed in a standardized manner so that an understanding of broad trends in the epidemic can be defined and compared to other populations within the same country, across different countries, and across time.

Determining The Time Frame Of Indicators

One of the most potentially confusing aspects of indicator development is the time frame of behaviors. For example, when asking a male respondent about whether he has had sex with a sex worker, should the time frame for this behavior be the past month, the past 6 months, or the past year?

People tend to remember recent behaviors more accurately and this seems to argue for the use of shorter time frames. However, if the behavior is not extremely frequent or common, too short a time frame will yield few respondents and make it difficult to track trends in the behavior over time with any degree of statistical confidence. Likewise, populations with very frequent risk behaviors will not remember the details of their behaviors over a long time period, such as 12 months.

The standardized set of indicators have taken these factors into account, and that is why some are over a 12-month period, while others only consider the past 6 months or the past 1 month. Consistency is critical, because changing the time frame for a behavioral indicator across survey rounds, or deviating from the standardized time frame for a given sub-population will yield dramatically differing results. While different time frames may be appropriate because of differences in local context, it should be kept in mind that comparisons of findings to other sub-populations and settings will be difficult if not impossible. Thus, time frames must be adopted that are both convenient to the respondent and analyzable to the researcher.

Developing Questionnaires For HIV/AIDS Behavioral Surveys

Standardized questionnaires should be used for behavioral surveillance to maximize the comparability of data between survey rounds and across sub-populations and geographic regions. Small variations in wording and the order of questions can greatly affect responses to questions so that observed changes in behaviors over time may, in fact, be due to these changes as opposed to any real changes in behavior. Preparing a well-developed questionnaire that can be maintained over multiple survey rounds with minimal changes is critical, therefore.

International experience in surveying key sub-population groups has generated a wealth of knowledge on what types of questions work and do not work when asking people about their sexual and drug-using behaviors. These questions have been brought together to form questionnaires that, in turn, are used to measure the indicators discussed in the preceding section. These standardized questionnaires have been extensively tested in international settings and are available at Family Health International's website (www.fhi.org) as well as in published behavioral surveillance system (BSS) guidelines. The website also provides information about how and when to use these questionnaires.

Five key sub-population groups are covered by these questionnaires:

• adults 15-49 years old;
• unmarried youth;
• female sex workers;
• men who have sex with men; and
• injecting drug users.

Using standardized questionnaires has many advantages. First, questionnaire development is a difficult process, and already developed instruments such as those cited above contain formulations of questions, time references, and skip patterns that have been tested and are known to produce high-quality data. Second, because these instruments have been used in numerous settings throughout the world, their continued use will allow behavioral surveillance results to be compared internationally to determine differences in the dynamics of behavior change and the characteristics of different population groups.

It is still essential, however, to pretest and adapt survey instruments for every local setting. This involves translating the instruments into local languages and using appropriate local terminology to ensure that the original meaning of the question is not lost. It is also necessary to conduct qualitative research and involve local members of the sub-population groups who can help interpret and adapt the questions and response categories. Back-translating the questionnaire to ensure that the translation into the local language maintains the original version's key concepts and meanings is an important additional step in ensuring the quality of the instrument.

It is also useful to develop a guide for interviewers and supervisors, which goes through the questionnaire one question at a time, explaining in full the rationale behind a question and its intended meaning. This guide can be used in training and in the field, to clarify any ambiguities or misunderstandings that may arise. An example of a supervisor/interviewer guide can be found at www.fhi.org as well as in BSS guidelines.

Informed Consent and Ethical Considerations
Confidentiality and informed consent are important for all research subjects, but when the research involves an illegal or stigmatized activity such as sex work, injecting drug use, or illegal migration, the importance of the privacy of the respondent is magnified.

Thus, behavioral surveys cannot take place without the informed consent of the respondent. Special efforts must be made to ensure that potential respondents understand any risks involved in taking part in the study, and every effort must be made to ensure that the community will derive some benefit from participating in the study. Involving the community in the planning and implementation of the study is one way to achieve this, as is the training of interviewers to ensure that respondents are informed of the purpose of the study and that their participation is requested in a factual and neutral manner. Measures taken to ensure the confidentiality of the respondent should be explained, and consent should be clearly given by the respondent before interviewing begins. In addition, the interviewer should sign the questionnaire at the time of consent in order to indicate that consent has been given. No respondent names or other identifying information should be recorded. It should be explained to respondents that they have the right to refuse to take part in the study as well as to drop out of the study at any time, and should by thanked politely for their time, whether or not they choose to participate.

Questionnaire Administration and Interview Settings
In situations in which some members of the sub-population group of interest may be illiterate, data should be collected by a trained interviewer who explains questions to the respondent and records answers. It is important to use the same data collection approach with all respondents because varying the way in which data are collected could bias results (for example, if some respondents are literate and others illiterate, an interview should be conducted with all respondents, rather than self-administered questionnaires being given to literate respondents and interviews conducted with the others). Where respondents are literate and educated (such as in student populations), respondents may record answers to questions themselves on an anonymous written questionnaire, submitting it to a data collection manager in a sealed envelope so that it cannot be distinguished from that of other respondents. It is important to ensure that the style and wording of this type of questionnaire is appropriate to self-administration, and that these questionnaires are pre-tested to ensure clarity and answerability.

When using trained interviewers to conduct a survey, it is important for them to conduct survey interviews in a setting where questions and answers cannot be overheard by others as well as to engage in a rapport-building conversation before asking survey questions. This will reduce the likelihood that respondents will give "socially desirable" answers rather than telling the truth. If a third person enters the room or is within hearing distance, the interviewer should explain that it is important to interview the respondent in privacy.

Assuring Quality Control Before and During Fieldwork
If care is taken to exercise quality control during fieldwork, two main sources of error that interfere with the ability to collect valid data will be avoided. One source comes from the people collecting data, and the other from the people from whom the data are being collected.

It is commonly said that people do not tell the truth about their sexual behaviors, and that they exaggerate, withhold information, or refuse to admit to behaviors that are culturally unacceptable. Experience has shown that certain techniques can increase the likelihood of honest sharing of information. When interviewers are well trained to discuss sensitive behaviors with respondents and make them feel at ease, research suggests that respondents will provide truthful information. Comprehensive interviewer guidelines (such as those cited above) can contribute to the comfort levels of both interviewers and respondents when discussing sensitive topics such as sex and drug taking.

Other quality control issues relate to interactions with the groups being surveyed. Often these surveys are conducted with communities who are vulnerable and therefore reticent to open up to strangers. Working through community organizations that have relationships with the sub-populations of interest in an essential component of ensuring access to community members. In some cases, such as with men who have sex with men and injecting drug users, it is necessary to use members of the community itself (or those working with them) to do the interviewing because it cannot be expected that sufficient rapport can be built between interviewers and respondents in a short period of time. If interviewers do not come from the community, then they must be carefully chosen individuals who will not threaten respondents in any way. Concern for the privacy and confidentiality of respondents and the community must be maintained at all times and winning the trust of the community is essential to obtaining valid results.

Conclusion

In summary, key considerations when identifying indicators and developing questionnaires include the following:

• Indicators most relevant to STI/HIV/AIDS programming are those related to sexual behavior, partner networking, and drug injecting. However, it has become increasingly clear that HIV cannot be considered from a static perspective in which all epidemics in all settings are considered to involve the same elements and dynamics. Therefore it is crucial to consider the stage and location of the epidemic when deciding which indicators to measure.

• When selecting indicators, consider how much a given indicator can be expected to change over time. Some knowledge indicators (or even behavioral indicators) may peak after several years and no longer indicate changes. How the information gathered through this indicator will be used in the program setting should also be considered.

• Use standardized questionnaires to increase generalizability and replicability across populations and geographic regions and pretest all questionnaires adequately and appropriately.

• Assurances of confidentiality to the respondent and community should be provided.

• All indicators should be measured and reported by gender, as it is widely recognized that risk behavior characteristics tend to vary greatly between men and women.

• Tracking indicators of condom availability and use represents an important means of assessing levels of risk among individuals and communities. When measuring condom use, it is essential that it be related to specific categories of partners, because research has shown that condom use varies with the perception of partner.

• In epidemics where there is a concentration of HIV infection among injecting drug users, it is important to include indicators that can measure awareness of the risk of unsafe injecting, as well as the frequency of unsafe injecting behaviors

• Equally important for IDU populations are those indicators that measure sexual risk-taking behaviors, because drug-taking may be associated with increased sexual risk, and because sexual partners of IDU are at increased risk of being infected.

• Regarding timeframes for indicators, people tend to remember recent behaviors more accurately and this seems to argue for the use of shorter time frames. However, if the behavior is not extremely frequent or common, too short a time frame will yield few respondents and make it difficult to track trends in the behavior over time with any degree of statistical confidence. Likewise, populations with very frequent risk behaviors will not remember the details of their behaviors over a long time period, such as 12 months. Thus time frames must be adopted that are both convenient to the respondent and analyzable to the researcher.

Recommended Reading

1. Brown T, Sittitrai W, Carl G, et al. Geographic and social mapping of commercial sex: a manual of procedures. Honolulu and Bangkok: Program on Population, East-West Center, and Thai Red Cross AIDS Research Center; 2000.
2. Catania J, Gibson D, Marin B, et al. Response bias in assessing sexual behaviours relevant to HIV transmission. Eval Progr Plann 1990;13:19-29.
3. Catania JA, Gibson DR, Chitwood DD, Coates TJ. Methodological problems in AIDS behavioral research: influences on measurement error and participation bias in studies of sexual behavior. Psych Bull 1990(3);8:339-362.
4. Catania JA, Moskowitz JT, Ruiz M, et al. A review of national AIDS-related behavioral surveys. AIDS 1996;10(Suppl A): S183-S190.
5. Cleland J, Ferry B, editors. Sexual behavior and AIDS in the developing world. London: Taylor & Francis; 1995.
6. Coates RA, Soskolne CL, Calzavara L, et al. The reliability of sexual histories in AIDS-related research: evaluation of an interview-administered questionnaire. Can J Public Health 1986;77(5):343-348.
7. Dare OO, Cleland JG. Reliability and validity of survey data on sexual behavior. In: AIDS impact and prevention in the developing world: demographic and social science perspectives. Health Transit Rev 1994; 4(Suppl):93-110.
8. Family Health International (FHI). Guidelines for repeated behavioral surveillance surveys in populations at risk of HIV. Arlington (VA): Family Health International; 2000.
9. Family Health International/IMPACT and Joint United Nations Programme on HIV/AIDS (UNAIDS). Meeting the behavioural data collection needs of national HIV/AIDS/STD programmes. Proceedings from a joint IMPACT/FHI/UNAIDS workshop. Arlington (VA) and Geneva: Family Health International/IMPACT and UNAIDS; 1998.
10. Figueroa JP, Brathwaite AR, Wedderburn M, et al. Is HIV/STD control in Jamaica making a difference? AIDS 1998;12(Suppl 2):S89-S98.
11. Fishbein M, Pequegnat W. Evaluating AIDS prevention interventions using behavioral and biological outcome measures. Sex Transm Dis 2000;27(2):101-110.
12. Hogle J, Sweat M. Qualitative methods for evaluation research in HIV/AIDS prevention programming. AIDSCAP evaluation tools, Module 5. Arlington, VA: Family Health International; 1996.
13. Konings E, Bantebya G, Caraël M, et al. Validating population surveys for the measurement of HIV/STD prevention indicators. AIDS 1995;9(4):375-382.
14. Lemp GF, Hirozawa A, Givertz D, et al. Seroprevalence of HIV and risk behaviors among young homosexual and bisexual men: the San Francisco/Berkeley Young Men's Survey. JAMA 1994 Aug;272(6):449-454.
15. Mills S, Saidel T, Bennett A, et al. HIV risk behavioral surveillance: a methodology for monitoring behavioral trends. AIDS 1998;12(Suppl 2):S37-S46.
16. Pequegnat W, Fishbein M, Celentano D, et al. NIMH/APPC workgroup on behavioral and biological outcomes in HIV/STD prevention studies: a position statement. Sex Transm Dis 2000;27(3):127-132.
17. UNAIDS/00.17E. National AIDS programmes: a guide to monitoring and evaluation. Geneva: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2000.
18. Watters J, Biernacki P. Targeted sampling: options for the study of hidden populations. Soc Prob 1989;36(4):416-430.
19. Zenilman JM, Weisman CS, Rompalo AM, et al. Condom use to prevent incident STDs: the validity of self-reported condom use. Sex Transm Dis 1995;22(1):15-21.