Multistate Evaluation of Anonymous HIV Testing and Access to Medical Care FREE

BOTH CONFIDENTIAL and anonymous antibody testing for the human immunodeficiency virus (HIV) have been available at public testing sites in the United States since 1985. In confidential testing, a person's name is linked to the specimen, and the test result is recorded in a medical chart with a name. Early in the epidemic, the stigma associated with testing positive for HIV focused attention on the potential for breaches in the confidentiality of an HIV test result. Concerned that anxiety about the potential loss of confidentiality would deter some at-risk persons from voluntarily seeking testing for HIV, many state and local public health departments made this test available on an anonymous as well as a confidential basis. In anonymous testing, a unique identifier (typically a number) rather than a patient's name is used to link the specimen and the result to the patient. Anonymous test results are not recorded in a medical chart that has a patient's name. The availability of an anonymous HIV testing option has differed over time across states and localities. Currently, 40 states have publicly funded anonymous testing sites for HIV, and all 50 states have publicly funded confidential HIV testing sites.

Human immunodeficiency virus is the only infectious disease for which anonymous testing is publicly funded, an exception that has been controversial. Proponents of anonymous testing believe that it encourages persons who would not otherwise seek testing to learn their HIV infection status by eliminating the concern about potential loss of confidentiality. Persons tested anonymously who learn that they are HIV positive may be motivated by their test result to seek medical care earlier in the course of the disease than they might had only confidential testing been available. Some studies have suggested that anonymous testing increases the number of people who are willing to be voluntarily tested for HIV. In North Carolina, counties that offered anonymous testing experienced greater growth in testing than did counties that continued to offer only confidential testing.¹ Similarly, with the introduction of anonymous testing in Arizona and Oregon,^{2 - 3} more people obtained testing than when only confidential testing was available. However, the findings have not been consistent; the Colorado State Health Department did not detect a meaningful increase in HIV testing with the introduction of anonymous HIV testing.⁴

Because people who test HIV positive anonymously cannot be individually identified, reporting systems that rely on the results of anonymous testing are prone to measurement error. It can be difficult to detect repeat tests, and the potential exists for duplicate reporting. Anonymous testing may undermine partner notification.⁵ Furthermore, anonymous testing eliminates the opportunity to recontact persons who do not return for their test results or to assist HIV-infected persons in obtaining medical care.

Because studies have been small, have been performed in only 1 state, or did little to control for differences in the characteristics of persons who used anonymous vs confidential testing, it has been difficult to draw clear conclusions about the value of anonymous HIV testing. We used data collected as a part of a cooperative project between the University of California, Berkeley, the Centers for Disease Control and Prevention (CDC), Atlanta, Ga, and several state health departments to evaluate anonymous HIV testing. We assessed the association of the type of HIV test (anonymous or confidential) with when in the course of the disease persons with acquired immunodeficiency syndrome (AIDS) (1) learned of their HIV infection and (2) sought HIV-related medical care.

The AIDS Patient Survey was conducted in Arizona, Colorado, Mississippi, Missouri, New Mexico, North Carolina, Oregon, and Texas. Because nearly all HIV-infected persons are thought to progress to AIDS eventually and because AIDS surveillance is estimated to be 80% to 96% complete,⁶ reported AIDS cases provide a population-based sample of the experience of HIV-infected persons that can potentially avoid biases that may be present in venue-based samples.

In each state we sought to interview, after obtaining consent, all persons who were described as having newly diagnosed AIDS in a 1-year period or a probability sample of new cases, depending on the projected incidence of new AIDS diagnoses in the state. The sampling frame was persons newly diagnosed as having AIDS reported to the state health department through the HIV/AIDS Reporting System (HARS) from May 1995 through December 1996, who were alive at time of report, who were at least 18 years old, and whose AIDS diagnosis had been made within 12 months of the date of their report to the health department.

An expected number of persons with newly diagnosed AIDS was estimated from the number reported from the previous year who met the sampling frame criteria. In states with an expected incidence of fewer than 500 cases, all new cases were sampled (Arizona, Mississippi, and New Mexico). In the remaining states, a probability sample was stratified by 4 HIV mode-of-exposure groups based on reported behavioral information in HARS: (1) men who have sex with men (MSM), including those with a reported history of injection drug use; (2) heterosexual injection drug users; (3) cases reported with no identified risk; and (4) all other modes of exposure (heterosexual contact, transfusion, hemophilia). To get adequate numbers in each stratum for analysis, we calculated sampling fractions with the goal of sampling equal numbers from each stratum. Colorado, Missouri, and Oregon sampled MSM and took all cases in the other 3 strata; North Carolina sampled 3 strata and took all in the other stratum; and Texas sampled all 4 strata. Uniform random numbers were generated for each new case in the 4 strata, and a new case with a random number equal to or less than the sampling fraction was selected for the study.

Sampled cases were considered eligible for the study if they were living in the state, English or Spanish speaking, and healthy enough to consent to and complete an interview. To avoid biasing our response rate upward by delaying the performance of the interview, patients who had died before the time of first contact were counted in the denominator of eligibles if contact had not been made within 6 months of report. Public health surveillance personnel in each state developed procedures for contacting and interviewing potential subjects.

All procedures were monitored by the University of California and CDC to ensure uniform methods across the states. Surveillance personnel completed an outcome report form for each sampled case, which indicated the consent process and the final outcome. Subjects were interviewed in either Spanish or English. The instrument was translated into Spanish and then back-translated to English before a final Spanish version was produced. Interviewers and supervisors from the state health departments were trained in joint training sessions in conducting a standard interview. States used between 1 and 4 interviewers to administer the survey and all study sites were visited at least once by University of California and CDC investigators to assess the consistency of their technique. All completed interviews and outcome report forms were stripped of personal identifiers, copied, and mailed to the University of California for data entry and conversion into electronic Statistical Analysis System (SAS Institute Inc, Cary, NC) files for analysis.

We compared the characteristics of respondents who were tested anonymously with those who were tested confidentially and examined whether the type of HIV test was associated with when in the course of the disease a subject sought HIV testing and HIV-related medical care. Date of AIDS diagnosis was extracted from the state HARS databases and combined with the interview data for analysis. Type of HIV testing was classified as anonymous or confidential depending on whether the subject reported giving a number (anonymous) or a name (confidential) to get the HIV test result. Subjects who in response to an explicit question said that they gave a false name were excluded from the analysis. To assess the validity of our method for classifying the type of HIV test, we compared the subject's report of having given a number or a name to obtain their test result with the type of testing site. Assuming that testing in a physician's office, hospital, jail or prison, or blood bank should have been reported as testing by name (confidential testing), we found that 96.4% of subjects tested in those settings reported they had received their results by name. Of those who reported that they had received their test result by number (anonymous testing), only 6.4% reported testing in one of those settings.

We limited our analysis to respondents who first tested HIV positive in the state from which they were sampled, lived in states that offered both anonymous and confidential testing (Mississippi excluded), and voluntarily sought HIV testing as opposed to being required to obtain a test because of regulations associated with prisons, drug treatment programs, the military, insurance plans, or blood banks. Thus, subjects were considered volunteers for testing if they, in response to a checklist of questions, reported that their reason for testing was (1) they felt sick and wanted to find out whether they had HIV, (2) they thought they might have HIV even if they did not feel sick, (3) someone told them that they should get tested, or (4) someone from the health department told them that they had had contact with an infected sex or needle-sharing partner.

In comparing the characteristics of persons tested anonymously vs persons tested confidentially, we tested differences in the proportions by using the χ² statistic. We examined the association of anonymous and confidential testing with several intervals: time from HIV-positive test result to AIDS and this interval's 2 subcomponents: (1) time of HIV-positive test result to first HIV-related medical care and (2) time from first HIV-related medical care to AIDS. We used the date of AIDS diagnosis to anchor comparisons of the HIV-positive test result date and the HIV-related medical care date. Date of first HIV-positive test result and date of first medical care for HIV infection were self-reported as a month and a year. Time intervals used in analysis were constructed from these dates and the date of AIDS diagnosis as reported to HARS. We compared the mean time intervals among HIV testing, HIV-related medical care, and AIDS diagnosis for persons tested anonymously and persons tested confidentially. Time intervals that included an AIDS diagnosis were also stratified by whether the diagnosis was based on an opportunistic infection or a CD4⁺ cell count of less than 0.20 × 10⁹/L (200/µL).

We compared subjects on the basis of whether they had symptoms of weight loss without dieting, fevers, heavy night sweats, diarrhea, oral thrush, frequent vaginal yeast infections, memory problems, shingles, pneumonia, Kaposi sarcoma, lymphoma, meningitis, or tuberculosis at the time they learned they had HIV. Subjects who said yes to any of these conditions were considered symptomatic at the time of the first HIV-positive test result. To estimate HIV disease severity at the time of first HIV-related medical care, we compared the mean self-reported first CD4⁺ cell counts of persons tested anonymously and persons tested confidentially. To estimate the quality of HIV-related medical care for persons tested anonymously and persons tested confidentially, subjects were asked to report whether their HIV-related medical care had ever included tuberculin skin testing, taking zidovudine for at least 1 day, and taking trimethoprim-sulfamethoxazole (Septra, Bactrim, Cotrim) or aerosolized pentamidine as a measure of Pneumocystis carinii pneumonia (PCP) prophylaxis.

To isolate the independent contribution of the type of testing on the time intervals and the first CD4⁺ cell count, we performed multivariate linear regression analyses that controlled for differences in the characteristics of persons tested anonymously vs confidentially. Marginal differences across states were controlled for using a state of residence indicator in the multivariate analyses. Means from multivariate analyses are the estimated least squares means from linear models. Because the distributions of time intervals and CD4⁺ cell counts were skewed by some higher values, we repeated our multivariate analyses by using log transformations of the time intervals and CD4⁺ cell counts. These analyses did not appreciably alter the significance of the results pertaining to anonymous vs confidential testing; therefore, for the purposes of providing measures of effect that are easily interpreted, we have chosen to display the results based on the nontransformed mean time intervals and CD4⁺ cell counts.

The study was approved by institutional review boards at the University of California, the individual states that required review, and review boards at local institutions as required within some states.

In the 8 participating states, 3321 AIDS cases were sampled from May 1995 through December 1996; of those, 2801 met eligibility criteria. Overall, 1913 (68.3%) of 2801 eligible AIDS cases were interviewed in the AIDS Patient Survey. We excluded 1078 respondents from the analysis because they initially tested HIV positive in a different state from the one in which they were sampled (363), they were from a state that did not have anonymous testing (262), their reason for testing was not voluntary (247), they provided a false name at a confidential testing site (55), or they did not have complete data for all the variables used in the analysis (151). Of the remaining 835 subjects, 192 (23%) reported that their first positive test result had been from an anonymous test (Table 1). Persons tested anonymously tended to be younger, white, slightly more educated than persons tested confidentially, and more likely at risk for HIV because they were MSM. Persons tested confidentially were significantly more likely to have had a regular source of care before their first HIV-positive test result and to have had HIV-related symptoms at the time they received the test; however, half of the persons tested anonymously were also symptomatic.

Persons tested anonymously presented earlier in the course of HIV disease for testing and care than did persons tested confidentially. The mean time from learning they were HIV positive to the diagnosis of AIDS was almost a year and a half longer (529 days) for those tested anonymously than for those tested confidentially (Figure 1). The mean time was 1246 days for persons tested anonymously and 718 days for persons tested confidentially. Most of this difference was in the length of time in HIV-related medical care. Persons tested anonymously received an average of 387 more days in HIV-related medical care before an AIDS diagnosis than did persons tested confidentially. Comparisons of the median times from knowledge of being HIV positive to AIDS were even greater between persons tested anonymously and persons tested confidentially. The median time was 929 days among persons tested anonymously and 90 days among persons tested confidentially. An additional indicator that persons tested anonymously came earlier for testing and medical care than did persons tested confidentially was the significantly higher first CD4⁺ cell count (0.427× 10⁹/L vs 0.267 × 10⁹/L) despite the longer unadjusted interval between the HIV-positive test result and medical care.

To isolate the independent contribution of the type of HIV test on the timing of testing and medical care, we adjusted our results to account for differences in the characteristics of persons who sought each type of test. In the multivariate analysis, several characteristics were associated with the length of time between a person's learning of a positive HIV test result and receiving an AIDS diagnosis. Among HIV exposure groups, MSM and injection drug users had a significantly longer period of knowing they were HIV positive before their AIDS diagnosis (Table 2). The strongest predictor of the length of time between knowledge of HIV positivity and AIDS was having symptoms at the time of the HIV-positive test result. Having symptoms at the time of the HIV-positive test result decreased the length of time between knowledge of being HIV positive and AIDS by 819 days.

After adjustment for the subject's age, sex, race/ethnicity, education, income, insurance status, HIV exposure group, if the respondent had a regular source of care or symptoms at the time of the HIV-positive test result, and the state of residence, anonymous testing remained significantly associated with earlier medical care (Figure 2). Although the difference in the number of days between the positive test result and first medical care was no longer significant between the 2 groups, the length of time in medical care before AIDS was almost 8 months longer (221 days) for persons tested anonymously compared with persons tested confidentially. The mean adjusted first CD4⁺ cell count was also 0.092 × 10⁹/L higher for persons tested anonymously than for persons tested confidentially.

Persons tested confidentially were more likely than those tested anonymously (35% vs 16%) to have an AIDS diagnosis based on an opportunistic infection rather than on a CD4⁺ cell count of less than 0.20 × 10⁹/L. Accounting for this difference in how AIDS was diagnosed in the 2 testing groups further expands the adjusted difference in the duration of HIV-related medical care before AIDS diagnosis from 221 to 230 days.

Comparisons of tuberculin skin testing and the use of zidovudine and PCP prophylaxis suggest that care was similar for the 2 testing groups. Ninety-one percent of persons tested anonymously vs 89% of persons tested confidentially reported that they had received tuberculin skin testing during the course of their HIV-related medical care. Ninety-eight percent of persons tested anonymously vs 95% of persons tested confidentially were offered zidovudine, and 73% in each testing group had been given PCP prophylaxis. None of these testing or treatment differences were significant between the 2 groups.

In this multistate study, we found that anonymous testing was sought by approximately a quarter of HIV-positive persons who had been tested voluntarily before an AIDS diagnosis. Anonymous testing for HIV infection was associated with earlier testing and medical care. As a result of this earlier testing and care, persons tested anonymously received the potential benefits of a significantly longer period of HIV-related medical care compared with persons tested confidentially.

Although the determination of the type of HIV test, CD4⁺ cell counts, and the intervals between HIV testing, medical care, and AIDS are in large part dependent on self-report, we suspect that the importance of this information for our respondents makes it reasonably likely that their reporting was accurate. Cunningham et al⁷ found that self-reported CD4⁺ cell counts were accurate when compared with values recorded in the medical record. To the extent that persons are misclassified by type of testing, this would tend to make the 2 testing groups look more similar. Acknowledging that there is also likely to be some error in the self-reported first CD4⁺ cell count and dates of HIV testing and HIV-related care, we do not have any reasons to suspect that this reporting is biased by the type of HIV test a person received.

Anonymous testing was not available in all the study states in the early years of HIV testing. However, since all the respondents were diagnosed as having AIDS within the same year, there is a bias toward a positive association between confidential testing and the longest intervals between knowledge of being HIV positive and AIDS. When we limited our sample to more recent years in which both anonymous and confidential testing were available, we find a proportionally even greater difference between persons tested anonymously and persons tested confidentially in the length of time between knowledge of being HIV positive and AIDS diagnosis (data not shown).

A question can be raised whether the benefit we observed for anonymous testing is attributable to the availability of this type of testing or to characteristics of persons tested anonymously that make them seek earlier testing and care. For example, among HIV exposure groups, MSM are more likely to seek anonymous testing. From a policy perspective the question is whether the same persons who seek early HIV testing at anonymous sites would do so at confidential sites if anonymous testing sites were eliminated.⁸ We cannot rule out the possibility that these same persons would have sought early testing and care even if anonymous testing were not available. However, we designed our analysis to isolate the independent contribution of type of HIV testing to our outcome measures. To avoid a potentially biased comparison of persons who voluntarily sought testing at either anonymous or confidential testing sites with those who were required to be tested in confidential settings, we limited our analysis to those whose reasons for testing suggested that the action was voluntary. To avoid a bias toward confidential testing among sicker persons who sought medical care, we included symptoms at the time of HIV testing in our adjusted analysis. Of the persons tested anonymously, 50% reported that they were symptomatic, suggesting that even sick persons were making testing choices. We also controlled for a wide variety of other characteristics that differentiated persons tested anonymously and those tested confidentially and still found that anonymous testing was independently associated with a substantially higher first CD4⁺ cell count and a longer period of HIV-related medical care before AIDS.

We explored the possibility that the longer duration of HIV-related medical care for persons tested anonymously could be due to explanations aside from their seeking medical care earlier. For example, if persons tested anonymously were diagnosed as having AIDS more often than persons tested confidentially on the basis of an opportunistic infection as opposed to a CD4⁺ cell count below 0.20 × 10⁹/L, this would create a bias toward lengthening the duration of HIV-related medical care before an AIDS diagnosis for persons tested anonymously. In general, opportunistic infections occur later in the HIV disease course than detection of a CD4⁺ cell count below 0.20 × 10⁹/L. However, since more persons testing confidentially than anonymously were diagnosed as having AIDS on the basis of an opportunistic infection, adjusting for this bias merely increases the duration of HIV-related medical care among persons tested anonymously compared with confidentially. A second explanation for the longer duration of HIV-related medical care for persons tested anonymously is that they were receiving better-quality medical care than were persons tested confidentially. However, comparisons between persons tested anonymously and confidentially in their receipt of several effective prevention and treatment services revealed no significant differences.

Some of the individual characteristics associated with earlier HIV testing and HIV-related medical care were expected, but others were not. For example, we had anticipated that persons who were symptomatic would seek care more quickly than persons who were asymptomatic. We found that after controlling for whether persons had HIV-related symptoms at the time they received a positive HIV test result eliminated the significant difference between persons tested anonymously and persons tested confidentially in the length of their delay between learning they were HIV positive and getting HIV-related medical care. However, we were surprised that neither health insurance nor having a regular source of care—2 traditional measures of access—was associated with early HIV testing or HIV-related medical care. This finding suggests that either physicians are not sufficiently identifying their high-risk patients and encouraging them to be tested early or that patients who have insurance or a regular source of care are reluctant to pursue HIV testing at any greater rate than is found among all at-risk individuals.

We found, as other reports have suggested, that black and Hispanic persons tended to have fewer days of knowing that they were HIV positive before AIDS and fewer HIV-related medical care days than whites⁹ ; however, the comparisons with whites were not significant in the adjusted analyses.

With the development of improved therapies for HIV-infected persons, the rationale for anonymous testing may be waning.¹⁰ In our companion study of persons at high risk for HIV, we found that in the 1990s the annual rate of choosing anonymous rather than confidential testing was 44% to 58% (mean, 48%) (A.B.B., D.O., F.M.H., et al, unpublished data, December 1995-November 1996). This suggests that, at least through 1996, anonymous testing has remained a consistently important testing option for a significant proportion of at-risk persons. It is also possible that more at-risk persons will be interested in anonymous testing now that the Council of State and Territorial Epidemiologists has revised its statement on HIV reporting to favor name reporting¹¹ and a growing number of states and Congress are actively considering the implementation of HIV name-reporting policies.¹² To the extent that name-reporting surveillance systems create a barrier to HIV testing for some persons, anonymous testing might serve as a "safety valve" for those who fear that confidential surveillance systems cannot adequately protect their privacy.

Observational studies may never be able to fully tease apart the contribution that anonymous testing makes to the timing of HIV testing and to HIV-related medical care. In reality, there is a complex interplay among the characteristics of persons at-risk for HIV, changes over time in the perceived benefit of knowing one's serostatus, the availability of anonymous testing, the implementation of name-reporting policies, the opportunity to circumvent surveillance strategies by using a false name at confidential testing sites, and the availability of anonymous home HIV testing kits. We were able to adjust for many, but not all, of these factors. Nonetheless, we believe that our study provides the strongest evidence to date that anonymous testing contributes at a population level to early HIV testing and medical care. Thus, to achieve the public health goal of providing early access to HIV testing and HIV-related medical care, public health departments should maintain and in some instances enhance the broad availability of anonymous testing options.