Towards an understanding of sexual risk behavior in people living with HIV: a review of social, psychological, and medical findings

Introduction

HIV continues to spread through sexual transmission in the United States and elsewhere [¹]. One source of transmission is unsafe sex among people who know they are infected with HIV. Of the estimated 800 000-900 000 people living with HIV in the United States [²], about two-thirds are aware of their seropositive status [³] and over 70% are sexually active after they learn they are infected [^4-8]. Many engage in safer-sex practices; however, a considerable percentage of seropositive persons (range 10% to 60% depending on the specific sex acts) [^9,10] continue to engage in unprotected sexual behaviors that place others at risk for infection and place themselves at risk for contracting secondary infections (e.g., syphilis, gonorrhea, herpesvirus-6) that may accelerate HIV disease [^11,12]. As more and more people with HIV live longer and healthier lives because of antiretroviral therapy [¹³], an increasing number of sexual transmissions of HIV may stem from those who know they are infected and engage in unprotected sex [^9,10,14]. It is, therefore, exceedingly important to understand the factors promoting risky sex in this population so that behavioral interventions can be designed optimally for HIV-positive persons. Indeed, these interventions may be a highly cost-effective approach to reducing sexual transmission of the disease.

Numerous studies have examined correlates of risky sex in HIV-seropositive persons. The diversity of results makes it difficult to attain an integrated understanding of the findings without a comprehensive review of the literature. Two papers [^9,14] provide qualitative reviews of factors associated with unsafe sex in persons living with HIV/AIDS. Those papers are valuable, but they did not cover the full range of psychosocial or medical constructs that have been investigated, did not include all of the studies that examined a specific construct, and did not provide data on strength of association with risky sex. Additionally, the previous reviews did not examine whether there are common and unique risk-promoting factors for HIV-positive men and women.

In the present paper, we comprehensively review studies that have examined psychological, social, interpersonal, and medical variables as correlates of sexual risk behavior in persons who know they are HIV positive. We review constructs central to many of the prevailing behavior models [e.g., health belief model (HBM) [¹⁵], social-cognitive theory (SCT) [¹⁶], theory of reasoned action (TRA) [¹⁷], theory of planned behavior (TPB) [¹⁸], AIDS risk reduction model (ARRM) [¹⁹], and information-motivation-behavioral skills model (IMB) [^20,21]. These general constructs include HIV knowledge (HBM, IMB, ARRM), perceived barriers to behavior (HBM), perceived risk (HBM, ARRM), perceived efficacy of a preventive behavior (HBM, ARRM), social support (ARRM), communication (ARRM), commitment (ARRM), health beliefs (TRA, TPB), attitudes (TRA, TPB), intentions (TRA, TPB, IMB), perceived social norms (TRA, TPB), perceived behavioral control (TPB), self-efficacy (SCT, IMB, ARRM), and outcome expectancies (SCT). Our intent was not to place these models into competition with each other, because few studies explicitly tested those models or systematically compared variables from the different theories. Rather, we attempted to determine whether individual variables that represented the general constructs received support in the literature. Additional constructs were also examined, such as emotional states, personality variables, coping, attributions about one's HIV infection, interpersonal and partner variables, medical status and treatment, and beliefs about antiretroviral therapy. HIV counseling/testing was not included as a variable in the review because its association with sexual behavior was reviewed in a recent paper [²²]. Effect sizes were calculated that indicated the magnitude of association between individual variables and unprotected sexual behaviors of people living with HIV, and the studies were stratified by gender of participants in an attempt to identify risk-promoting factors for HIV-positive men and women.

Methods

Literature available in AIDSLINE, MedLINE, and PsychINFO from 1980 to June, 2001 was searched using the following keywords: HIV, AIDS, HIV-positive, HIV-infected, sexual behavior, risk behavior, sexual risk-taking, unprotected sex, unsafe sex, and condom use. Additional studies were identified through the bibliographies of the articles located. English language articles published in peer-reviewed journals that contained any of the key words were screened for inclusion. Studies were included if they met all of the following criteria.

1. Contained a measure of a social, psychological, interpersonal, or medical variable as well as a measure of sexual risk behavior in HIV-positive men or women.

2. Measured any of the following sexual behavior variables:

1. any unprotected insertive or receptive anal intercourse, unprotected vaginal intercourse, or unprotected oral sex

2. consistency of condom use

3. number of unprotected sex partners

4. other sexual risk measures that combined two or more components from the previous categories

5. the presence of sexual transmitted infections after having been diagnosed HIV-seropositive. Instances of unprotected sexual behaviors had to have occurred after study participants became aware of their seropositive status. Twelve studies [^6,7,23-32] included unprotected oral sex as a risk behavior; in each case it was part of an overall risk index that included unprotected vaginal, anal, or oral sex. Studies that used an index that combined sexual risk behavior with other risky behaviors such as drug use [³³], needle sharing, exchanging sex for money or drugs, or pregnancy [³⁴] were excluded in order to focus unambiguously on unprotected sex as the primary outcome variable. The term 'unprotected' rather than 'unsafe' was used because some studies did not specify the HIV serostatus of the sex partners, thus clouding the meaning and potential consequences of the behavior.

1. Reported that some type of statistical test (e.g., chi square, t-test, analysis of variance, regression analysis, correlation, discriminant analysis) was used to examine the association between unprotected sex and a psychological, social, interpersonal, or medical variable. Studies conducted with HIV-positive and HIV-negative samples had to provide separate analyses of the seropositives or statistically test an interaction with HIV serostatus.

If more than one article examined similar risk factors and reported findings from the same dataset, only one of the articles was included. For example, Robins et al. [³⁵] was included instead of Robins et al. [³⁶] because the former focused explicitly on unprotected sex acts. Heckman et al. [³⁷] was included instead of Heckman et al. [³⁸] because they reported findings based on a larger sample.

Calculation of effect sizes (r)

Effect sizes indicating the magnitude of association between unprotected sex and correlates are presented as correlation coefficients calculated according to procedures described by Rosenthal [³⁹]. A positive effect size indicates that a specific variable (as described in Table 1) was associated with an increased likelihood of unprotected sex, whereas a negative effect size indicates that a variable was associated with a decreased likelihood of unprotected sex. When a single study examined more than one correlate of sexual risk behavior, an effect size was calculated for each finding.

Table 1

Ninety-three percent of the reviewed articles reported univariate/bivariate findings, and only 39% of all of the studies reported some multivariate analyses. In many cases, it was difficult to interpret the multivariate results across studies because each multivariate model included a different set of predictor variables in the equation. Conversely, the univariate/bivariate analyses provided an opportunity to detect an association unclouded by other variables. Therefore, effect sizes were based on the univariate/bivariate results. In the few cases in which only multivariate results were reported [^4,23,40-42], effect sizes were based on those findings.

Effect sizes were calculated with the DSTAT program [⁴³]. Study results given in frequencies and chi-square were translated into a phi coefficient. Means and standard deviations, Student's t, and F values were translated into point-biserial coefficients. For a few studies [^{4,37,40-42,44-47}], effect sizes were estimated based on the sample sizes of the subgroup comparisons and the reported significance levels (e.g., P < 0.05, P < 0.01; see [³⁹]). Nine papers [^{32, 48-55}] reported that a finding was significant but did not provide sufficient information to estimate an effect size. In those cases, a significance level of P < 0.05 was simply reported. Finally, some papers described a finding as 'not significant' and did not provide information necessary for estimating an effect size. Those results are denoted as NS.

Construct dimensions

The findings are organized into 19 construct dimensions (see Table 1). Each dimension includes one or more specific variables examined in relation to sexual risk behavior. The construct label captures the general essence of the variables within a dimension; however, those variables may differ in specific meaning or definition. Consequently, there was no attempt to aggregate effect sizes within a dimension. Effect sizes or other information about significance are provided for each variable in the category. The findings are presented separately for studies that examined HIV-seropositive men, HIV-seropositive women, and HIV-seropositive men and women combined.

Results and discussion

The review included 61 English language published articles contributing tests of association for 126 psychological, social, interpersonal, and medical variables. The majority of the studies were conducted in the United States; 11 were conducted elsewhere (Thailand [⁵⁶], Canada [^57,58], Europe [^{24,40,49,53,59-61}], Australia [⁶²]). Thirty-seven investigations reported sexual risk behavior among men [primarily men who have sex with men (MSM)]; 13 studies provided analyses on women [mainly heterosexuals or injection drug users (IDU)], and 17 studies pooled men and women in the analyses. These pooled analyses included considerably more men (mostly MSM) than women (mostly heterosexual or IDU). In general, participants were recruited from HIV outpatient clinics, sexually transmitted disease (STD) clinics, local or state health departments, or other community locations. Most of the IDU samples were recruited from drug treatment outreach programs. Only two studies used probability-sampling methods [^45,63]. Some constructs (HIV/AIDS knowledge, self-efficacy) were not examined in samples of women. Further, there were several instances in which specific variables were examined in only one study. In those cases, caution must be used in drawing conclusions about an association or lack of association. The findings are summarized in Table 1. Information is provided on the study reference number, the effect size or other statistical information about the finding(s), sample size, and type of sample. Significant results (P < 0.05) are indicated in bold.

Variables central to prevailing behavior models

The findings support several constructs of the prevailing behavior models. Specifically, unprotected sex was associated with having less knowledge about HIV/AIDS, its transmission, and health risks (TRA, IMB, ARRM, HBM), believing that safer sex decreases sexual pleasure (HBM, TRA), having less intention to engage in safer sex (TRA, TPB), having little commitment to self or others to practice safer sex (ARRM), lack of confidence in one's ability to enact safer sex practices (SCT, IMB, ARRM), perceiving that one has little behavioral control over condom use (TPB), having problems communicating to partners about safer sex (ARRM), and perceiving barriers to condom use (HBM).

Other variables that may be viewed within the confines of these models received little or no support in the literature. There was only limited evidence that unprotected sex was associated with not perceiving a social norm for safer sex (TRA, TPB) or lack of social support from family, friends, and partners (ARRM). Surprisingly, having a negative attitude toward using condoms (TRA, TPB) was not found to be associated with unprotected sex in five studies. This surprising finding suggests that other protective processes (e.g., sexual communication) may overcome these negative attitudes in HIV-seropositive persons. Further, there were no significant effects for outcome expectancies (beliefs about the consequences that follow from specific behaviors; SCT). This construct, however, was examined in only one study [⁶⁴] and it was operationalized in a very specific manner (belief that condoms can be sexy and erotic, belief that negotiating safer sex will gain partner's trust, and belief that disclosing seropositive status to sex partners will increase sexual pleasure).

Additional psychological, interpersonal, and social factors

Our review included many other variables that were not tied directly to these theoretical models. Several partner variables were found to be associated with unprotected sex in HIV-seropositive persons. The evidence was quite clear in 9 out of 11 studies that seropositive men and women are significantly more likely to engage in unprotected sex with partners reported to be seropositive than with uninfected partners. This finding indicates that many seropositive persons attempt to prevent transmission of HIV. The finding also suggests, however, that seropositive persons who have unprotected sex with seroconcordant partners place themselves at risk for contracting secondary infections that may accelerate their HIV disease [^11,12].

Studies indicate that unprotected sex was not more likely to occur with primary than non-primary partners, which is inconsistent with the elevated rates of risky sex with primary partners observed in general MSM populations [^65-73]. It is plausible that HIV-positive persons have unprotected sex with seroconcordant partners regardless of whether partners are primary or non-primary [^62,73]. When the partners are seronegative or of unknown serostatus, unprotected sex might be more likely to occur with non-primary partners than with primary partners. This trend is seen in three of five studies, although the results failed to reach statistical significance [^62-64].

Other studies uncovered risk for HIV transmission with other types of partners. HIV-positive MSM were more likely to have unprotected sex with anonymous than known partners at risk for infection. Having an attractive partner and having a partner who was willing to engage in risky sex also increased the likelihood of unprotected sex in MSM [^50,74]. HIV-positive IDU were more likely to have unprotected sex with their IDU partners than with non-IDU partners [^58,75], and seropositive women who shared needles with their significant others also tended to have unprotected sex with those partners [⁷⁶]. Finally, living with a sex partner was a risk-promoting factor, and a male partner's desire for children, but not the women's desire for children, was associated with unprotected sex in HIV-positive women.

With regard to previous sexual experiences, having a greater number of past sex partners was associated with unprotected sex in seropositive MSM; the association was less consistent in seropositive women or in combined samples. For those groups, having more sexual intercourse episodes was a risk-promoting factor (variable not examined in MSM). Therefore, for both men and women, having more sexual contacts (either in terms of number of partners or number of episodes) is risk promoting.

Interestingly, although several studies indicated that having problems communicating with partners was associated with unprotected sex, there was little evidence that withholding disclosure of one's seropositive status from sex partners was associated with risky sexual behavior. Indeed, as shown in a recent study [³⁰], the prevalence of safer sex among non-disclosers was very similar to the prevalence of safer sex among disclosers. Non-disclosers may fear negative consequences from disclosing (e.g., refusal to have sex, loss of privacy, stigmatization) but still attempt to be safe with those uninformed partners. Moreover, disclosure does not assure that safer sex will prevail, because some partners may engage in risky sexual activity even after being informed of their risk [³⁰].

There was little evidence that emotional states such as depression, anxiety, or emotional distress in seropositive men or women were associated with sexual risk behavior, although there was some evidence that anger was a risk factor. Experiencing anger, especially anger directed toward other persons, may lessen motivation to protect a partner, which may reduce self-monitoring of one's behavior or reduce attention focused on the safety of the partner [^7,77]. Conceptually similar findings were seen for external attributions about one's HIV infection. MSM who attributed their infection to something that another person intentionally did to them or attributed responsibility/blame for their infection to other persons were much more likely than other men to have engaged in unprotected anal intercourse with partners perceived to be HIV negative or of unknown HIV serostatus [⁷]. Further, MSM and heterosexual men who attributed responsibility for protection more to sex partners than to themselves were at increased risk for engaging in unprotected anal or vaginal intercourse [⁷⁸].

In studies of coping strategies, two of four investigations of MSM [^24,64] found that use of avoidance coping (e.g., distracting oneself from thinking about one's infection) was associated with unprotected sexual practices. There was little evidence that men or women who used fewer behavioral, problem-focused, or cognitive-coping strategies were at significant risk for engaging in unprotected sex. Of the many personality variables that were examined, only a few were found to be significant. Not surprisingly, unprotected sex was associated with being impulsive [^64,76] and sexually compulsive [^26,45]; and among women it was related to being less empathetic, less assertive, and more rebellious [⁷⁶].

Only a couple of studies examined societal- or cultural-level variables. Having experienced AIDS-related discrimination or feeling that society stigmatizes people with HIV were not found to be risk factors for unprotected sex in a combined sample of seropositive men and women [³⁸]. Another study examined the association between acculturation and sexual risk behavior. HIV-positive Latino men and women living in the United States who were acculturated to American practices were more likely than their less-acculturated counterparts to have engaged in unprotected sex [⁷⁹].

Medical factors

Five recent studies examined whether taking highly active antiretroviral therapy (HAART) was associated with sexual risk behavior. Scheer et al. [⁵²] found that HIV-positive persons (majority were MSM) taking HAART showed an increase in the risk of developing an STD (gonorrhea, chlamydial infection, syphilis, non-gonococcal urethritis) compared with those who were not taking HAART. Similarly, another longitudinal study found that HIV-seropositive MSM were three times more likely to report unprotected sexual practices with HIV-negative or unknown partners after they started taking protease inhibitors compared with their behavior before initiating that therapy [⁵⁹]. However, this study did not find an association for other partner groups (HIV-positive, regular, occasional partners) or in other participant subsamples (heterosexual men and women). Two other longitudinal studies also did not replicate the effects of receiving HAART on sexual behavior in a sample of MSM [⁵³] or a sample of men and women combined [⁴²]. Further, a cross-sectional study by van der Straten et al. [⁸⁰] showed that heterosexual serodiscordant couples were less likely to engage in unprotected sex if the seropositive person was taking protease inhibitors. Taken as a whole, these findings do not provide compelling evidence that taking HAART is a risk factor for unprotected sex.

Other studies examined viral load, CD4 cell level, and symptomatic status in relation to sexual risk behavior. Again, the findings are very mixed. One study found that having an undetectable viral load was associated with increased unprotected sex in MSM [⁵³], but two other studies did not confirm the effect in MSM or in other samples [^59,80]. Three studies found that a higher CD4 cell count was associated with increase sexual risk behavior [^46,52,53], whereas four other studies did not [^23,28,32,44]. A similarly inconsistent picture emerged for symptomatic status, number of HIV illness symptoms, and length of time since testing seropositive.

A few studies examined whether beliefs about combination therapy were associated with unprotected sexual behavior. Vanable et al. [⁸¹] found that having reduced concerns about engaging in unsafe sex because of the availability of combination therapy was significantly associated with risky sexual behavior in MSM. However, another MSM study [⁶⁰] and a study of heterosexual serodiscordant couples [⁸⁰] did not find support for such an association. Further, the belief that combination therapy makes people with HIV less infectious was not associated with unprotected sex in a sample of MSM [⁶⁰].

In summary, the results of the medical studies suggest that behavioral interventions for HIV-seropositive persons will need to be cast widely and not focused only on those under treatment with HAART, those who have undetectable viral load, those who are asymptomatic or have higher CD4 cell counts, or those who hold a specific type of medical belief.

Common and unique risk-promoting factors for HIV-positive men and women

Relatively few variables were examined in both seropositive men and women, so there was only limited opportunity to identify common and unique risk-promoting factors. Nevertheless, the literature offers some hints. Although most of these variables discussed above were on an individual basis, they are re-listed below to emphasize their common effects in men and women. In both genders, unprotected sex was associated with having HIV-positive (versus uninfected) partners, believing that safer sex decreases sexual pleasure, having less intention to use condoms, less perceived personal control over condom use, anger, impulsiveness, having had a greater number of previous sex partners or intercourse episodes, and acculturation (among Latinos).

It was more difficult to identify gender-specific processes confidently in the current literature. However, one consistent pattern emerged for women: unprotected vaginal intercourse was more likely among those who perceived that they had little personal control over their male partner's use of a condom, those who were less assertive, those whose male partner desired to have children, and those who used contraceptive methods other than condoms. These findings suggest that gender power differences in controlling sexual situations may contribute to unsafe sex in serodiscordant or seroconcordant heterosexual couples [⁸²].

Limitations of reviewed studies and suggestions for future research

A number of methodological aspects of the reviewed studies merit comment. Three quarters of the investigations did not specify the HIV serostatus of sex partners. Consequently, in many cases it was unclear whether HIV-seropositive persons were having unprotected sex knowing that there was a possibility of transmitting HIV. Future investigations of the sexual behavior of HIV-positive persons need to address empirically the distinction between unprotected sex with a partner of unknown or serodiscordant status, which presents risk of HIV transmission, and unprotected sex with seroconcordant partners, which presents risk for secondary infections [^62,74,83].

Most analyses were based on cross-sectional tests of associations and, therefore, the findings cannot specify cause-effect relationships. Indeed, the link between some psychosocial factors and unprotected sexual behavior may be bidirectional. For example, being less confident in one's ability to enact safer sex, perceiving that one has little behavioral control over condom use, having little commitment to self or others to practice safer sex, or feeling angry may be as much a result as a cause of unprotected sex. Future studies are needed to help to clarify direction of causality.

Another methodological issue concerns the statistical models used to analyze the data. The majority of the studies reported only univariate/bivariate analyses. These types of analysis provide an opportunity to detect an association unclouded by other variables; however, in reality, more complicated correlations among variables may alter the effects that a given variable may have on sexual behavior. For example, being sexually compulsive may manifest itself in having a greater number of sex partners. The association between sexual compulsiveness and unprotected sex may diminish appreciably after statistically controlling for number of partners. Further, variables may interact with each other, and their interactions may have stronger associations with risky sex than either of the variables considered alone. For example, having poor sexual communication skills and a sex partner of unknown serostatus may combine to produce very high risk for unsafe sex. To attain a more refined understanding, multivariate models including tests of interaction effects are needed in future studies.

In addition to the methodological issues, the studies as a whole provide limited information on wider sociocultural variables that may be associated with sexual risk behavior. Research is needed to examine interpersonal power imbalances based on gender, age, and ethnicity as well as stigmatization and discrimination stemming from a person's HIV-seropositive status or sexual orientation. More attention is also needed to the effects of racism, sexism, and oppression on risk behavior. Further, given the increasing incidence of HIV infections in ethnic-minority groups [⁸⁴], there is an urgent need to conduct studies that examine these sociocultural as well as psychological and interpersonal variables in minority populations.

Implications for interventions

The findings of this review illustrate the types of process that need to be addressed in behavioral interventions for HIV-seropositive men and women. The challenge is to identify approaches and settings for conducting these interventions. A full range of approaches is needed, including interventions that focus on psychological processes, behavioral skills, and the dynamics of sexual interactions and relationships. Societal-level approaches that focus on changing community attitudes, values, and practices are also important, because individual-level interventions do not occur in a vacuum.

One logical setting in which to conduct prevention interventions for HIV-seropositive men and women is the HIV outpatient clinic. This setting, as well as STD clinics, affords the opportunity to integrate behavioral prevention with routine medical care and address behavior change across time. Health care providers or support staff can deliver brief interventions to patients each time they attend clinic. This approach may be useful for addressing some types of individual-level risk factors (e.g., knowledge, motivation, behavioral intentions). Specifically, it is important that brief interventions educate HIV-seropositive persons about HIV/AIDS and its treatment and provide information that instills accurate knowledge about transmission risks. Informing HIV patients that secondary infections can accelerate their HIV disease may also be necessary. Further, brief interventions are a perfect vehicle for delivering prevention messages that attempt to increase motivation and commitment of participants to protect partners and themselves. Self-protective messages may be especially effective when communicated to HIV-positive patients by their primary care providers. Brief interventions can also include prominent 'cues' (e.g., brochures, posters) in HIV care settings to help to remind HIV-seropositive persons in a non-invasive manner of the importance of safer sex.

Of course, some HIV-positive people may need more intensive client-centered counseling from professionals to address personality (e.g., sexual compulsiveness, impulsiveness), emotional (anger), and attributional (blaming others for one's HIV infection) dynamics underlying sexual risk behavior. The success of client-centered interventions will depend on whether health-care providers, social-service providers, and people at community-based organizations (CBO, e.g., AIDS service organizations) are able to recognize specific needs of HIV-seropositive persons and provide appropriate referrals to services.

In addition to these individual-level approaches, it is also important that interventions focus on the dynamics of relationships [102]. As shown in this review, several interpersonal and partner variables are associated with risky sex in HIV-seropositive persons, and unprotected sex may occur within serodiscordant relationships. Behavioral interventions are needed to help HIV- positive persons to increase their skills in negotiating safer-sex practices, controlling sexual situations, and communicating effectively with sex partners. This focus may be especially useful for women. Small-group sessions designed to enhance these types of skills can be implemented at HIV clinics and CBO. These skill-building trainings can involve couples as well as individuals, be led by peers or professionals, and utilize behavior modeling and modification techniques.

Finally, wider societal-level processes must also be addressed [102]. Attitudes and practices that discriminate against and stereotype people based on a medical diagnosis or sexual orientation, and that reinforce differences in social status and social power based on gender, race/ethnicity, occupation, or location of residence, may hinder an individual's attempt to protect self and others. Accordingly, it is important to encourage collaboration among researchers, policy-makers, and communities to identify these and other ecological factors that may hinder HIV prevention efforts and to form strategies for constructing approaches to encourage supportive attitudes and practices within communities [^10,102]. Combining this type of ecological approach with individual- and relationship-level interventions for HIV-positive persons will provide a strong opportunity for reducing the incidence of HIV infections.

Acknowledgement

The authors would like to thank Cherilyn McMonigle for her contributions on the early planning phases of this review article.