|Home | About | Journals | Submit | Contact Us | Français|
HIV behavioral surveillance in the United States is conducted among three groups: infected populations, high-risk populations, and the general population. We describe the general population component of the overall U.S. HIV behavioral surveillance program and identify priority analyses. This component comprises several data systems (ongoing, systematic, population-based surveys) through which data on risk behaviors and HIV testing are collected, analyzed, and disseminated. Multiple data systems are needed to balance differences in scope and purpose, as well as strengths and weaknesses of the sampling frames, mode of administration, and frequency of data collection. In a concentrated epidemic, such as in the United States, general population data play a small but important role in monitoring the potential spread of infection more broadly, particularly given increases in HIV transmission through heterosexual contact.
As part of its strategic plan to control the spread of HIV, the Centers for Disease Control and Prevention (CDC) has established the overarching goal of reducing new HIV infections in the United States by 50%. To achieve this goal, four specific goals were identified, one of which is to strengthen the national capacity to monitor the HIV epidemic to better direct and evaluate prevention efforts.1 As part of these monitoring activities, CDC has defined a program of HIV behavioral surveillance that focuses on collecting behavioral data from three population groups: those who are infected, those at increased risk for infection, and the general population.2
The HIV epidemic in the United States is characterized as “concentrated,” meaning the highest prevalence of infection is in specific groups and the epidemic is not well established in the general population. Specifically, concentrated epidemics are defined as having HIV prevalence higher than 5% in at least one subgroup and below 1% in the general population. Given this concentrated nature of the epidemic in the U.S., the major focus of behavioral surveillance efforts is on infected and high-risk populations. Despite the concentration of infection and, therefore, surveillance efforts in specific groups such as men who have sex with men (MSM) and injecting drug users (IDUs), general population behavioral surveillance is critical, particularly given the increase in HIV attributed to heterosexual contact.3
The World Health Organization's “second-generation HIV surveillance” recommends the use of national, population-based surveys such as demographic and health surveys in combination with surveys of high-risk populations.4 The conceptual framework for HIV behavioral surveillance in the U.S. similarly recommends that data are needed from the general population, as well as infected and high-risk populations.
Specifically, according to the HIV behavioral surveillance framework for the U.S., information on risk and testing behaviors should be collected among the general population.2 Risk behavior data are used to assess the prevalence of risk behaviors at a point in time, monitor trends in these behaviors over time, assess the co-occurrence of risk behaviors, and identify demographic and social correlates of risk.5,6 Monitoring testing behaviors in the general population is appropriate because HIV testing programs are not always targeted to specific groups. Data from general population surveys can be used for assessing progress toward health objectives related to HIV testing, such as CDC's HIV Prevention Strategic Plan1 or Healthy People 2010;7 evaluating national policies or guidelines on HIV testing;8,9 and evaluating the use of new testing technologies, such as oral fluid tests and rapid tests. Data from both risk and testing behaviors in the general population can play a part in evaluating the reach of HIV prevention programs, policies, and guidelines.
In this article, we describe the general population component of the overall U.S. HIV behavioral surveillance program and identify priority analyses. The contribution of general population data is discussed in the context of the strengths and limitations of these population-based surveys.
The general population component of the HIV behavioral surveillance program includes data from several systems. The Division of HIV/AIDS Prevention (DHAP) at CDC works in partnership with various organizations that conduct ongoing, systematic behavioral surveys to collect, analyze, and disseminate general population HIV behavioral surveillance data. There are several reasons for relying on partnerships to collect general population HIV-related data. One is cost-effectiveness. In addition, the different purposes of each survey (none of which is specific to HIV) means that no single existing data system is able to (1) reach an appropriate sample at the national, state, or local level, (2) employ the best mode of survey administration given the nature of the questions asked, (3) collect data with sufficient frequency, or (4) meet the HIV behavioral surveillance data needs described in this article.
Population-based surveys are expensive, time-consuming, and complex. Therefore, it is not cost-effective to run a general population survey that only meets the needs of HIV behavioral surveillance. The complex sampling designs require skilled staff for planning and implementing the survey as well as for analyzing the data.
The sampling frames for population-based health surveys in the United States differ. Some are nationally representative while others are state-based, and some restrict their samples or HIV-related question sets to individuals of a certain gender or age range.
Different modes of data collection are used, including face-to-face, telephone, self-administered, or some combination of these methods. Mode of administration can affect the responses participants give, particularly to sensitive questions such as sexual behavior and drug use. Each mode of data collection has its advantages and disadvantages; having multiple data sources to compare on similar measures allows for a better estimation of bias from any particular mode of administration or sampling frame.
In terms of frequency of data collection, to be used as part of a surveillance approach, the surveys must be ongoing and systematic. Some systems have periodic data collection and others are continuous.
Because the HIV epidemic in the United States is concentrated (i.e., not widespread in the general population), a limited amount of information is needed from the general population for the purposes of HIV behavioral surveillance. Measurement of behaviors and data analysis for the purpose of HIV behavioral surveillance should be focused on the measures listed in the Figure. These indicators were chosen based on the model for HIV behavioral surveillance2 and reflect common routes of HIV transmission (sexual behavior, injection drug use) and common, widespread prevention strategies in the U.S.
Given the limited scope of data needed for the purposes of general population HIV behavioral surveillance, developing and maintaining partnerships between DHAP at CDC and organizations that conduct population-based surveys is a more efficient strategy than fielding a new surveillance system. The addition of a few key HIV-related questions to various population-based surveys—such as the Behavioral Risk Factor Surveillance System (BRFSS), the National Survey of Family Growth (NSFG), and the General Social Survey (GSS)—provides a sufficient picture of the behaviors driving the epidemic in the general population.
To illustrate the manner in which general population HIV behavioral surveillance is implemented via partnerships with other behavioral data systems, there are three surveys that are part of the HIV behavioral surveillance program: NSFG, GSS, and BRFSS. Although these three surveys were chosen to demonstrate the diversity in the purpose of the surveys and the types of partnerships with DHAP, they are not the only systems that collect HIV-related data from ongoing or repeated population-based surveys.
NSFG and BRFSS are CDC-run surveys, administered by the National Center for Health Statistics (NCHS) and the National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP); information about these surveys can be found at www.cdc.gov/nchs/nsfg.htm and www.cdc.gov/brfss. GSS is administered by the National Opinion Research Center (NORC) at the University of Chicago (http://www.norc.uchicago.edu/projects/gensoc1.asp). The Table summarizes the key features of each survey. For all three surveys, as part of the partnership, staff at DHAP provide subject-matter expertise in the development of HIV-related survey questions; data analysis is a collaborative effort among staff from each system and DHAP.
NSFG is a periodic survey of a nationally representative sample of people aged 15–44 and examines factors that affect the formation, growth, and dissolution of families, such as marriage, divorce, births, sexual activity, and contraceptive use.10 NSFG supplements the birth registration system by providing data on behaviors that explain birth and pregnancy rates, as well as serving the needs of other federal programs by including additional demographic and epidemiologic topics. To date, NSFG has been conducted six times: in 1973, 1976, 1982, 1988, 1995, and 2002. Its scope and methods have changed somewhat over time; for example, men were included for the first time in 2002.
In 2002, about 12,600 men and women completed NSFG, and a nationally representative multistage area probability sampling method was used. Trained female interviewers visited the homes of selected people in 120 areas across the country and surveyed one person from each selected household. Interviewers used computer-assisted personal interviewing (CAPI), although parts of the survey—including questions related to HIV risk behaviors—were administered by audio computer-assisted self-interview (ACASI). Data were weighted to produce national estimates.
DHAP contributes funding for questionnaire item development and data collection for NSFG and provides input on decision-making with regard to the survey's design and schedule. DHAP representatives are invited to periodic and ad hoc workshops on NSFG held by the NCHS.
In 1995 and 2002, NSFG questionnaires included a section on sexually transmitted disease (STD)/HIV risk behaviors. This section included questions for women on having male sex partners who had sex with other men and, for men, questions on having male sex partners. Both men and women were asked about number of sex partners, condom use, and injection drug use. The survey also included questions about HIV counseling and testing experiences. With its reproductive health focus, NSFG included a wide range of questions on sexual activity, such as age at first sex and nonvoluntary sex that can be analyzed in the context of HIV risk.
BRFSS is a system of state-based health surveys that focus on health risk behaviors, clinical preventive health practices, and health care access.11 BRFSS methods allow for annual estimates at both the state and national levels; the indicators of health included in BRFSS are used to measure progress toward disease prevention goals established by the federal government. Approximately 200,000 interviews with adults (aged 18 and older) are conducted each year in all 50 states, the District of Columbia, Puerto Rico, the Virgin Islands, and Guam.
The BRFSS sampling method is a disproportionate stratified sample of telephone numbers in each state. States stratify the sampling frame for high and low household density groups and a probability sample of all households with telephones is obtained in each strata. One person per household is selected to participate. CATI is used to administer the survey, which takes about 20 minutes. Data are weighted for probability of selection of telephone number, the number of adults in a household, and the number of telephones in a household. Post-stratification adjustments are created for nonresponse and households without telephones.
The BRFSS survey is comprised of three components: the core questionnaire, optional modules, and local questions. The core questionnaire is developed each year with input from the states and representatives from the federal government, including DHAP. While BRFSS is primarily focused on risks for chronic diseases, HIV-related questions have been included since the late 1980s. The DHAP representative is responsible for attending BRFSS national meetings and for proposing questions for emerging issues that could impact the HIV epidemic. Efforts are made to ensure the consistency of questions from year to year and to use questions that have been used on other national surveys. Each new question proposed to BRFSS is evaluated through cognitive and field testing before it is added to the questionnaire.
The HIV/AIDS section of the current BRFSS core questionnaire collects the following information: whether the respondent was ever tested for HIV and, if so, the month and year of last test and the facility where last tested; and a general risk variable in which risk behaviors are measured in a single question: respondents only answer “yes” or “no,” but do not indicate a specific risk. In 1997, funding was provided for a sexual behavior module that was used by 23 states, the District of Columbia, and Puerto Rico. In subsequent years, the number of states using this module declined, making multistate analyses difficult and less meaningful. Historically, the HIV/AIDS section of BRFSS has included questions about knowledge and attitudes about HIV and HIV prevention strategies.12
GSS collects information from the general public on a variety of subjects, including attitudes toward social issues, education, religion, government, and other institutions; jobs and the economy; and politics and policy issues. Since 1988, GSS has also collected data on sexual behaviors, including number of sex partners, frequency of intercourse, extramarital relationships, and sex with prostitutes.
GSS, conducted by NORC, was conducted annually for most years between 1972 and 1994, and biennially thereafter. GSS is a national probability sample of all noninstitutionalized English-speaking people 18 years of age or older living in the United States. The interview is conducted in person; starting in 2002, interviewers conducted the survey by using CAPI.
The questionnaire contains a standard core of demographic and attitudinal variables, plus certain topics of special interest selected for rotation (called “topical modules”). The exact wording of questions is retained over time to facilitate trend analyses as well as replications of earlier findings.
The interview takes approximately 90 minutes. In 1994, GSS implemented a split-sample design. The sample consists of two parallel subsamples of approximately 1,500 people each. The two subsamples are administered the identical core questions but differ in the topical modules included. Thus, sample sizes for questions in specific topic modules will be about half the size of the total sample. In 2002, the total sample size was 2,765.
Since 1988, HIV/AIDS-related items have been included on GSS. Currently, these include: number and gender of sex partners, sex with risky partners, condom use at last sex by type of partner (steady or not), injection drug use, and use of crack cocaine. CDC has supported the collection and analysis of these data through a series of contracts with NORC. CDC's current contract with NORC includes the addition of three questions to GSS that measure history of HIV testing for the 2006 GSS.
These general population data systems have individual strengths and weaknesses that must be considered for data to be included in a behavioral surveillance program. As previously noted, key features include the purpose of the data system, sampling frames, mode of administration, and frequency of data collection.
A major strength of BRFSS is that it produces state-based and national estimates. For more recent years, city-based estimates are also produced. The utility of state- and local-level data makes this system a recommended part of the HIV epidemiologic profiles that are used to obtain and allocate funds for HIV prevention.13
HIV-related questions on each of the three survey instruments vary in number and topic, based on the purpose of the systems. NSFG and GSS collect data for the indicators in the Table for sexual behavior, injection drug use, and HIV testing. The BRFSS questionnaire includes each of the HIV testing measures noted in the Table. Risk behaviors are measured using the single question previously described.
While BRFSS administers the survey by telephone, NSFG and GSS are household-based, face-to-face surveys. Recent response rates in these surveys reflect the current difficulties in conducting telephone surveys: response rates for NSFG (79%) and GSS (70% to 75%) are considerably higher than BRFSS (median response rate in 2000 was 48.9%). Several studies indicate more risk behaviors are reported using ACASI;14, 15 therefore, NSFG may have more reliable sexual behavior data than the other surveys.
GSS has produced a national sample every two years, and BRFSS produces state and national estimates annually; these data systems provide consistent and comparable sources of data to assess changes in HIV-related behaviors over time. NSFG traditionally has been fielded on a periodic basis; the utility of NSFG data for HIV behavioral surveillance has increased with the addition of men to the sample, and plans to begin continuous data collection in 2007 will increase it further.
Despite data from each of the three systems being available for a number of years, few trend analyses have been conducted on behaviors related to HIV infection. Thus, a general priority for HIV behavioral surveillance is to examine trends in data from each system, particularly to monitor changes in sexual behavior and drug use. More common than trend analyses are comparisons of similar data across surveys at a single point in time.16–19
Having multiple sex partners increases an individual's chances of coming into contact with an infected partner. In addition, the rate of sex partner change affects how efficiently the virus can be spread within a population.20 Therefore, assessment in the U.S. population of the prevalence of having multiple sex partners during a specified time period can be used as a broad-based measure of risk for HIV infection. For example, data from 23 states participating in BRFSS in 1997 indicated that 11% of respondents reported multiple sex partners in the past year.21 Data from other surveys in more recent years show a higher prevalence of multiple partners: 18% of men and 14% of women in the 2002 NSFG reported multiple partners in the past year;17 from the 2004 GSS, 15% overall had multiple partners in the past year (unpublished data).
To better assess the chances that a partner is infected, information about the partner's risk behaviors is important. Specifically, ascertaining whether the partner injects drugs or is a man who has sex with men can give a better assessment of risk to a particular respondent. GSS provides estimates of the percentage of MSM; from 1996 through 2000, this estimate was trending upward.22 Data from the 2002 NSFG (the first year in which males were surveyed) showed that 3% of males aged 15–44 have had oral or anal sex with another male in the last 12 months; about 1% of men aged 15–44 had both male and female sexual partners in the last 12 months.17
For those at risk, consistent and correct use of condoms can lower the probability of transmission of HIV and other STDs.23 Ongoing assessment of lack of condom use among those who are sexually active can be used to measure risk, while the prevalence of condom use serves as a measure of adoption of preventive actions. In the 1997 BRFSS, among those with multiple partners, 65% reported using a condom at last sex.21 In the 2002 NSFG, 40% of men and 22% of women used a condom at their most recent sex.17 In the 2000 GSS, condom use at last sex was reported by 19.7% of the sample.16 While BRFSS has not had states collect sexual behavior data consistently over time, NSFG and GSS have done so and are good sources for examining trends in these sexual behavior indicators.
Collaborations between DHAP and NSFG staff resulted in a major report on sexual behavior based on data collected in the 2002 NSFG.17 Reports from GSS on sexual behavior are prepared for CDC as part of the contractual agreement. To the extent that these reports provide data on the sexual behavior indicators shown in the Table, further analyses for the purposes of behavioral surveillance can focus on comparing prevalence and trends in these indicators over time and across surveys, as well as comparing to data collected from high-risk and infected populations.
HIV prevention messages for IDUs encourage them to stop using drugs; however, it is recognized that this is a difficult behavior to change and, therefore, drug treatment is recommended for those who cannot stop on their own. For those who cannot or will not stop injecting drugs, using sterile needles and syringes is recommended. Given that these are the priority strategies for preventing IDU-acquired HIV infection, it is necessary to monitor the prevalence of injection drug use, utilization of drug treatment by IDUs, and use of sterile needles and syringes.
IDUs may acquire HIV infection from or transmit it to their sex partners. Therefore, for the purposes of behavioral surveillance, it is also important to collect data about sexual risk behaviors from IDUs.
The National Survey of Drug Use and Health, (NSDUH), sponsored by the Substance Abuse and Mental Health Services Administration, provides yearly national and state estimates of drug use in the United States and is the most comprehensive general population survey on drug use among adults.24 This survey includes questions about injection drug use, needle sharing, and drug treatment; however, it does not include any sexual behavior data.
Based on 2002 and 2003 data from the NSDUH, an annual average of 354,000 people (0.2%) aged 12 or older used a needle to inject drugs during the past year.24,25 Adjusting for various factors that affect the validity of these data, Friedman, et al., estimated the number of injectors in the United States to be approximately 1.4 million.26 Preliminary data from the 2004 GSS indicate that 2.6% of respondents had ever injected drugs, and 5.9% of those had injected within the past year (unpublished data). With any of these estimates, injection drug use is not a common behavior in the general U.S. population. This low prevalence is reflected in the limited information needed to monitor injection behaviors in general population surveys (the Figure) and the limited data collected (the Table).
Priority analyses regarding injection drug use for the purposes of behavioral surveillance should include prevalence and trend data on injection behaviors, use of sterile needles, and access to/use of drug treatment. These indicators are all available from NSDUH. Data from NSFG and GSS provide the ability to monitor prevalence of and trends in injection drug use, with the added benefit of examining drug use in conjunction with sexual risk behavior data. However, such analyses may be limited by small cell sizes for analysis, as the prevalence of injection drug use is low.
In April 2003, CDC announced a refocused commitment to HIV prevention through funding of the Advancing HIV Prevention (AHP) initiative.9 The initiative emphasizes increased HIV testing in medical and nonmedical settings to identify infected people who are not aware of their own infection, and getting them into treatment and prevention services. HIV/AIDS surveillance provides testing and clinical data on HIV-infected individuals. Data are also available through the counseling and testing system, collected in some publicly funded clinics. Estimates of HIV testing in the general population can provide a broad picture of testing in a variety of settings, among those at both high and low risk.
The BRFSS data have been used to determine the proportion of the U.S. population that has been tested for HIV and where they are getting tested; generally, these data are reported each year.19 Over the past several years, BRFSS data have been used to describe prevalence and trends in HIV testing. Recently, HIV testing data from the 2002 NSFG were published.18 In 2001, BRFSS data showed that 23.4% of black Americans and 14.8% of Hispanics tested recently (in the past 12 months), compared to 10.8% of Caucasians.27 Among Native Americans, during the period from 1998 to 2000, about half had ever been tested.28 From the 2002 NSFG, overall 15% of men and women had been tested in the past 12 months; like BRFSS, recent testing was more common among black Americans (24.2%) and Hispanics (16.1%) than Caucasians (12.5%).18
Trend analysis of BRFSS data found that the proportion of pregnant women who reported being tested increased from 41% in 1995 to 53% in 1996, after the publication of U.S. Public Health Service guidelines that recommended all pregnant women be counseled and offered testing for HIV infection.29 Data from 2002 across several surveys indicate a range of prevalence of testing among pregnant women from 48% (NHIS)19 to 69% (NSFG).18 Differences in results for the same study year may be attributable to survey methods, question wording, or other factors; however, cross-survey comparisons are useful to assess the extent of variation.
Because BRFSS has included the same questions over a number of years, it is possible to identify trends in HIV testing in the general population and, thus, evaluate the reach of HIV testing strategies and policies. Analyses of prevalence and trends in HIV testing based on data from BRFSS and NSFG will be used to assess progress in specific AHP strategies and the overall renewed emphasis on HIV testing as a preventive measure.
In a concentrated epidemic, such as in the United States, general population data play a small but important role in monitoring the potential spread of infection more broadly, particularly given increases in HIV transmission through heterosexual contact.3 The general population component of the HIV behavioral surveillance system is based on partnerships with existing ongoing surveys, including BRFSS, NSFG, and GSS. Analysis of general population data for HIV behavioral surveillance must focus on the broad-based questions relevant to the general population and capitalize on the strengths of different data systems to answer these questions.