Current Trends Pilot Study of a Household Survey to Determine HIV Seroprevalence
A survey based on a probability sample of U.S. households was suggested as a method to determine the number of persons infected with human immunodeficiency virus (HIV) (1). To test the feasibility of such a survey, CDC's National Center for Health Statistics (NCHS) recently conducted a pilot study in two sites through a contract with the Research Triangle Institute. The first field test, conducted in January 1989, used a random sample of adults from households in Allegheny County (Pittsburgh), Pennsylvania (2); the response rate of 81% suggested that the majority of the sample population might participate in a carefully planned household HIV survey. This report summarizes the second field test, which was conducted in Dallas County, Texas, from September through December 1989.
The study in Dallas County was designed to resemble a national project. In both Allegheny County and Dallas County, to address public sensitivities and concerns about acquired immunodeficiency syndrome (AIDS)-related issues, planning and implementation activities included the establishment of policy advisory panels representing relevant constituencies and organizations. Procedures were developed to protect the privacy of all persons contacted in the survey.
Survey participants (adults aged 18-54 years) were randomly selected from a sample of 2528 households that were visited by teams of interviewers and phlebotomists. Persons selected were shown a videotape on the purpose and content of the survey; blood specimens obtained from those who consented were tested for antibodies to HIV and hepatitis B core antigen. Each participant provided basic demographic information and answered questions about risk behaviors on an anonymous self-administered questionnaire. The survey sample was geographically stratified and of adequate size to provide a reliable HIV prevalence estimate for the county.
Of the 1724 eligible persons identified, 1446 (84%) consented to participate in the survey. Primary reasons for refusal to participate were lack of interest in participating in surveys (48%) and fear of giving blood (22%).
To assess and reduce potential nonresponse bias, a follow-up study was conducted using a random sample (n=184) of those who initially declined to participate: 50% of these persons were asked only to complete the questionnaire and 50% to complete the questionnaire and provide a blood sample. In the follow-up study, a larger proportion of participants reported intravenous (IV)-drug use (7.0%) and male-to-male sex (16.8%) than those who consented to participate when first contacted (3.1% and 5.1%, respectively).
This evidence of nonresponse bias was used to adjust the overall HIV seroprevalence estimate among 18- to 54-year-old residents of households in Dallas County from 0.3% to 0.4%,* indicating that approximately 4000 persons among an estimated 950,000 persons aged 18-54 years would test positive for antibody to HIV (95% confidence interval=2200-7500 HIV-positive persons) (Table 1). HIV seroprevalence was highest among males, persons aged 25-34 years, and persons who were unmarried. Antibody to hepatitis B core antigen was detected in 7.3% of participants who provided blood samples.
Overall, 20.2% reported having engaged in one or more risk behaviors for HIV infection (Table 2). An estimated 7.3% of males in Dallas County reported having had sex with another male since 1978 and no IV-drug use; 4.5% of males reported having had receptive anal intercourse. Of all adult household residents, an estimated 3.8% reported using IV drugs since 1978 and no male-to-male sex. In addition, 10.3% of respondents reported a history of gonorrhea; 2.4%, syphilis; and 2.3%, genital herpes.
The prevalence of HIV infection was highest among those who reported behavior previously associated with HIV transmission. Males who reported having had sex with another male (but not IV-drug use) since 1978 had an estimated HIV infection prevalence of 8.7%, which increased to 14.0% among those who reported having had receptive anal intercourse. The estimated HIV infection prevalence among those who reported IV-drug use (but no male-to-male sex) in the preceding year was 1.3% (Table 2). All HIV-infected persons in the Dallas survey reported having engaged in risk behavior for HIV infection.
To evaluate the survey estimate of HIV prevalence, several model estimates were prepared using back-calculation, a statistical method that uses observed AIDS incidence and estimated AIDS incubation period distribution (3-7). Comparison of the survey prevalence estimate with back-calculation models indicates that the point estimate (4000 infected persons) is lower than the range of estimates from these models; however, the upper boundary of the survey estimate is within the ranges obtained from two of the models (Table 3). Reported by: BW Dixon, MD, EJ Streiff, MPH, AH Brunwasser, MPH, Allegheny County Health Dept, Pittsburgh, Pennsylvania. CE Haley, MD, A Freeman, MSPH, HG Green, MD, Dallas County Health Dept, Dallas, Texas. Div of HIV/AIDS, Center for Infectious Diseases; Office of the Deputy Director (HIV); Office of Vital and Health Statistics Systems, National Center for Health Statistics, CDC.
Editorial Note: Findings in this feasibility study indicate that persons who engage in risk behaviors for HIV infection are likely to report those behaviors in a self-administered anonymous questionnaire. However, the extent of underreporting of these risk behaviors cannot be readily determined because there are no comparable studies for Dallas County and reporting in the survey could not be verified through comparison with other records because of ethical and privacy concerns. Nonetheless, the behavioral findings are consistent with previous reports (8,9), and the serologic and behavioral findings are consistent within this survey.
The follow-up study permitted adjustment to improve the estimate by compensating for nonresponse bias. Despite this adjustment, estimates derived using back-calculation models exceeded those based on the household survey. Although these models are not a definitive standard for assessing the accuracy of the survey estimate, they suggest that the serologic survey point estimate for Dallas County is too low.
The feasibility study demonstrated that substantial resources are needed to implement such a household survey and that public concern requires intensive community preparation. This study confirmed that nonresponse bias is an important problem and that methods for measurement of nonresponse bias need to be field tested.
CDC is estimating national seroprevalence levels by using AIDS surveillance data and back-calculation models (5) and by using data from HIV serosurveys such as that of childbearing women (10). Based on the findings in this study and the availability of other methods for estimating HIV seroprevalence, CDC has recommended that a national household survey to estimate HIV seroprevalence not be done. However, applications of this methodology in the future could employ approaches for measuring nonresponse bias that 1) include persons for whom reliable independent records of HIV status and risk behaviors are available (this approach may not be feasible because of HIV research-related privacy concerns); and/or 2) employ a dual-frame survey design, including a sample of the general population and samples of special populations (i.e., clients of sexually transmitted diseases clinics, drug-treatment centers, and hospitals). The latter approach might produce more accurate HIV seroprevalence estimates by compensating for biases in each of the surveys.
Copies of the final report of the National Household Seroprevalence Survey Feasibility Study are available from the Office of Vital and Health Statistics, NCHS, CDC, Room 1120, 6525 Belcrest Road, Hyattsville, MD 20782.
In: Turner CF, Miller HG, Moses LE, eds. AIDS sexual behavior and intravenous drug use. Washington, DC: National Academy Press, 1989:447-70.
2. Research Triangle Institute. National Household Seroprevalence Survey Feasibility Study final report. Vols I and II. Research Triangle Park, North Carolina: Research Triangle Institute, 1990.
3. Brookmeyer R, Gail MH. A method for obtaining short-term projections and lower bounds on the size of the AIDS epidemic. J Am Stat Assoc 1988;83:301-8.
4. Brookmeyer R, Damiano A. Statistical methods for short-term projections of AIDS incidence. Stat Med 1989;8:23-34.
5. CDC. HIV prevalence estimates and AIDS case projections for the United States: report based upon a workshop. MMWR 1990;39(no. RR-16).
6. Byers RH Jr, Hessol NA. Back-calculation of HIV infection distributions: convolution of Weibull and log-logistic (Abstract). Vol 2. VI International Conference on AIDS, San Francisco, June 20-24, 1990:255.
7. Longini IM, Clark WS, Byers RH, et al. Statistical analysis of the stages of HIV infection using a Markov model. Stat Med 1989;8:831-43.
8. Fay RE, Turner CF, Klassen AD, Gagnon JH. Prevalence and patterns of same-gender sexual contact among men. Science 1989;243:338-48.
9. CDC. Number of sex partners and potential risk of sexual exposure to human immuno deficiency virus. MMWR 1988;37:565-8. 10. Pappaioanou M, Dondero TJ, Peterson LR, Onorato IM, Sanchez CD, Curran JW. The family of HIV seroprevalence surveys: objectives, methods, and uses of sentinel surveillance for HIV in the United States. Public Health Rep 1990;105:113-9.
Disclaimer All MMWR HTML documents published before January 1993 are electronic conversions from ASCII text into HTML. This conversion may have resulted in character translation or format errors in the HTML version. Users should not rely on this HTML document, but are referred to the original MMWR paper copy for the official text, figures, and tables. An original paper copy of this issue can be obtained from the Superintendent of Documents, U.S. Government Printing Office (GPO), Washington, DC 20402-9371; telephone: (202) 512-1800. Contact GPO for current prices.**Questions or messages regarding errors in formatting should be addressed to firstname.lastname@example.org.
Page converted: 08/05/98
This page last reviewed 5/2/01