Tracking the Epidemic
Two decades after AIDS was first recognized in the United States, the epidemic continues to evolve in new directions. After a decade and a half of steady increase, AIDS incidence and AIDS deaths declined for the first time in 1996 (CDC, 1997), due to the development of newer and more effective antiretroviral therapies that slow disease progression and extend the lives of people with AIDS, and in part, to a combination of successful HIV prevention efforts. However, recent data suggest that the declines in AIDS incidence and deaths may be stabilizing (CDC, 2000a). At the same time, the number of people living with AIDS has steadily increased. Today, the number of people living with AIDS is at an all-time high (CDC, 2000b). The demographics of the epidemic also have changed considerably over time. While the proportion of AIDS cases among men who have sex with men has declined, the proportion of cases in women, youth, and racial and ethnic minorities has increased (CDC, 2000b). In addition, although the epidemic remains largely concentrated in urban areas, growth in AIDS cases in rural areas has been dramatic in recent years (CDC, 2000b) (see Appendix A for a more detailed discussion of the changing epidemic).
Much of what is known about the epidemic comes from the national AIDS case surveillance system. In this chapter, the Committee discusses the present system and its limitations, the rationale for implementing a surveillance approach to track the incidence of HIV, and the limitations of current HIV surveillance systems in measuring incidence. Further, the Committee recommends a new surveillance approach that can provide population-based estimates of HIV incidence.
NATIONAL AIDS SURVEILLANCE SYSTEM
Since the beginning of the epidemic, surveillance efforts have emphasized determining the number and characteristics of individuals diagnosed with AIDS. The current national AIDS case surveillance system, which was implemented prior to both the identification of HIV as the etiologic agent of AIDS and the development of an antibody test to determine HIV infection, was originally based on epidemiological investigations of an end-stage syndrome (Gostin et al., 1997). Each state requires that all patients diagnosed with AIDS be reported by name to the local, state, and/or territorial health departments. These reports are then forwarded (without names but with unique identifiers) to the Centers for Disease Control and Prevention (CDC), where a national surveillance database is created and analyzed. This surveillance system provides uniform data on trends and distribution of individuals diagnosed with AIDS. Standard records for each case include information on sex, race and ethnicity, state of residence (and metropolitan area, if relevant), mode of exposure to HIV, age at diagnosis, month of AIDS diagnosis, date reported, and other data. Although there are some reporting delays in the system, the data are relatively complete (more than 85 percent of AIDS cases) (Schwarcz et al., 1999; Buehler et al., 1992; Rosenblum et al., 1992) and statistical methods are available to adjust for both reporting delays and incompleteness (Green, 1998). Data from this surveillance system have been used as the basis for allocating federal resources for HIV treatment and care, and as the basis for planning local HIV prevention services (CDC, 1999a).
Until the era of potent antiretroviral therapies, AIDS case reporting, although imperfect, provided a relatively accurate picture of trends in HIV infection, especially relative HIV prevalence in groups defined by geography, race and ethnicity, and primary mode of infection. Estimates of HIV incidence and prevalence were made by statistical techniques, such as calculating backward from reported AIDS cases according to well-established patterns of disease progression (Brookmeyer and Gail, 1994). Recent developments in therapy for HIV and AIDS have at least temporally decoupled HIV infection and its progression to AIDS (Hammer et al., 1996; Collier et al., 1996). As a result, the timing of the progression from HIV infection to AIDS and from AIDS to death is increasingly difficult to predict, making HIV incidence and prevalence estimates based on AIDS cases much less accurate (CDC, 1999a). Consequently, AIDS case reporting is no longer adequate to monitor trends in HIV infection. The United States now faces the challenge of developing an effective HIV surveillance system that can predict where the epidemic is headed.
RATIONALE FOR A NATIONAL SYSTEM OF HIV SURVEILLANCE
In considering how the current surveillance system might be optimally restructured, it is useful to examine the goals of surveillance and the types of data needed to achieve those goals. Public health officials, policy makers, researchers, and community groups use surveillance data to accomplish a variety of goals, including:
tracking the course of the epidemic over time and place, in subpopulations, and by risk factors in order to assess its public health impact and to identify possible prevention strategies;
evaluating access to and effectiveness of treatment and prevention efforts; and
allocating resources for prevention, treatment, and care services, and for determining research priorities (Johri et al., 1998).
A variety of different types of data are needed to meet these goals. For example, data on HIV and AIDS incidence and prevalence, demographics, and risk factors are needed to track the course of the epidemic and to identify the populations needing treatment services. Incidence data are needed to judge the effectiveness of prevention programs and to set priorities for prevention programs (see Chapter 3). Optimally, an HIV surveillance system would also allow tabulation of sociodemographic and behavioral risk information on newly infected individuals which would offer a more precise picture of the HIV epidemic in the United States (see Text Box 2.1).
In its analysis and recommendations, the Committee makes the distinction between “surveillance” and “case finding.” Surveillance is a statistical activity intended to provide for the “ongoing, systematic collection, analysis, interpretation, and dissemination of risk factor, exposure, and/or outcome-specific data for use in public health practice” (Thacker, 1994). Case finding, on the other hand, is intended to identify individuals who can benefit from early intervention in a disease process. Both surveillance and case finding rely on HIV testing, but for different purposes—surveillance to gather population-based data and case finding to identify infected individuals.
HIV CASE REPORTING
In response to concerns about the limitations of the current AIDS surveillance system in providing accurate information about trends in the HIV epidemic, the CDC recommends that all states and territories extend
TEXT BOX 2.1
To help in improving the design and evaluation of prevention strategies, additional surveillance information is needed on behaviors that put people at risk for HIV. Currently, assessment of HIV risk behaviors is conducted on three levels. First, the CDC regularly conducts behavioral surveillance using such instruments as the Behavioral Risk Factor Surveillance System, the Youth Risk Behavior Surveillance System, and the National Survey for Family Growth (IOM, 1997). These surveys provide very general information about HIV testing and some sexual and drug use-related behaviors, but are very limited in the amount of data they provide regarding specific risk practices, particularly among high-risk subgroups, such as men who have sex with men and injection drug users. Behavioral surveys conducted among HIV-infected populations may provide information on practices that increase risk for viral transmission to sex and drug-using partners, but do not yield data on at-risk, uninfected persons. Behavioral assessments among high-risk populations fill this void, but often are constrained in their representativeness and generalizability due to sampling biases (e.g., as with convenience sampling).
Although each of these surveillance methods has limitations, together they might provide adequate information with which to better develop, implement, and evaluate prevention programs among populations at high risk of infection. The usefulness of data derived from these assessments, however, is often hampered by the lack of comparability between survey instruments and items that are intended to measure the same behavior (e.g., frequency of condom use). In order to provide some guidance in this matter, the CDC currently is developing a set of “core” items for use in HIV/STD behavioral surveillance surveys. This effort will be particularly useful in providing consistency for the measurement of sexual and drug use-related risk practices. Further, detailed information about the social contexts in which risk behaviors occur is sorely needed but rarely assessed. Assessments of such information should be integral components of behavioral assessments, as they serve to provide a better understanding of the dynamics of risk behavior and the social issues that need to be better addressed by prevention interventions. Therefore, the Committee endorses prior recommendations regarding the establishment of a national survey that can determine the prevalence and correlates of HIV risk-taking behavior (IOM, 1995).
their AIDS surveillance activities to include case reporting of HIV infection (CDC, 1999a). The CDC maintains that HIV case reporting will provide additional epidemiological data about HIV-infected populations to enhance prevention efforts, improve allocation of treatment resources, and assist in evaluating the impact of HIV prevention programs (CDC, 1999a). However, data from the HIV reporting system are incomplete in several important ways. In contrast to the AIDS case reporting system, which is relatively complete, the HIV reporting system collects data only from persons who choose to be tested and who do so at non-anonymous testing sites (i.e., where the HIV test result is linked with identifying
information, including patient and provider names). Thus, HIV case reporting data exclude individuals who are infected but have not been tested, as well as those who use anonymous testing sites or home collection test kits (CDC, 1999a). Because of this selectivity, HIV case reporting by name is not representative of the larger population of infected persons. Further, because reported HIV cases could represent infections that are anywhere from a few weeks to a few years old, the data would reflect the time that individuals chose to be tested rather than when the individual became infected. As a result, HIV case reporting data provide only partial information about the number of existing HIV cases (HIV prevalence), rather than information about new HIV infections (HIV incidence) (Johri et al., 1998).
HIV case reporting can be implemented in one of two ways: through use of name reporting or through use of coded identifiers. As of June 1, 2000, a total of 35 states and the U.S. Virgin Islands had implemented HIV case surveillance using the same confidential name-based case reporting system used for AIDS cases. Two of these states (Connecticut and Oregon) conduct only pediatric HIV surveillance (CDC, 2000c). Name-based HIV case reporting has several potential benefits. First, name-based reporting can facilitate linkages between HIV case registries with those of other communicable diseases (e.g., syphilis and tuberculosis) at the state and local level. This cross-referencing can be used by public health officials to obtain a more comprehensive picture of their local epidemic and effectively target local level prevention and care services (K. Toomey, personal communication). Name-based reporting has also been suggested as a case-finding method by which HIV-infected individuals could be linked to treatment services (Colfax and Bindman, 1998). However, reporting HIV cases to state health departments does not automatically ensure that HIV-infected persons will receive beneficial linkages to care.1 In a recent study, Osmond and colleagues (1999) found that contact with a health department after testing HIV-infected was not associated with receipt of timelier care. Finally, HIV name-based reporting may enhance partner notification programs that are used to identify individuals at high risk of infection (Colfax and Bindman, 1998).2
One of the major concerns with name-based reporting is the potential for breaches of confidentiality. For example, groups that were both heavily affected by HIV/AIDS and already stigmatized by society (e.g., gay men
and injection drug users) initially felt that compiling a list of the names of HIV-infected persons would compound the stigma they were already experiencing. There also were concerns that such a list could be used for discriminatory purposes should it become publicly available (Gostin et al., 1997). However, there have been very few reported incidents of intentional misuse of this information (Colfax and Bindman, 1998). In addition, there is substantial concern that name-based reporting may deter individuals at risk of infection from being tested, thus delaying their access to counseling or treatment services (Colfax and Bindman, 1998; Woods et al., 1999). Some studies have suggested that name-based reporting policies might deter or postpone some high-risk individuals from seeking testing (CDC, 1998; Hecht et al., 1997, 2000). For these reasons, the CDC supports adherence to strict confidentiality protections of testing and surveillance data and the availability of anonymous testing options (CDC, 1999a).
Concerns that stigma, discrimination, and breaches in confidentiality might deter some individuals from being tested have led some states such as Maryland and Massachusetts to implement a coded system of HIV case reporting based on “unique identifiers” (CDC, 1999a; Solomon et al., 1999). This system is similar to HIV name-based case reporting, except a code number is created and reported rather than the individual’s name. Each code number is based on information specific to the individual (e.g., Social Security number). If all elements of the code are complete and accurate, the code number is unique enough to avoid duplicate records (Solomon et al., 1999), while still allowing for follow-up to obtain additional information.
While the unique identifier case reporting system does address concerns about confidentiality, the data obtained from this type of surveillance have the same statistical drawbacks as name-based reporting with regard to timeliness and test site selection bias. Although both systems provide data that are, in a sense, more timely than AIDS case reporting for estimating HIV incidence, they still provide an incomplete and inaccurate picture of the HIV epidemic.
POPULATION-BASED HIV INCIDENCE ESTIMATION
In order to meet the goal of providing accurate HIV incidence information, a new surveillance approach is needed. With this in mind, the Committee proposes the use of sentinel surveillance, a method that would obtain incidence data from targeted samples of “sentinel” populations using advanced testing technology, in combination with statistical modeling to extrapolate these incidence data to larger subsets of the population or the population as a whole (Johri et al., 1998).
Current estimates of HIV/AIDS prevalence and incidence are based in part on a “family of surveys” in a variety of populations: screening results of blood samples derived from programs testing special populations (e.g., blood donors and applicants to the military and Job Corps) and testing of anonymous blood specimens from smaller studies (e.g., at sexually transmitted disease clinics, drug treatment centers, and adolescent medical clinics) (Pappaioanou et al., 1990). These surveys, however, draw from non-representative convenience samples rather than from probability samples.3 As a result, many sources of potential bias exist. Changes in measurement techniques also introduce bias. For example, the group tested may not be representative of the population, or the populations being tested may not be stable over time.
Restructuring the family of HIV surveys to permit better statistical estimates would optimize knowledge of HIV incidence. Indeed, the National Research Council (NRC) has recommended that the CDC family of surveys be reformulated as two-stage probability samples (NRC, 1989). First, a random sample of nationwide drug treatment centers, sentinel hospitals and primary care facilities, clinics devoted to sexually transmitted diseases and tuberculosis, and clinics serving women of reproductive age would be chosen. For reasons of statistical efficiency, settings that are likely to serve a large number of HIV-infected persons, such as drug treatment centers, should be more intensively sampled. However, in order to get nationally and locally representative estimates, sampling should not be limited to such sites. Second, random samples of blood from clients of these facilities would be chosen and tested anonymously for HIV. Thus, the survey sites must be ones in which blood is normally drawn for routine testing. The clustered nature of the final sample must be taken into account in determining the appropriate sizes of the first and second size samples. Additionally, in order to extrapolate to a broader population of individuals who visit these sites, questions would be added to nationally representative surveys, such as the National Health Interview Survey, about whether individuals had visited the kind of facilities targeted by the family of HIV serosurveys. With these results, accurate statistical estimates can be made of the prevalence of HIV in demographically defined groups. Reinstating some representative subset of the Survey of Childbearing Women, which was based on anonymous HIV testing of newborn
children and which was discontinued by the CDC in 1994, also would be an important part of this effort.
Further, the recommended serosurveillance effort should also incorporate the use of recent advances in HIV testing, such as the detuned assay (Janssen et al., 1998). The detuned assay is an advanced testing technique that determines whether an individual has been infected with HIV within approximately the past four to six months4 (Janssen et al., 1998, CDC, 1999b). Combined with probability sampling approaches (Kaplan and Brookmeyer, 1999) that would improve estimates of HIV prevalence, the use of detuned assays could yield more accurate national estimates of HIV incidence. It should be noted that because blood samples are tested anonymously, sentinel serosurveillance methodologies are not suitable for facilitating linkages to care.
Surveillance systems can be evaluated in terms of a number of characteristics, including simplicity, flexibility, acceptability, sensitivity (the ability to include existing cases), positive predictive value (the likelihood that included cases do in fact have the intended disease or condition), representativeness (the ability to describe the occurrence of a health event over time and its distribution in the population by place and personal characteristics), and timeliness (Buehler, 1998). Table 2.1 compares the alternative approaches using three of these criteria (timeliness, representativeness, acceptability).
From this analysis, the Committee concludes that a new surveillance system focused on HIV incidence is needed in order to more effectively guide HIV prevention planning, resource allocation, and evaluation decisions at the national, state, and local levels. To the extent possible, the system would provide estimates at the state and local level and for the population groups at highest risk for HIV infection. We believe that a system of population-based HIV incidence estimation will provide the most accurate and timely data for these objectives. Too often, surveillance data are haphazardly collected as incidental by-products of clinical or social services to selected patients rather than as part of a coordinated strategy. To reflect the importance of HIV/AIDS to our society and to do a better job of allocating scarce resources for HIV prevention (see Chapter 3) a more effective surveillance system is needed. Therefore, the Committee recommends that:
TABLE 2.1 Comparison of HIV Surveillance Approaches
The Centers for Disease Control and Prevention create a surveillance system that can provide national population-based estimates of HIV incidence. The recommended surveillance system would estimate new HIV infections using blinded serosurveys of well-characterized sentinel populations (e.g., drug users in treatment, people attending sexually transmitted disease clinics and tuberculosis clinics, clinics serving women of reproductive age), surveys that characterize the populations served by those sites, and advanced testing technologies that are able to identify recent HIV infections.
The Committee is aware that previous attempts to do HIV surveillance have been controversial (Bayer, 1997), particularly before effective therapies became available. In the 1980s, some groups perceived the “public health” responses to AIDS as being aimed at identifying HIV carriers and protecting the blood supply, without any regard for the rights or protection of those infected. In the 1990s, legislation to mandate HIV testing of newborn children without the consent of their mothers (whose HIV antibodies were actually being tested) added to the distrust of public health officials that was already felt by some groups (IOM, 1999). Thus, in
many quarters, HIV case reporting seemed to offer little to HIV-infected people.
Now that effective therapies are available and individuals have a variety of opportunities for confidential HIV testing, there are strong reasons for people to be tested if for no other reason than to protect their own health. However knowing one’s serostatus is only helpful if it results in obtaining medical care and treatment for HIV infection. The Committee’s proposal for HIV surveillance based on anonymous test results at sentinel sites separates surveillance (a statistical activity) from case finding, and it has the potential to both provide better estimates of HIV incidence and avoid the controversies of case reporting. HIV case finding, used for linking infected persons to care, partner notification, or contact tracing, is a separate prevention activity and should be judged in reference to other uses of prevention funds (see Chapter 3).
Bayer R. 1997. Science, politics, and AIDS prevention policy. Journal of Acquired Immune Deficiency Syndrome and Human Retrovirology 14 (Suppl 2):S22–S29.
Brookmeyer R and Gail MH. 1994. AIDS Epidemiology: A Quantitative Approach. New York: Oxford University Press.
Buehler JW. 1998. Surveillance. In Rothman KJ and Greenland S, (Eds.), Modern Epidemiology, 2nd ed. Philadelphia: Lippincott Williams and Williams.
Buehler JW, Berkelman RL, Stehr-Green JK. 1992. The completeness of AIDS surveillance. Journal of Acquired Immune Deficiency Syndrome 5(3):257–264.
Centers for Disease Control and Prevention (CDC). 2000a. U.S. AIDS Cases, Deaths, and HIV Infections Appear Stable. AMA Press Briefing, July 8.
Centers for Disease Control and Prevention (CDC). 2000b. HIV/AIDS Surveillance Report, 1999 11(2). Atlanta: CDC.
Centers for Disease Control and Prevention (CDC). 2000c. Current Status of HIV Infection Reporting—May 2000. Atlanta: CDC.
Centers for Disease Control and Prevention (CDC). 1999a. Guidelines for national human immunodeficiency virus case surveillance, including monitoring for human immunodeficiency virus infection and acquired immunodeficiency syndrome. Morbidity and Mortality Weekly Report 48(RR-13):1–31.
Centers for Disease Control and Prevention (CDC). 1999b. New HIV Test Provides Unprecedented Look at Level of New Infections in High-Risk Groups. Press Briefing. 1999 National Prevention Conference. August 29 – September 1. Atlanta, GA.
Centers for Disease Control and Prevention (CDC). 1998. HIV testing among populations at risk for HIV infection—nine states. Morbidity and Mortality Weekly Report 47:1086–1091.
Centers for Disease Control and Prevention (CDC). 1997. Update: Trends in AIDS incidence—United States, 1996. Morbidity and Mortality Weekly Report 46:861–867.
Colfax GN and Bindman AB. 1998. Health benefits and risks of reporting HIV-infected individuals by name. American Journal of Public Health 88(6):876–879.
Collier AC, Coombs RW, Schoenfeld DA, Bassett RL, Timpone J, Baruch A, Jones M, Facey K, Whitacre C, McAuliffe VJ, Friedman HM, Merigan TC, Reichman RC, Hooper C, Corey L. 1996. Treatment of human immunodeficiency virus infection with saquinavir, zidovudine, and zalcitabine. AIDS Clinical Trials Group. New England Journal of Medicine 334(16):1011–1017.
Gostin LO, Ward JW, Baker AC. 1997. National HIV case reporting for the United States. A defining moment in the history of the epidemic. New England Journal of Medicine 337(16):1162–1167.
Green TA. 1998. Using surveillance data to monitor trends in the AIDS epidemic. Statistics in Medicine (2):143–154.
Hammer SM, Katzenstein DA, Hughes MD, Gundacker H, Schooley RT, Haubrich RH, Henry WK, Lederman MM, Phair JP, Niu M, Hirsch MS, Merigan TC. 1996. A trial comparing nucleoside monotherapy with combination therapy in HIV-infected adults with CD4 cell counts from 200 to 500 per cubic millimeter. AIDS Clinical Trials Group Study 175 Study Team. New England Journal of Medicine 335(15):1081–1090.
Hecht FM, Chesney MA, Lehman JS, Osmond D, Vranizan K, Colman S, Keane D, Reingold A, Bindman AB. 2000. Does HIV reporting by name deter testing? MESH Study Group. AIDS 14(12):1801–1808.
Hecht FM, Coleman S, Lehman JS, Vranizan K, Keane D, Bindman AB, Chesney M, and the MESH Study Group, San Francisco General Hospital, UCSF, San Francisco, CA, and CDC, Atlanta, GA. 1997. Named reporting of HIV: Attitudes and knowledge of those at risk (Abstract). Journal of General Internal Medicine 12(Suppl 1):108.
Institute of Medicine. 1999. Reducing the Odds: Preventing Perinatal Transmission of HIV in the United States. Stoto MA, Almario DA, and McCormick MC (Eds.). Washington, DC: National Academy Press.
Institute of Medicine. 1997. The Hidden Epidemic: Confronting Sexually Transmitted Diseases. Eng T and Butler W (Eds.). Washington, DC: National Academy Press.
Institute of Medicine. 1995. AIDS and Behavior: An Integrated Approach. Auerbach J, Wypijewska C, and Brodie K (Eds.). Washington, DC: National Academy Press.
Janssen RS, Satten GA, Stramer SL, Rawal BD, O’Brien TR, Weiblen BJ, Hecht FM, Jack N, Cleghorn FR, Kahn JO, Chesney MA, Busch MP. 1998. New testing strategy to detect early HIV-1 infection for use in incidence estimates and for clinical and prevention purposes. Journal of the American Medical Association 280(1):42–48.
Johri M, Kaplan EH, Levi J, Novick A. 1998. New approaches to HIV surveillance: Means and ends. A Summary Report of the Law, Policy, and Ethics Conference on HIV Surveillance, May 21–22, 1998. Yale University Center for Interdisciplinary Research on AIDS.
Kaplan EH and Brookmeyer R. 1999. Snapshot estimators of recent HIV incidence rates. Operations Research 47(1):29–37.
National Research Council. 1989. AIDS: Sexual Behavior and Intravenous Drug Use. Turner CF, Miller HG, and Moses LE (Eds.). Washington, DC: National Academy Press.
Osmond DH, Bindman AB, Vranizan K, Lehman JS, Hecht FM, Keane D, Reingold A. 1999. Name-based surveillance and public health interventions for persons with HIV infection. Multistate Evaluation of Surveillance for HIV Study Group. Annals of Internal Medicine 131(10):775–779.
Pappaioanou M, Dondero TJ Jr, Petersen LR, Onorato IM, Sanchez CD, Curran JW. 1990. The family of HIV seroprevalence surveys: Objectives, methods, and uses of sentinel surveillance for HIV in the United States. Public Health Reports 105(2):113–119.
Rosenblum L, Buehler JW, Morgan MW, Costa S, Hidalgo J, Holmes R, Lieb L, Shields A, Whyte BM. 1992. The completeness of AIDS case reporting, 1988: A multisite collaborative surveillance project. American Journal of Public Health 82(11):1495–1499.
Schwarcz SK, Hsu LC, Parisi MK, Katz MH. 1999. The impact of the 1993 AIDS case definition on the completeness and timeliness of AIDS surveillance. AIDS 13(9):1109–1114.
Solomon L, Flynn C, Caldeira E, Wasserman MP, Benjamin G. 1999. Evaluation of a statewide non-name-based HIV surveillance system. Journal of Acquired Immunodeficiency Syndromes 22:272–279.
Thacker SB. 1994. Historical Development. In Teutsch SM and Churchill RE. (Eds.), Prin ciples and Practice of Public Health Surveillance. New York: Oxford University Press.
Toomey K. Georgia Department of Human Resources, Personal Communication, June 16, 2000.
Wilcox LS and Marks JS. 1994. Overview. From Data to Action: CDC’s Public Health Surveillance for Women, Infants and Children. Atlanta: CDC.
Woods WJ, Dilley JW, Lihatsh T, Sabatino J, Adler B, Rinaldi J. 1997. Name-based reporting of HIV-positive test results as a deterrent to testing. American Journal of Public Health 89(7):1097–1100.