Unprotected sex with an HIV-positive individual is the primary vector of HIV infection in the United States.1 And since the advent of highly active antiretroviral therapy (HAART) in 1996, there have been growing concerns about unprotected sex—particularly among HIV-positive individuals receiving HAART who may have unprotected sex because they feel healthier or because they erroneously believe that they are no longer infectious because of reduced viral loads.2 The prevalence of unprotected sex among HIV-positive persons has been increasing since 1996;3 meanwhile, the incidence of STDs,4 including HIV,5 has also increased, especially among men who have sex with men.

Studies have examined how safer-sex practices among HIV-positive individuals have changed since the advent of HAART. According to a meta-analysis of work done on the association between HAART and unprotected sex at the individual level, 21 of 25 studies examined suggest that HAART treatment has little effect on the decision to engage in unprotected sex.6 Moreover, of the seven that found significant associations, four indicate that HAART is associated with a reduced propensity to engage in unprotected sex. Taken together, these findings suggest that the social context after the advent of HAART, not the characteristics of individuals receiving HAART, is driving the rise in unprotected sex and the increased spread of STDs, especially among men who have sex with men.

Several psychosocial factors have repeatedly been linked to unprotected sex among HIV-positive persons. Among men who have sex with men, using avoidance as a coping style has been associated with increased rates of unprotected anal intercourse.7 Blaming others for acquiring HIV has consistently been associated with increased risk of unprotected sex.8 Similarly, unprotected anal sex has been associated with an increased level of sexual arousal, having sex under the influence of drugs, having difficulty in discussing safer sex with partners, believing that safer-sex practices reduce pleasure, having a greater number of male partners and experiencing emotional distress;9 however, these relationships have not been consistently observed.10

Although individuals' coping styles and emotional states influence their propensity to engage in unprotected sex, sex is a social process that occurs within a social decision-making context. The nature of the sexual partnership is an important determinant of unprotected sex among HIV- positive persons. Seropositive men and women are significantly more likely to engage in unprotected sex with partners reported to be HIV-positive than with uninfected partners.11 Moreover, among seropositive men who have sex with men, having a greater number of partners is associated with having more unprotected sex.12

HIV-positive men who have sex with men appear to protect their primary sexual partners more than those who are HIV-negative do: HIV-negative men who have sex with men report more unprotected sex with primary partners than with other partners;13 seropositive men do not.14 Among HIV-positive individuals, having sex without disclosing one's serostatus is more common for men who have sex with men than for other men or for women.15 Moreover, those who disclose their serostatus do not have significantly higher rates of safer sex than those who do not disclose their serostatus, as might be expected.16

None of the studies examining how safer-sex practices among HIV-positive individuals have changed since the advent of HAART has focused on youth. Young people are of particular importance because an estimated 13% of reported cases of HIV in the United States between 1996 and 1999 occurred among 13–24-year-olds,17 and 50% of all new HIV infections worldwide are among people aged 15–24.18 This study aims to compare the factors associated with unprotected sex among young people living with HIV before and after the advent of HAART.

In keeping with previous studies of adult populations, we examine the association between unprotected sex and mental health, partner's serostatus, serostatus disclosure, number of sex acts and type of partner (primary or casual). We examine these variables in two samples of youth living with HIV—one recruited prior to the advent of HAART and the other after. In addition, because individual-level associations between HAART and unprotected sex are of increasing interest, we examine characteristics associated with unprotected sex among youth living with HIV in the post-HAART era, comparing those individuals who were receiving HAART with those who were not.


Study Design

Baseline samples from two similar longitudinal studies of interventions that targeted the sexual behaviors of HIV- positive youth provided the data for our study.19 The first sample (pre-HAART) consisted of 351 youth aged 13–24, recruited between January 1994 and August 1996 at adolescent clinical care and social service sites in four locations—Los Angeles, New York, San Francisco and Miami.20 The second sample (post-HAART) consisted of 253 youth aged 13–29, recruited between May 1999 and December 2000. Recruitment occurred primarily at similar sites and in the same cities (except Miami) as those used for the pre-HAART study.21 To make the two samples more comparable, the age of participants was restricted to include only those aged 24 or younger; therefore, 87 participants in the post-HAART sample and four participants in the pre-HAART sample were removed.

In both studies, participants were eligible if they provided informed consent and were determined at baseline to be sufficiently healthy to remain in the study. Parental consent was obtained for any participant younger than 18 who was not an emancipated minor if the parent was aware of the youth's HIV status. Each participant was paid $20–25 to complete a baseline interview.

Participants in both studies completed similar interviews that covered a variety of demographic, social, medical and behavioral topics. Both instruments were primarily interviewer-administered to participants on a computer, although the post-HAART study contained a section using audio computer-assisted self-interview; pre-HAART interviews typically lasted three hours, whereas post-HAART interviews averaged approximately two hours. Interviewers were chosen to reflect the ethnic diversity of the participants and were trained to administer the assessment on-site, using laptop computers. Interviewer training covered research ethics, emergency procedures for suicidal and homicidal ideations, psychosocial and substance abuse assessments, procedures for reporting sexual abuse, intensive review of assessment protocols and mock interviews.

The assessments measured lifetime and current behaviors, defined as activities in the previous three months. Several strategies were employed to enhance overall recall. For example, participants were probed in-depth about life events, such as job or school changes, celebrations, relationships and changes of residence. Using these events as anchors, we generated a visual time line of the past 12 months for each participant and made frequent references to it during the interviews to assist them in accurately answering questions. Interviewers were trained to aid participants in summing behaviors over the different time frames and to confirm the accuracy of reporting using consistency checks and frequent probing.

The most significant difference between the two interventions was that the pre-HAART intervention occurred in a group setting, whereas in the post-HAART intervention, participants worked one-on-one with a trained HIV prevention facilitator. Also, the second intervention limited its enrollment to youth who were recent substance users at the time of their baseline interview; however, the baseline data included youth who were not recent substance users. These differences had no impact on the current analysis, as enrollment criteria for the two baseline samples analyzed were the same, except for the age range, which we have truncated in the second study to facilitate cross-study comparisons. An analysis not shown here indicated that the inclusion of drug use as a predictor had no effect on the models that were tested.


Independent variables. Measures related to participants' and their sexual partners' demographic characteristics, participants' health status and participants' sexual behavior, were considered as independent variables. Demographic variables included age, gender, research site, educational background, and race and ethnicity. Being a man who has sex with men was based on whether male participants reported ever having had a male sexual partner.

Physical health status was measured by participants' self-report of their most recent CD4 count and by the time elapsed since their reported date of HIV diagnosis. The Brief Symptom Inventory (BSI)—a composite index of 53 items, each of which is measured on a five-point Likert scale ranging from 0 (not at all) to 4 (extremely)—was used as a measure of mental health; an increased BSI score signifies greater emotional distress. Current antiretroviral use was assessed for the post-HAART study with a yes-or-no question.

For sexual behavior, we used participants' reported number of sex partners over the past three months. In their interviews, participants in the post-HAART sample were asked to give a detailed sexual history of up to 15 recent partners; however, those in the pre-HAART group were asked about only five. Thus, we did not consider data for the sixth and subsequent partners of participants in the post-HAART sample. Participants reported the age, gender, perceived serostatus and relationship type (casual or serious) of each partner.

Also, participants reported whether they had disclosed their serostatus to each partner and answered questions regarding the number of sex acts and condom use with each partner. Partner serostatus was asked as a multiple-choice item for the pre-HAART group, but as a yes-or-no question for the post-HAART group. This difference in question wording and the difference in number of reported partners dictated that we run separate regression analyses for the two samples.

Dependent variable. The outcome variable—unprotected sex in the three months prior to baseline—was based on youths' responses to four questions regarding sexual behavior with each reported partner: the number of times the participant had vaginal sex with the partner, the number of times the participant or partner used a condom for vaginal sex, the number of times the participant had anal sex with the partner and the number of times the participant or partner used a condom for anal sex. If a participant's reported number of times using a condom with a partner was less than the reported number of sex acts (for vaginal sex, anal sex or both) with that partner, we deemed the participant to have had unprotected sex with that partner. Those who did not report any vaginal or anal sex in the three months prior to baseline were excluded from the analysis.


The data were structured hierarchically, with one record produced for each of the sex partners detailed by participants, allowing for multiple records for each participant. The first level of the hierarchy consisted of participant-based variables, and the second level consisted of partner-based variables.

The analysis included data for 501 partners of 224 participants in the pre-HAART sample and 175 partners of 104 participants in the post-HAART sample. Seventy pre-HAART and 45 post-HAART participants had been abstinent in the three months prior to the interview and were consequently excluded from the analysis.

We used SAS version 8.1, primarily the GENMOD procedure, for the analysis. An informal selection procedure, based on choosing the most highly predictive variables, was used to create a parsimonious model. Participants with missing data (51 from the pre-HAART sample and 17 from the post-HAART sample) were excluded.

To identify the covariates that were predictive of unprotected sex, we used a generalized estimating equations procedure, because unprotected sex was treated as a binary outcome variable and the hierarchical nature of the data set precluded a simpler generalized linear model analysis. The two levels represented by the hierarchical linear model were the participants (level one) and the partners of each participant (level two). A follow-up analysis involving only the post-HAART sample was conducted to investigate differences in the predictors of unprotected sex between those participants who were receiving HAART treatment and those who were not.

Initial models included all of the potential predictors in the two data sets and selected two-way interaction terms for these variables. All of the interaction terms and many of the main effects failed to reach significance (alpha=.05). To create a more parsimonious model, we eliminated the variables that failed to meet this criterion for both data sets. The following variables were retained: being a man who has sex with men, log of the time since participant's HIV diagnosis, overall BSI score, partner's serostatus, relationship status, number of sex acts with a particular partner and disclosure of HIV serostatus to a particular partner. Current antiretroviral use was included only for the post-HAART sample. We tested the two final models for outliers using residual plots. The plots showed very few outliers, and removal of outliers did not result in a different outcome for the generalized estimating equation analysis.


Seventy-seven percent of the youth in the pre-HAART sample and 71% of the post-HAART sample were male (Table 1); of the males in the two samples, 76% and 68%, respectively, reported ever having had sex with a man. The mean age of each sample was about 21 years, and about half of each sample had had at least a high school education. The greatest proportion of the pre-HAART sample were from Los Angeles, whereas the greatest proportion of the post-HAART sample were from New York; more than one-third of each sample were Latino. The pre- and post-HAART samples differed in terms of mean amount of time since HIV diagnosis (2.1 years and 3.1 years, respectively), perhaps because of the growing epidemic in the youth population at the time of the pre-HAART study.

The proportion of youth in the post-HAART sample who had had unprotected sex with any partner in the past three months was more than twice that in the pre-HAART sample (62% vs. 25%). On average, youth in the post-HAART sample reported smaller proportions of casual partners (28% vs. 46%), HIV-negative partners (54% vs. 70%) and partners to whom they had disclosed their serostatus (42% vs. 53%).

Variables Associated with Unprotected Sex

Among the pre-HAART sample, being a man who has sex with men and being in a casual relationship were negatively associated with the odds of having unprotected sex (odds ratios, 0.5 and 0.2, respectively—Table 2). Among the post-HAART sample, the odds increased with BSI score and were negatively associated with having an HIV-negative partner (1.02 and 0.2, respectively); the participant's number of sex acts with and disclosure of serostatus to a partner were marginally associated with increased odds of unprotected sex. The time since HIV diagnosis was not significantly associated with unprotected sex in either sample, and HAART use was not significantly associated with unprotected sex in the post-HAART sample.

For our separate analyses of the interactions between HAART use and other predictors of unprotected sex, we divided the post-HAART sample into two subgroups: the 54 participants who were using HAART (who had 76 partners) and the 50 who were not (who had 99 partners). Among both groups, having an HIV-negative partner was associated with significantly reduced odds of unprotected sex (0.2 and 0.3, respectively—Table 3). In addition, BSI score was positively associated with the odds of unprotected sex among nonusers (1.02).


Several important findings emerged from this analysis. First, our data suggest that the characteristics associated with unprotected sex among youth living with HIV have changed since the advent of HAART. In the pre-HAART sample, sexual behavior and relationship status were significantly associated with unprotected sex; however, these factors were nonsignificant in the post-HAART sample. By contrast, mental health and partner's serostatus were significant in the post-HAART sample, but not in the pre-HAART sample. This suggests that mental health has become an increasingly salient comorbidity for HIV/AIDS, even in the face of improving medical treatments.

Second, counter to findings by Crepaz and Marks,22 but consistent with results of other studies,23 we found that poorer mental health among youth living with HIV in the post-HAART era who were not receiving the treatment was associated with an increased likelihood of unprotected sex. As part of Advancing HIV Prevention: New Strategies for a Changing Epidemic, the Centers for Disease Control and Prevention (CDC) plans to conduct demonstration proj- ects through health departments to provide prevention case management and counseling for people living with HIV.24 Our findings suggest that for young people living with HIV, such services aimed at improving overall mental health, coping and well-being could be an effective vehicle for prevention interventions.

Third, partner-level characteristics may contribute to unprotected sex among youth living with HIV. In the post-HAART sample, youth living with HIV were less likely to have unprotected sex with a partner they knew to be HIV-negative than with a partner who was HIV-positive; this finding is consistent with results of previous research.25 Also consistent with earlier findings,26 the number of sex acts was associated—albeit marginally—with an increase in the likelihood of having unprotected sex among youth living with HIV in the post-HAART sample, but not among those in the pre-HAART sample. Whether a partner was casual or primary was not significantly associated with unprotected sex among the post-HAART sample, as has been reported previously.27 The absence of a significant association between unprotected sex and casual partner status, coupled with the dramatic increase in the percentage of HIV-positive youth who are having unprotected sex, suggests that young people living with HIV in the post-HAART era are no longer systematically engaging in safer-sex practices with partners outside of their primary relationships, as had been the practice in the pre-HAART period. This is a troubling development because it presents added opportunities for the spread of the virus into the more expansive sexual networks of casual partners.

Finally, youth living with HIV in the post-HAART era who disclose their serostatus to their partners may be more likely than others to have unprotected sex, a finding that is counter to those among adults.28 Although this finding was only marginally significant (likely because of the relatively small sample size), the implications are troubling and point to a need for further research. Youth appear to be acting under the assumption that disclosing one's HIV- positive status places the burden of transmission and protection onto one's partner. Other studies have found that seropositive persons who believe that the responsibility of safer sex lies primarily with their partners are less likely than others to engage in safer-sex practices.29 Perceptions of personal responsibility for protecting sex partners from contracting HIV are likely to be an important factor in sexual decision making for HIV-positive youth.30

There are some limitations to the present analysis. We were unable to compare changes in the effect size of covariates of unprotected sex behaviors across the two time periods. Also, because our analysis used data from two studies and the questions asked in each study differed slightly, we were unable to combine the samples for analysis. Thus, we were limited to comparing what characteristics were associated with unprotected sex at each time period, but we cannot assess how the strength of these associations may have changed from the pre- to the post-HAART period.


One of the primary plans for advancing HIV prevention currently advocated by the CDC is to work with infected people and their partners; the present findings lend weight to this strategy. Among youth living with HIV in the post-HAART era who are not taking HAART, those with poorer mental health had an elevated likelihood of engaging in risky sex. This finding is especially alarming because those not taking HAART are also the most infectious. Future interventions that focus on improving overall mental health, coping and well-being could be an important tool in addressing the complex needs of youth living with both HIV and poor mental health.