Similarities and Differences by Race/Ethnicity in Changes of HIV Seroprevalence and Related Behaviors Among Drug Injectors in New York City, 1991-1996

Objective: To measure differences and similarities in the prevalence of HIV infection and of related risk and protective behaviors among New York City black, white, and Hispanic drug injectors during a period of decreasing HIV prevalence.

Methods: Drug injectors were interviewed at a drug detoxification clinic and a research storefront in New York City from 1990 to 1996. All subjects had injected drugs within the last six months. Phlebotomy for HIV testing was conducted after pretest counseling. Analysis compares the first half (period) of this recruitment interval with the second half.

Results: HIV seroprevalence declined among each racial/ethnic group. In each period, white drug injectors were significantly less likely to be infected than either blacks or Hispanics. Similar declines were found in separate analyses by gender, length of time since first injection, and by recruitment site. After adjustment for changes in sample composition over time, blacks and Hispanics remained significantly more likely to be infected than whites. Interactions indicate that the decline may be greatest among Hispanics and slowest among blacks. A wide variety of risk behaviors declined in each racial/ethnic group; and syringe exchange use increased in each group. Few respondents reported injecting with members of a different racial group at their last injection event.

Conclusions: HIV prevalence and risk behaviors seem to be falling among each racial/ethnic group of drug injectors. Black and Hispanic injectors continue to be more likely to be infected. Declining prevalence among whites poses some risk of politically based decisions to reduce prevention efforts. Overall, these results show that risk reduction can be successful among all racial/ethnic groups of drug injectors and suggest that continued risk reduction programs may be able to attain further declines in infection rates in each group.

NOTE

We use the term race/ethnicity rather than using either race or ethnicity alone. This is because objectionable connotations may be perceived in using either term alone; these are to some extent linked to the various uses of these terms in different disciplines and in the discourse of political leaders. Further discussion of these points is presented elsewhere ^(1-4). Because we see race/ethnicity as the product of sociohistorical forces, we define racial/ethnic categories as sociohistorical constructs rather than as biologic entities.

INTRODUCTION

There is a wide literature on racial/ethnic differences in HIV and AIDS prevalence ^(5-13). An earlier paper found that, during the long period of overall stable HIV seroprevalence among injecting drug users (IDUs) in New York City, differences in seroprevalence between white, black, and Hispanic IDUs remained stable, with approximately 30% to 35% of white IDUs and approximately 50% of black IDUs and of Hispanic IDUs infected ⁽¹⁴⁾. That paper suggested that these differences were able to be maintained because of relatively weak intergroup linkages among injection networks, that is, of a tendency toward members of a racial/ethnic group not to inject with members of other groups. Racially/ethnically specific network ties have been found to be a factor that may help explain racial/ethnic differences in infection rates in another study in New York City ⁽⁹⁾.

The period of stable seroprevalence seems to have ended. Since the early 1990s, the HIV epidemic among New York City IDUs has been in a phase where HIV seroprevalence, incidence, and risk behaviors declined ^(15-17).

Here, we consider whether there are differences by race/ethnicity in trends in HIV seroprevalence during this declining phase of the New York City epidemic. Such differences could occur as a result of relatively weak intergroup linkages allowing the epidemics in the different racial/ethnic groups to develop in accordance with their own dynamics, that is, as three separate, only weakly interacting, epidemics. Under these conditions it would be possible, for example, for seroprevalence to decline in one group but to remain stable or even to increase in another. In addition to reporting on seroprevalence, we also consider interracial/interethnic differences in risk behaviors, protective behaviors, and in injecting with members of other racial/ethnic groups during this period.

METHODS

Subjects and Specimens

IDUs were interviewed at a drug detoxification clinic and at a research storefront in New York City from January 1990 through November 1996. All study subjects had injected drugs within the previous 6 months. Phlebotomy for HIV testing and other indicators was conducted after pretest counseling.

To avoid methodologically induced trends in observed data, sampling was conducted to create independent annual samples, that is, no respondent was included twice in any given calendar year. Potential participants were eligible to be interviewed in any calendar year for which they remained behaviorally eligible members of the sampling frame. This is similar to the procedures used by Battjes and colleagues and by the CDC Family of Surveys ^(18,19). Thus, at the storefront, study subjects who continued to inject drugs were eligible to be reinterviewed after 1 year had elapsed from a prior interview; similarly, at the detoxification clinic, study subjects who received multiple episodes of treatment were eligible for reinterview if at least 12 months had elapsed.

The drug detoxification clinic is the largest detoxification center in New York City and attracts patients from the entire city. Thus, as elsewhere described ⁽¹⁵⁾, it provides a useful window on the HIV epidemic among drug injectors in New York City. Clinic patients were selected for participation as follows. Staff rotated through general admission wards of the program to identify the most recent entrant aged 18 or older who, at program intake, had reported injection of illicit drugs within the previous 6 months. The study was then described to the potential participant and a signed informed consent obtained. Most detoxification patients who were asked agreed to participate.

Subject recruitment at the research storefront was primarily through word-of-mouth and peer referral. It supplements the detoxification sample by providing a sample of IDUs who are not entering treatment. Drug injection within the preceding 6 months and age of 18 or older were the only eligibility criteria for the analyses reported here. The research storefront, which has been in continuous operation since 1989, is well known among drug users in the community. Staff make referrals for drug users who need services. There were no changes in study subject recruitment procedures for IDUs at the detoxification program or at the storefront during the period of data collection.

At both sites, a structured face-to-face interview was administered by a trained interviewer. HIV pretest counseling was provided before collection of blood samples. It was not possible to obtain blood from 34% of participants, primarily for two organizational reasons. First, to obtain same-day lymphocyte determinations, which were part of the overall study protocol, blood must be collected before 2:00 P.M. on Monday through Friday, although interviews can be conducted throughout the day. Some study subjects who were interviewed after 2 P.M. left the detoxification program before they could be seen by the HIV counselor/phlebotomist. Second, counselor/phlebotomist staff vacancies also sometimes prevented our obtaining blood specimens. Respondents without HIV test results were excluded from this paper. Rates of such exclusion varied only slightly by race/ethnicity (32% for whites, 36% for blacks, 33% for Hispanics, and 30% for others), gender, age, recruitment site (storefront or detoxification program), or years of injection; and there were no significant differences between those with and without test results for any of the behaviors in Table 1.

Table 1

HIV-1 antibody testing was performed at the New York City Department of Health's Bureau of Laboratories. Replicate enzyme linked immunosorbent assays were done for all samples using commercially available assays. Western blot testing was performed on positive or indeterminate samples.

Questionnaire

All study subjects were interviewed using a supplemented version of the WHO Multi-Centre Study of AIDS and Injecting Drug Use questionnaire ⁽²⁰⁾. Study subjects classified themselves by race/ethnicity as white, black, Hispanic, or other; further specifications of other were elicited and coded. In addition to standard questions on injection with used syringes, sex with primary partners, and similar variables, questions were asked about the last injection event. These included questions on the race/ethnicity of other participants in this event. To guard against social desirability effects, a two-factor scale measuring Impression Management and Self-Deception ^(21,22) was added to the questionnaire in 1995. Each subscale contained seven true/false items, some of which were negatively loaded. Socially desirable responding was indicated by a high total score on each subscale. These subscales have been construct validated and were used by Latkin et al. in a study of self-reported HIV risk behaviors among drug injectors ⁽²²⁾; although Latkin suggested that these subscales might have limited utility for drug injectors, they were used here because they were nonetheless the best ones available to test for such social desirability effects.

Analysis

To investigate changes over time within racial/ethnic groups, we dichotomized the period of observation into a first half (a total of 1276 days from January 10, 1990 through June 30, 1993) and a second half with a total of 1240 days from July 1, 1993 through November 22, 1996). Changes in seroprevalence and other variables were analyzed by comparing the earlier with the later period.

Limitations

Certain limitations of this study should be kept in mind. As in all studies of drug injectors, it was impossible to recruit a probability sample of the population. However, the use of both storefront and detoxification samples and the wide geographic catchment area of the detoxification program suggest some degree of representativeness to important populations of drug injectors in New York City. As in all studies using self-report, issues of inaccuracy of recall, self-presentation bias (the effects of which are studied below using social desirability controls), and varying interpretations of questions all can reduce the validity of measures.

It is possible that the sampling procedure in the storefront sample might recruit the highest-risk subjects first, and then those at less risk, as has been found to be a possible problem with targeted sampling ^(23,24), which could produce a tendency to find decreasing risk over time. This problem is limited by the overall design of this paper, however, in that this sampling-induced tendency would not affect recruitment in the detoxification sample. When analyses of overall trends in HIV and in behaviors were run separately for each subsample as well as for the pooled sample of both together, there were few substantial differences (see Results), and these did not suggest that the storefront sample is any more likely to have declining risk behaviors than the detoxification sample.

RESULTS

Sample Composition

Table 2 presents data on the racial/ethnic composition of the sample during each period. There were small changes in sample composition, with the sample proportion of blacks decreasing and the proportion of whites and others increasing. These changes were specific to the storefront sample. The proportion of women fell from 25% to 22%; that of new IDUs rose from 16% to 21%; that in those 40 years old or older increased from 32% to 40%. The extent to which these changes are artifacts of sampling rather than of the proportions of IDUs in the storefront neighborhood and of those using the detoxification clinic cannot be determined.

Table 2

HIV Seroprevalence

HIV seroprevalence declined significantly among whites, blacks, and Hispanics (declining 14% for whites, 10% for blacks, and 17% for Hispanics, respectively;Table 3). Although the numbers of research subjects in the Other classification is too small to enable definitive determination of whether seroprevalence is declining among them, these data certainly are consistent with what would be expected if it is declining.

Table 3

Whites were significantly less likely to be infected than either black or Hispanic IDUs in each period. For whites versus blacks, the odds ratios (OR) were 0.41 (95% confidence interval [CI], 0.31, 0.53) and 0.25 (95% CI, 0.18, 0.35) in the first and second periods, and for whites versus Hispanics, the ORs were 0.35 (95% CI, 0.27, 0.46) and 0.29 (95% CI, 0.21, 0.41) in the two periods.

Trends in seroprevalence by race/ethnicity (for whites, blacks, and Hispanics), examined separately by subcategories of subjects, are summarized here. Among male IDUs, there was evidence of decline in seroprevalence within each of these racial/ethnic groups. Within the cohort of female IDUs, declines were also present, although the decline between periods did not quite attain statistical significance for black women (OR, 0.65; 95% CI, 0.40, 1.06.) Among new IDUs (who first injected <7 years before being interviewed), HIV declined precipitously among whites (from 10% to 1.5%); there was a significant decline among Hispanics (from 23% to 11%) and a trend to decline among the relatively small number (81) of black new IDUs (from 29% to 17%). Among long-term IDUs (who had first injected ≥7 years before being interviewed), HIV declined among all three groups. When looked at by recruitment site, there were significant declines among each group of storefront-recruited participants, and among white and Hispanic detoxification clinic clients. Among black detoxification clinic clients, seroprevalence was relatively stable at 55% during the first period and 51% during the second (OR, 0.84; 95% CI, 0.60, 1.18).

In studying racial/ethnic trends in seroprevalence, it is important to take account of compositional changes that may differ by race/ethnicity, at least where subsets may differ in infection levels or dynamics. In the data reported here, there was a significant increase (from 21% to 33%) in the proportion of white IDUs who were new injectors (who are more likely to be HIV-negative ^[25]), and a small increase in the proportion of blacks who were new IDUs (from 10% to 13%); but the proportion of new IDUs among Hispanics was stable at around 20%. The proportion of males increased among Hispanics from 75% to 79%. The proportion of whites recruited at the detoxification clinic decreased from 45% to 34% and the proportion of blacks recruited there increased from 39% to 47%. The proportion of young injectors (<30 years old) increased for whites from 24% to 34%, but decreased for Hispanics (from 25% to 17%). Among blacks, the proportions were 8% and 6%, respectively.

After adjustment for these compositional changes and for period (Table 4), blacks and Hispanics remained significantly more likely to be infected with HIV than whites. Addition of interaction terms between period and race/ethnicity significantly improved the fit of the model, providing evidence that the decline in seroprevalence differed between racial/ethnic groups. During the early period both blacks (OR, 2.11; 95% CI, 1.59, 2.79) and Hispanics (OR, 3.00; 95% CI, 2.26, 3.99) were more likely to be seropositive than whites. By the later period, the OR for seropositivity for Hispanics had dropped to 2.66 (95% CI, 1.86, 3.79), whereas for blacks it had risen to 2.94 (95% CI, 2.06, 4.18). Thus, between the earlier and later periods, the decline in seroprevalence tended to be less among blacks (OR, 0.72; 95% CI, 0.56, 0.92) than among whites (OR, 0.52; 95% CI, 0.35, 0.75) or Hispanics (OR, 0.46; 95% CI, 0.36, 0.59).

Table 4

In summary, HIV seroprevalence declined among each racial/ethnic group. In each period, white IDUs were less likely to be infected than either blacks or Hispanics. After adjustment for changes in sample composition over time, blacks and Hispanics remained significantly more likely to be infected than whites. The decline in seroprevalence may be greatest among Hispanics and slowest among blacks.

Risk Behaviors and Protective Behaviors

Risk behaviors are declining among white, black, and Hispanic drug injectors (Table 1). The proportions of IDUs who engaged in receptive syringe sharing in the previous 6 months (and also the proportions who did so at their last injection) declined among whites, blacks, and Hispanics. Whites were more likely to have engaged in such sharing in the prior 6 months during both the early and the later period, although receptive sharing at the last injection did not vary by race/ethnicity in either period. Questions about backloading (i.e., syringe-mediated drug sharing) were added to the questionnaire well into the first period; from the data that are available, it appears that backloading decreased among black and perhaps Hispanic detoxification subjects, but remained stable among storefront recruits and among white detoxification recruits. As a result, 26% of whites reported backloading during the prior 6 months in the second period as compared with only 19% of blacks and 15% of Hispanics (p< .001). The proportion of black IDUs who reported unsafe sex with primary partners decreased, as did the proportion of black IDUs who reported unsafe sex with casual partners. As a result, by the second period, rates of unsafe sex were similar among all three racial/ethnic groups. The results in this and the following paragraph remained substantially unchanged under logistic regression control for changes in sample composition (data not shown).

Syringe exchange use in the prior 6 months increased considerably in all groups: by 18% for Hispanics, 22% for blacks, and 33% for whites. By the second period, storefront-recruited whites (74%) were significantly (p< .001) more likely to report syringe exchange use than blacks (56%) or Hispanics (58%). Among study subjects recruited at the detoxification clinic in the second period, syringe exchange use was lower (43%) but not associated with race/ethnicity.

Of course, data on risk and protective behaviors are based on self-reports. Examination of the 825 cases for which measures of impression management and self-deception were available (all of which were from interviews in 1995 and 1996) revealed that Hispanics had a significantly higher score (mean, 3.96) than whites (mean, 3.60), with blacks (mean, 3.80) not statistically significantly different from either of the other groups by the Tukey-Kramer post-hoc test. On the self-deception scale, blacks (mean, 4.02) and whites (mean, 3.82) were significantly higher than Hispanics (mean, 3.17) by the Tukey-Kramer post-hoc test. Logistic regression analyses of these 825 cases were carried out to check for modifying or confounding effects of socially desirable responding on racial/ethnic differences on six self-reported risk and protective behaviors (receptive syringe sharing, backloading, unsafe sex with a primary partner, unsafe sex with a casual partner, syringe exchange use during the last 6 months, and any syringe sharing at the last injection). Results for each of these behaviors showed that the interactions between race and social desirability scores were nonsignificant (all p> .20); and the effect of race/ethnicity on the behavior was not changed by the inclusion of the social desirability measures in the equations. In other words, the relation between race/ethnicity and behavior was not confounded by socially desirable reporting. There was, however, some indication that persons high on impression management were less likely to report unsafe sex with casual partners (OR, 0.84; 95% CI, 0.71, 0.98).

In summary, it appears that risk behaviors are declining, and syringe exchange use increasing, in each group.

Injection Behaviors With Members of Other Racial/Ethnic Categories

Relatively few members of any racial/ethnic group reported injecting with one or more members of another group at their last injection event (Table 5). The extent of such cross-group linkages was similar for whites and blacks at each time period. Hispanics began with this same rate (14%) during the first period, but decreased their injecting with members of other groups to 8% in the second period. Sharing syringes with members of other groups at the last injection was ≤5% in each group during the first period. It fell significantly for whites (from 5% to 2%) and Hispanics (from 5% to 1%), whereas 3% of blacks reported event-specific sharing with someone from another group in each period. These rates reflect the overall data: only 13% reported sharing with anybody during the first period and only 8% during the second period.

Table 5

DISCUSSION

The first major finding of these data is that HIV seroprevalence and risk behaviors are declining within each of the three racial/ethnic groups considered (and possibly among Other drug injectors in New York as well;Table 2). This is clearly a very positive development.

The second major finding is that black and Hispanic IDUs continue to have higher HIV seroprevalence than white IDUs. Black IDUs recruited at the detoxification clinic, in particular, continue to have a very high HIV seroprevalence rate, with 51% of 245 being infected during the second period (which is statistically indistinguishable from the 55% of 289 infected during the first period). Thus, there is a need for expanded research and prevention efforts aimed at black detoxification clinic attenders and their recruitment base outside of the clinic.

Based on our analysis of a subset of cases, furthermore, there is no reason to believe that the effect of social desirability on self-reports about behavior differs between racial/ethnic groups, although some risk behaviors may nonetheless be underreported.

It would be useful to know whether similar patterns- decline in HIV prevalence within each racial/ethnic group, but continuing higher seroprevalence among Hispanics and blacks-holds in other populations (e.g., Asbury Park, NJ, methadone maintenance entrants) where the epidemic may be declining ⁽¹⁸⁾.

The data on cross-group mixing patterns, which suggest that approximately 8% to 14% injected with someone of a different racial/ethnic group at last injection, are inherently ambiguous: is the glass half empty or half full? This level of cross-group injecting seems to be low enough not to disturb the long-term pattern that white IDUs in New York have lower seroprevalence rates than black or Hispanic IDUs, at least in the city's mature and then declining stages. Data from other cities, and particularly from cities in which HIV prevalence is rising among IDUs, might help define the relation between such mixing patterns and patterns of viral spread.

In interpreting the data on racial/ethnic trends and differences in risk behavior and seroprevalence, there is little reason to expect parallel trends in risk behaviors and prevalence. This is because, although seroprevalence is a function of current risk behaviors together with infection rates among the partners in high-risk behaviors ^(26,27), seroprevalence is also a function of prior rates of infection, deaths among the infected and uninfected, rates of recruitment of new persons into drug injection and of departure of IDUs from sampling frames, and much else.

Although the declines in HIV prevalence have been sizable in all three racial/ethnic groups, the declines among whites have been particularly marked. By the second period, only 14% of the white IDUs being interviewed for this study were infected. Among white new IDUs, seroprevalence was only 1.5%. This suggests the possibility that the epidemic among white IDUs in New York City might be reaching a point where new infections would become very rare and where seroprevalence might decline to ≤5%. This would clearly be a very positive development. Nevertheless, even this good news could pose difficult problems. First, it should be clear that this apparent near victory over the epidemic among white IDUs is extremely unstable. White IDUs engage in the highest rates of risk behavior, so renewed epidemic spread among them remains a strong possibility. Furthermore, they do engage in a degree of receptive syringe sharing and backloading with blacks and Hispanics, so there remain potential routes for infection to reach them even if their own seroprevalence were to shrink to zero. Second, even the degree of risk reduction white IDUs have attained may be highly dependent (perhaps even more so than that of black and Hispanic injectors) on their continued access to syringe exchange and other prevention services. Should political or other changes decrease this access, the epidemic among white IDUs could recur. Third, an apparent victory over HIV among white IDUs implies the possibility that politicians and community leaders may view minority drug injectors as being the only ones still at risk. This could lead to reductions in prevention resources and to greatly increased stigmatization of minority drug users.

Another possible situation is also disturbing: that funding for prevention programs might be sufficient to reduce the HIV epidemic among white IDUs to very low levels, whereas seroprevalence among black and/or Hispanic drug injectors remains at seroprevalence levels of ≥30%. This, of course, would also imply continuing sexual and perinatal transmission to noninjectors, most of who will be black or Hispanic. Although this might occur despite vigorous public health and medical efforts, it would more likely be the consequence of limited and underfunded provision of prevention and medical services. Such a pattern would be quite consistent with past American responses to diseases among blacks and Hispanics ^(2,28-31).

These hypotheses should not cause us to lose sight of the current picture. HIV seroprevalence and risk behaviors have been decreasing among black, Hispanic, and white IDUs and seem to have been decreasing among other drug injectors as well. This shows that risk reduction is possible and strongly suggests that, with appropriate prevention efforts, further declines in seroprevalence can be attained in each racial/ethnic group.

Acknowledgements:

We would like to acknowledge support by National Institutes on Drug Abuse grant DA03574 and Cooperative Agreement ##U64/CCU209685 from the Centers for Disease Control and Prevention. Many colleagues, including Michael Marmor of the New York University School of Medicine, provided useful suggestions during the analysis and interpretation of these data. Other National Development and Research Institutes research staff, including Martin Blasco, Eldon Garcia, Carole Johnson, Ivette Moloney, and Leonard Wright, and the HIV Counseling and Testing Team of the Beth Israel Chemical Dependency Institute, assisted in gathering and processing data. We would especially like to acknowledge the assistance provided by the thousands of drug-injecting participants who answered questions and provided blood for analysis.