A Plausible Causal Model of HAART-Efficacy Beliefs, HIV/AIDS Complacency, and HIV-Acquisition Risk Behavior Among Young Men Who Have Sex with Men
- 364 Downloads
Despite considerable research, the causal relationship remains unclear between HIV/AIDS complacency, measured as reduced HIV/AIDS concern because of highly active antiretroviral therapy (HAART), and HIV risk behavior. Understanding the directionality and underpinnings of this relationship is critical for programs that target HIV/AIDS complacency as a means to reduce HIV incidence among men who have sex with men (MSM). This report uses structural equation modeling to evaluate a theory-based, HIV/AIDS complacency model on 1,593 MSM who participated in a venue-based, cross-sectional survey in six U.S. cities, 1998–2000. Demonstrating adequate fit and stability across geographic samples, the model explained 15.0% of the variance in HIV-acquisition behavior among young MSM. Analyses that evaluated alternative models and models stratified by perceived risk for HIV infection suggest that HIV/AIDS complacency increases acquisition behavior by mediating the effects of two underlying HAART-efficacy beliefs. New research is needed to assess model effects on current acquisition risk behavior, and thus help inform prevention programs designed to reduce HIV/AIDS complacency and HIV incidence among young MSM.
KeywordsHIV/AIDS complacency HAART optimism HIV treatment beliefs Structural equation modeling Men who have sex with men
Notwithstanding considerable public-health efforts, the annual HIV incidence among men who have sex with men (MSM) in the United States has increased steadily since the early 1990s . Young black and Hispanic MSM are particularly affected [2, 3]. MSM aged 13–29 years accounted for 38% of the estimated 30,000 new infections among MSM in 2006, and 52 and 43% of the estimated new infections among black and Hispanic MSM, respectively .
To help reduce HIV incidence among MSM and other high-risk persons, the Centers for Disease Control and Prevention (CDC) announced in 2009 a new national Act Against AIDS campaign . The prevention strategy of this $45 million prevention campaign is based on the hypothesis that many high-risk persons believe that HIV/AIDS is no longer a serious health threat because of highly active antiretroviral therapy (HAART), and that those who are less concerned (i.e., more complacent) about the disease are more likely to acquire HIV by engaging in greater risks [5, 6]. In apparent support of this hypothesis, 16 cross-sectional studies found that although a minority of HIV-negative or unknown-status MSM endorsed HAART-related optimistic beliefs and/or reduced HIV/AIDS concern, MSM who endorsed at least one optimistic belief or attitude were more likely to engage in risk behavior [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22].
Because of their cross-sectional designs, however, these studies could not evaluate the directionality of observed associations: whether HAART-optimistic beliefs or attitudes were a determinant or consequent of risk behavior. Two contrasting theories that attempt to explain these associations have emerged from longitudinal studies that found evidence for both directions.
First, motivational health-behavior theories such as the health belief model (HBM) and protection motivation theory (PMT) hypothesize that persons are less motivated to enact behaviors to prevent disease if they perceive less susceptibility to that disease or that the disease is not severe [23, 24, 25]. Thus, under HBM/PMT, risk behavior is determined, in part, by the motivational underpinnings of perceived susceptibility and severity. In support of HBM/PMT, two longitudinal studies conducted in the Netherlands found that MSM who endorsed reduced HIV/AIDS concerns because of HAART were more likely to subsequently engage in sexual risks and to acquire a sexually transmitted disease (STD) including HIV [26, 27].
In serial cross sectional studies conducted in England and Scotland, however, annual increases in the prevalence of risk behavior occurred among both MSM who were and were not less concerned about HIV/AIDS because of HAART [28, 29, 30]. Moreover, in a longitudinal study of MSM in the U.S., reduced HIV/AIDS concern due to HAART did not predict subsequent risk behavior; however, risk behavior predicted subsequently measured reduced HIV/AIDS concern . Based on their findings, Elford and colleagues, Williamson and Hart, and Huebner and colleagues proposed that cognitive dissonance theory (CDT) might better explain observed cross-sectional associations [28, 29, 30, 31].
In contrast to HBM/PMT, CDT posits that persons experience negative emotional states (e.g., stress) upon recognizing that their stated ideas or enacted behaviors contradict internalized beliefs or attitudes (i.e., cognitions are dissonant) . Under CDT, beliefs or attitudes are modified to diminish the stress associated with the behavior that produced the cognitive dissonance . The motivation to modify beliefs or attitudes (e.g., concern about HIV/AIDS) can be powerful if the behavior that produces the cognitive dissonance is perceived as particularly harmful to oneself or others .
Inconsistent findings from the above longitudinal studies, thus, suggest complexity in the causal relationship between the HAART-related attitude, reduced HIV/AIDS concern, and risk behavior. Moreover, the literature yields inconsistent findings on two plausible underlying beliefs of theoretical import. Of 13 studies that measured the belief that HAART reduces susceptibility to HIV, for example, seven found significant associations between this belief and risk behavior [16, 17, 18, 19, 20, 21, 22] and six did not [13, 14, 15, 28, 29, 30]. Of six studies that measured the belief that HAART reduces HIV/AIDS severity, none found that this belief was significantly associated with risk behavior, contrary to theoretical expectations under HBM/PMT [15, 16, 18, 20, 21, 27].
Public Health Significance and Prior Research Limitations
Understanding the complexity of the causal relationship between HAART-related reduced HIV/AIDS concern and risk behavior, and of the relative importance of underlying beliefs, is critical for prevention programs that target HIV/AIDS complacency as a means to reduce HIV incidence among MSM. Two principal limitations of past research have prevented a more complete understanding of this relationship.
First, previous studies evaluated HAART-related beliefs or attitudes as independent predictors of risk without regard to a theoretical framework that maps out direct and indirect (i.e., mediated) effects on risk behavior. Through structural equation modeling (SEM), a HAART-efficacy belief and HIV/AIDS complacency model can be evaluated that might help explain inconsistent findings on both the directionality of direct effects between reduced HIV/AIDS concern and risk behavior, and the relative importance of the indirect effects of underlying beliefs.
Second, none of the prior HAART-optimism studies in the United States evaluated HIV infection as the outcome of interest. Simple behavioral measures do not typically include partner risks for infection, and thus can be poor surrogates of STD risk and in differentiating MSM known to have considerably different HIV-infection risks [33, 34, 35, 36]. Thus, it is important to evaluate whether HIV/AIDS complacency, the target of a new national social marketing campaign, is associated with both increased risk behavior and HIV infection, and to assess whether HBM/PMT, CDT, or both theoretical frameworks might explain observed associations.
To help meet these needs, this report uses data from the second phase of CDC’s Young Men’s Survey (YMS) to evaluate a plausible HAART-efficacy belief and HIV/AIDS complacency model among young MSM, and whether HIV/AIDS complacency, as the hypothesized mediating construct in the model, presumably operates under HBM/PMT to increase both risk behavior and HIV-infection risk. In this paper, “efficacy” is used rather than “optimism” to better reflect our measures on perceived effects of HAART, which are now generally recognized as true [6, 37, 38, 39, 40, 41].
Plausible Causal Model
The effects of the two belief constructs on reduced HIV/AIDS concern are hypothesized to either be completely mediated (mitigate-susceptibility belief) or partially mediated (mitigate-HIV/AIDS belief) by reduced susceptibility concern (Fig. 1). Finally, the model posits that risk behaviors measured since onset of sexual activity (i.e., sexual lifetime) act to reduce concerns about HIV/AIDS (path B) and affect new (i.e., recent) risk behavior. In the model, HAART-efficacy beliefs and reduced susceptibility and HIV/AIDS concern are measured as latent constructs, lifetime and recent risk behaviors are indices of observed variables, and all paths are hypothesized to have positive signs (Fig. 1).
The posited causal relationships of the model are based on principles derived from theories of planned behavior (TPB), health-belief model, protection motivation, and cognitive dissonance [23, 24, 25, 32, 42]. In accordance with TPB, the model posits that beliefs are underlying determinants of behavior, and that the effects of beliefs on behavior are mediated, in part, by attitudes . Mitigate-susceptibility belief, for example, is posited to be completely mediated by its corresponding attitude in accordance with TPB (Fig. 1). Considerable observational and experimental research supports the hypothesis that attitudes can mediate the effects of beliefs on behavior [25, 43].
In accordance with HBM/PMT and CDT, the key complacency construct, reduced HIV/AIDS concern, is hypothesized as both a determinant (path A) and consequent (path B) of risk behavior (Fig. 1). Empirical evidence for this dual role is provided by serial cross-sectional and longitudinal studies mentioned earlier [26, 27, 28, 29, 30, 31]. Finally, lifetime risk behavior is posited to serve as a surrogate for unmeasured extrinsic or intrinsic pleasures or rewards that are posited under HBM/PMT to influence decisions to engage in new risk behavior (Fig. 1) [23, 24].
YMS was a cross-sectional, venue-based survey of young MSM conducted in two phases in select metropolitan areas of the United States. The purpose of YMS was to evaluate the prevalence of HIV infection and associated risk behaviors among diverse samples of young men who attended MSM-identified venues [2, 44]. In Phase I (1994–1998), MSM 15–22 years of age were recruited in seven metropolitan areas. In Phase II (1998–2000), MSM 23–29 years of age were recruited in six of the Phase I cities: Baltimore, Dallas, Los Angeles, Miami, New York, and Seattle. This report uses data from Phase II participants who were asked about their awareness, beliefs, and attitudes about HAART.
YMS methods are described in detail elsewhere . In summary, formative research was conducted in each city to construct monthly sampling frames of the days, times, and venues attended by young MSM. From these sampling frames, 12 or more venues and their associated day/time periods were selected randomly and scheduled as recruitment events each month. During recruitment events, men were approached consecutively to assess their eligibility. Men aged 23–29 years who resided in a locally defined area and who reported having never previously participated in the second phase of YMS were eligible and encouraged to participate. Participants had blood drawn for HIV testing, and were interviewed using a standard questionnaire, provided counseling and referral for care, and reimbursed $50 for their time. Specimens were tested at local laboratories with assays approved by the Food and Drug Administration. The YMS protocol was approved by institutional review boards at CDC, and at state and local institutions that conducted the survey.
Distribution of scores on construct items and risk-behavior indices of 1,593 23–29 year-old MSM who were aware of HAART and who had never HIV tested or last tested HIV-negative, six U.S. cities, 1998–2000
Construct items and indices
Range of meansa
HAART mitigates HIV/AIDS belief
1. If I became infected with HIV today, I probably wouldn’t get AIDS given the combination drug treatments that are available
2. If I got infected with HIV today I could live a long and healthy life by taking the combination drug treatments that are available
3. HIV is now a manageable disease much like diabetes
4. If I became HIV infected today, the combination drug treatments would prevent me from getting AIDS for many years
HAART mitigates HIV susceptibility belief
5. I would be less likely to get infected by an HIV positive partner with undetectable virus than a HIV positive partner with detectable virus
6. If I were having anal sex with an HIV-positive man and his condom broke, it would be less risky for me if he had no detectable virus
7. If my partner had a low viral load it would be less risky for me to have receptive anal sex with him than if he had a high viral load
Reduced susceptibility concern
8. If I had an HIV positive sex partner who was taking the new combination drug treatments for HIV, I would be less worried about getting infected by him
9. If I had an HIV positive sex partner who had a low viral load, I would be less worried about getting infected by him
10. If my partner had a high viral load I would worry about having sex with him (reverse coded)
Reduced HIV/AIDS concern
11. Because of the combination drugs available for HIV, I am less concerned about becoming infected
12. Because of the combination drugs available for HIV, I’m not as concerned about slipping and having unsafe sex
13. With the good news about combination drugs for HIV, I worry less about having sex with partners that might be HIV-positive
14. I’m not as concerned about HIV infection now that there are combination drugs available for HIV
Risk behavior indices
The 14 items were included in a separate section of the questionnaire on knowledge and beliefs about HAART. Responses were measured on a 5-point scale ranging from (1) strongly disagree to (5) strongly agree. The 14 items were administered to participants who (1) reported either having never tested for HIV or having last tested HIV-negative and (2), who responded “yes” to the following question: “Have you heard about the new combination-drug treatments for HIV and AIDS that include protease inhibitors? By combination-drug treatment I mean a protease inhibitor taken with at least one other anti-HIV drug to treat HIV infection.”
To evaluate the association between reduced HIV/AIDS concern and specific sexual risk behaviors and HIV infection, scores on items 11, 12, 13, and 14 (Table 1) were summed into a composite score and dichotomized into two endorsement levels labeled weak and moderate/strong. Weak endorsement included composite scores between 4 and 8 (i.e., representing disagreement with the items). Moderate/strong endorsement included all other responses because very few MSM strongly agreed with the items.
Risk Behavior Indices
We used lifetime (ever) and recent (prior 6 months) behavior indices to evaluate plausible effects on and from reduced HIV/AIDS concern (Fig. 1). Indices were based on factors associated with prevalent and incident HIV infection among MSM (Table 1) [2, 45]. The lifetime risk behavior index was composed of the following four variables weighted in accordance with reported adjusted odds ratios for prevalent HIV infection: number of lifetime male sex partners, ever having anal sex with another male, ever being diagnosed with an STD, and ever having used needles or “works” to inject non-prescription drugs . The recent risk behavior index was composed of the following five variables measured in the prior 6 months weighted in accordance with reported adjusted hazard ratios for incident HIV infection: number of male sex partners; amphetamine use; heavy alcohol use; use of alcohol or drugs before sex; and unprotected anal intercourse (UAI) with HIV-infected/unknown-status male partners . Additional information on the weighting mechanism and predictive validity of both indices is available upon request.
Because YMS was cross-sectional, the temporal relationship between HAART-related reduced HIV/AIDS concern and risk behaviors that composed each index is unknown. For example, because many MSM initiate risk behaviors at a young age , some but not all of the behaviors reflected in the lifetime index occurred before protease inhibitors became available (1996) and before the development of attitudes about HAART. Moreover, because a 6-month recall period was used, all behaviors reflected in the recent index may have occurred either before or after the development of attitudes about HAART. Because these temporal relationships are unknown, evidence for the directionality of effects on and from reduced HIV/AIDS concern can only be evaluated from a theoretical perspective (see “Directionality of Effects” section).
Perceived Risk for HIV Infection
Perceived risk for HIV was measured with the following item: “Using this card, choose a number that best describes how likely it is that you are HIV-positive today.” The card included six possible responses (1, very unlikely; 2, unlikely; 3, somewhat likely; 4, likely; 5, very likely; 6, HIV-positive). In this analysis, all participants except one reported values ≤5 for this risk-perception question. Because the one participant who perceived being “HIV-positive” reported that his last HIV test result was negative, he was also included in this analysis. Responses to the risk-perception question were categorized into two levels (1, labeled “very low perceived HIV risk” vs. 2–6, labeled “some perceived HIV risk”) required for stratified SEM analyses (see “Directionality of Effects” section).
Analyses were first performed to evaluate (1) recruitment outcomes, socio-demographic characteristics, risk behaviors, and item distributions overall and by YMS city; (2) the normality of items and indices used for SEM; and (3) correlations between the 16 items and indices (correlation matrix is available upon request). Severe univariate non-normality was defined as >|3| for skew or >|8| for kurtosis . Multivariate non-normality was defined as having a relative multivariate kurtosis value >|2|. The magnitude of proportions of explained variance was interpreted in accordance with Cohen’s recommendations: <9%, small; 9–25%, moderate; >25%, large [25, 47]. For all analyses, P values <0.05 were considered statistically significant. All univariate data screening, internal consistency, SEM, and contingency table analyses were conducted using FREQ, CORR, and CALIS procedures in SAS version 9.1 (SAS Institute Inc., Cary, NC).
Parameter Estimation and Model Fit Criteria
To estimate SEM parameters, maximum likelihood (ML) estimation was used on covariance matrices. For all models, fit was evaluated using the model χ2 fit statistic, root mean square error of approximation (RMSEA, 90% confidence interval), root mean square residual (RMR), and non-normed (NNFI), incremental (IFI or Bollen’s Delta2), and comparative (CFI) fit indices [46, 48].
Because of the large sample size of this survey and the highly constrained measurement and structural models, the model χ2 fit statistic was expected to be statistically significant, indicating that the model does not fit the data perfectly (i.e., relative to the just-identified model in which all possible paths are modeled) . Thus, although the model χ2 fit statistic is reported for reference purposes, model fit is based on the remaining indices in accordance with the following interpretative criteria: RMSEA <0.05, close approximate fit, 0.05–0.08, reasonable approximate fit, >0.08, poor fit; RMR <0.05, good fit, 0.05–0.09, adequate fit, >0.09, poor fit; NNFI, IFI, CFI: <0.90, inadequate fit; 0.90–0.95, adequate fit, >0.95, good fit .
The original measurement model included the four latent constructs and their respective items (Table 1). Two methods were used to assess reliability of the model and constructs. First, separate confirmatory factor analyses (CFA) were performed to evaluate model fit and parameter estimates across all 20 combinations of sub-samples restricted to three YMS cities [49, 50]. To derive the final measurement model, changes to the original model were considered only when parameter estimates justifying the change (e.g., % variance explained) were consistent across all 20 sub-samples. Sub-samples from three YMS cities were chosen to provide approximately 20 cases for each estimated parameter in accordance with SEM sample-size recommendations . Second, the reliability of specific constructs was assessed with Cronbach’s alpha on items retained for the final measurement model.
Stability and Mediation
The original hybrid model included the final measurement model and the original structural model (Fig. 1). To assess the stability of fit and parameter estimates, SEM was used to evaluate the original hybrid model for the entire sample and for all 20 combinations of sub-samples. To assess mediation of HAART-efficacy beliefs, the original hybrid model was compared against an alternative model that included four direct effects that were constrained to zero in the original hybrid model. Improvement of fit between original and alternative hybrid models was evaluated using the χ2 difference test for nested models .
Directionality of Effects
To assess evidence on directionality of effects, the original hybrid model was evaluated separately for MSM who perceived themselves at very low and at some risk for HIV infection. If CDT alone explains hypothesized associations, the hybrid model should fail (i.e., have poor fit and statistically non-significant path coefficients) when evaluated among MSM who perceive themselves at very low risk for HIV. Model failure is expected because the only presumed motivation underlying the causal relationships (i.e., cognitive dissonance) should be nearly eliminated among MSM who perceive themselves at very low risk for HIV. Alternatively, if HBM/PMT alone explains hypothesized associations, excluding path B (Fig. 1), all other path coefficients should be statistically significant and the model should fit adequately for both risk-perception strata of MSM. Finally, if both CDT and HBM/PMT operate, the hybrid model should fit adequately for both strata; however, path B should be statistically significant only among MSM who perceive at least some risk for HIV.
Reduced HIV/AIDS Concern, Sexual Behaviors, and HIV Infection
Finally, contingency table analyses using chi-squared tests and odds ratios (OR) and 95% confidence intervals (CI) were conducted to evaluate whether the key complacency construct of the hybrid model, reduced HIV/AIDS concern, is associated with testing HIV-positive, and reporting in the last 6 months ≥10 male sex partners and UAI with ≥1 HIV-positive/unknown status male partners. These behavioral outcomes were chosen because they had the highest adjusted hazards for incident HIV infection in a large contemporary cohort of MSM . Contingency tables were evaluated on all data combined and on data stratified by perceived risk for infection and interval in time since last HIV-negative test result (never tested/≥1 year vs. <1 year).
Derivation of Analytic Sample
At 181 venues in the 6 cities, staff enrolled 3,137 (57.6%) men of 5,443 who were identified as eligible. Of the 3,137 participants, the following were removed from analyses: 53 (1.7%) duplicates; 13 (0.4%) who gave contradictory responses or who were impaired by alcohol or drugs; 11 (0.4%) who reported never having sex; 121 (3.9%) who reported never having sex with men; 199 (6.3%) who reported previously testing HIV-positive (n = 104), indeterminate (n = 5), or who either didn’t know their last result (n = 89) or who refused to report their last result (n = 1); and 1055 (33.6%) who reported either being unaware of HAART (n = 1047) or who had missing information on awareness of HAART (n = 8).
Of the remaining 1,685 MSM, 92 (5.5%) were excluded from analyses because they either reported not knowing, had missing responses, or refused to respond to one or more of the 14 items (n = 90, 5.3%), or reported not knowing or refused to respond to the measure on perceived HIV risk (n = 2, 0.1%). Of the 90 MSM, 74 (82.2%) were excluded because they reported not knowing whether they agreed or disagreed with one or more of the 14 items.
The 92 MSM with incomplete responses were not statistically significantly different from the 1,593 MSM with complete responses by age group and race/ethnicity, or on median scores on lifetime and recent risk behavior (data not shown). Analyses were restricted to the 1,593 MSM who reported being aware of HAART, and had either never tested for HIV or had last tested HIV-negative, and on whom analyzable responses were obtained for each of the 14 items and perceived risk for HIV infection.
Recruitment, demographic, and risk characteristics of 1,593 23–29 year-old MSM who were aware of HAART and who had never HIV tested or last tested HIV-negative, six U.S. cities, 1998–2000
No. of venues
Participation rate (%)a
No. of enrolled and analyzed
Some college or tech. school
Part or full time
Last HIV-test result (%)b
Never previously tested
Negative, <1 year ago
Lifetime risks (%)b,c
≥20 male sex partners
Engaged in anal sex
Diagnosed with an STD
Recent risks (%)b,d
≥10 male sex partners
UAI with HIV +/unk status male
Daily alcohol use
Under influence of drugs/alcohol during sex
Very low perceived HIV risk (%)b,e
Normality Assessment and ML Estimation
The relative multivariate kurtosis statistic (1.32) and univariate skew and kurtosis statistics for all items and risk-behavior indices did not meet criteria for severe non-normality for the entire sample (Table 1) and for all 20 sub-samples (data not shown). Maximum likelihood estimation of all original, alternative, and final models using the entire sample and all sub-samples successfully converged without any estimation irregularities such as negative error variances or R2 values >1.
Original and Final Measurement Models
For all 20 sub-samples, the original measurement model demonstrated adequate fit, and with the exception of item 10, HAART-efficacy belief and complacency constructs explained similar moderate to large proportions of observed variance of respective items (data not shown). The amount of variance of item 10 explained by reduced susceptibility concern was very small and ranged from 0.5 to 3.3% for the 20 sub-samples. Because of these concordant findings, the final measurement model excluded item 10; no other changes were made.
Original Hybrid Model
Fit and Stability
For all 20 sub-samples, the original hybrid model demonstrated adequate fit (data not shown); all estimated path coefficients had a positive sign; nearly all path coefficients were similar in magnitude; and similar proportions of variance in HIV/AIDS complacency constructs and recent risk behavior were explained (Data Table, Fig. 3). Unstandardized coefficients for four paths, including the effect of reduced HIV/AIDS concern on recent risk behavior, were statistically significant in all 20 sub-samples. For 3 paths, 2 of the 20 sub-samples produced statistically non-significant path coefficients; no city combination produced >1 statistically non-significant path coefficient (Data Table, Fig. 3).
Although the fit of the original model + C was statistically significantly better than the fit of the original model (χ2 original = 338.7, χ2 original + C = 317.9; χ2-difference = 20.8, df = 1, P < 0.01), a meaningful improvement in the proportion of variance explained in recent risk behavior was not observed (original = 15.0% vs. original + C = 15.1%) (Data Table, Fig. 4). Because (1) paths D, E, and F were statistically non-significant, (2) the original hybrid model predicted essentially the same amount of variance in recent risk behavior as original + C, and (3) because the fit of the original hybrid model was adequate, the original model was retained on grounds of increased parsimony.
Directionality of Effects
Among MSM who perceived themselves at very low HIV risk, the original hybrid model demonstrated adequate fit [χ2 = 176.1 (n = 675, df = 82, P < 0.001), RMR = 0.045, RMSEA = 0.041 (0.033–0.050), NNFI = 0.956, IFI = 0.966, CFI = 0.965]; the effect of reduced HIV/AIDS concern on recent risk behavior (path A) was statistically significant (P < 0.05); the effect of lifetime risk behavior on reduced HIV/AIDS concern (path B) was nearly zero and non-significant; and removing path A resulted in a model that had statistically significantly poorer fit (χ2 original = 176.1, χ2 original − A = 182.5; χ2-difference = 6.4, df = 1, P < 0.05) and a 9.5% (1/10.5) loss in explained variance of recent risk behavior (Data Table, Fig. 4).
Among MSM who perceived themselves at some HIV risk, the original hybrid model demonstrated adequate fit [χ2 = 260.3 (n = 918, df = 82, P < 0.001), RMR = 0.048, RMSEA = 0.049 (0.042–0.055), NNFI = 0.938, IFI = 0.952, CFI = 0.951]; the effect of path A was statistically significant (P < 0.001); the effect of path B, although small, was statistically significant (P < 0.001); and removing path A resulted in a model that had statistically significantly poorer fit (χ2 original = 260.3, χ2 original − A = 281.6; χ2-difference = 21.3, df = 1, P < 0.01) and a 17.1% (2.4/14.0) loss in explained variance of recent risk behavior (Data Table, Fig. 4).
Final Hybrid Models
Relative Effects of HAART-Efficacy Beliefs
Among MSM who perceived themselves at very low HIV risk, the HAART-efficacy beliefs had similar effects on reduced HIV/AIDS concern (total standardized effect of mitigate-susceptibility belief on reduced concern: 0.533 × 0.603 = 0.321; total standardized effect of mitigate-HIV/AIDS belief on reduced concern: (0.166 × 0.603) + 0.200 = 0.300; total standardized effect ratio: 0.321/0.300 = 1.07) (Fig. 5).
Among MSM who perceived themselves at some HIV risk, the effect of mitigate-susceptibility belief on reduced concern was approximately 69% greater than that of mitigate-HIV/AIDS belief (total standardized effect of mitigate-susceptibility belief on reduced concern: 0.677 × 0.622 = 0.421; total standardized effect of mitigate-HIV/AIDS belief on reduced concern: (0.126 × 0.662) + 0.166 = 0.249; total standardized effect ratio: 0.421/0.249 = 1.69) (Fig. 6).
Reduced HIV/AIDS Concern, Sexual Behaviors, and HIV Infection
Associations between moderate/strong endorsement of reduced HIV/AIDS concern because of HAART and risk behavior and testing HIV-positive among 1575 23–29 year-old MSM, overall and by perceived HIV risk and interval since last negative HIV test result, six U.S. cities, 1998–2000
Strength of endorsement
OR (95% CI)
OR (95% CI)
OR (95% CI)
Reduced HIV/AIDS concern
Very low perceived HIV risk
Reduced HIV/AIDS concern
Some perceived HIV risk
Reduced HIV/AIDS concern
Never tested/tested HIV-neg ≥1 year ago
Reduced HIV/AIDS concern
Tested HIV-neg <1 year ago
Reduced HIV/AIDS concern
In a venue-based, cross-sectional survey conducted in six U.S. cities 2–4 years after HAART became available, we evaluated a theory-based, HAART-efficacy belief, HIV/AIDS complacency, and risk behavior model among young MSM who reported having never HIV tested or having last tested HIV-negative. Our model demonstrated adequate fit and stability over multiple geographic samples with different socio-demographic and risk-behavior distributions. Among both MSM who perceived themselves at very low and at some risk for HIV, our model (1) demonstrated that HIV/AIDS complacency, measured as reduced HIV/AIDS concern because of HAART, mediated the effects of beliefs that HAART mitigates HIV/AIDS severity and susceptibility to HIV; and (2) predicted statistically significant variance in risk behavior known to be associated with incident HIV infection among MSM.
Although the amount of variance of recent risk behavior explained by HAART-efficacy belief and complacency constructs was small in our sample of MSM, reduced HIV/AIDS concern had a moderately strong (OR = 2.51) association with testing HIV-positive. Notably, the magnitude of this association was similar among MSM who perceived themselves at very low and at some risk for HIV, and among MSM who had remotely and recently tested HIV-negative.
We found, thus, consistent associations between HAART-related reduced HIV/AIDS concern and both recent risk behavior and presumably recent HIV infection. In concordance with the Health Belief Model and Protection Motivation Theory, our findings support the plausibility that HIV/AIDS complacency, shaped in part by beliefs that HAART reduces susceptibility to and severity of HIV/AIDS, increases HIV-acquisition risk among young MSM.
HIV/AIDS Complacency, Risk Behavior, and HIV-Acquisition Risk
Our findings are consistent with 11 of 13 studies that evaluated the association between similar HAART-related reduced HIV/AIDS concern constructs and risk behavior among MSM who had not previously tested HIV-positive [13, 14, 15, 16, 17, 21, 27, 28, 29, 30, 31]. The two studies that did not find an association between these variables were either conducted in the year following the availability of HAART or included only MSM ≤25 years of age, many of whom had not yet heard of HAART [20, 51]. Our findings are also consistent with the only study of its kind that found reduced HIV/AIDS concern was associated with incident STD/HIV infection among MSM in the Netherlands .
Because HIV-acquisition risk is dependent on having partners who are HIV infected, excluding partner-risk variables from behavioral measures reduces the validity of these measures in predicting HIV infection [34, 35, 36]. Thus, because of imperfect measurement, it is not surprising that our reduced HIV/AIDS concern measure had a moderately strong association with testing HIV-positive even though it predicted a small proportion of variance in our recent risk index, which excluded partner risks. Future research on HAART-related HIV/AIDS complacency should consider incorporating partner-risk variables in indices used as behavioral outcomes.
Dual Role of HIV/AIDS Complacency
Our findings also suggest that among MSM who perceive themselves at some HIV risk, HAART-related HIV/AIDS complacency may act as both a consequent and determinant of risk behavior. This finding is not unexpected given considerable empirical support for contrasting HBM/PMT and CDT theories, and may thus help to explain the inconsistent findings on the presumed causal pathway between reduced HIV/AIDS concern and risk behavior [27, 28, 29, 30, 31].
We are not aware of any theoretical rationale that would preclude both HBM/PMT and CDT pathways from operating within individuals over time. While our findings suggest that a reciprocal relationship may exist between HIV/AIDS complacency and heightened risk behavior among MSM, we chose not to include reciprocal pathways in our model because (1) we had no basis on which to assume the reciprocal relationship has reached equilibrium, a necessary assumption with cross-sectional data , and (2) based on factors known to be associated with prevalent and incident HIV infection, we were able to construct two behavioral indices reflecting different time periods from which unidirectional paths could be evaluated.
Mitigate HIV Susceptibility Belief
Our findings suggest that the relative influence of HAART-efficacy beliefs on HIV/AIDS complacency depends on perceived risk for HIV. Among MSM who perceived themselves at very low risk, the effect of mitigate-susceptibility belief attenuated, and both beliefs had similar effects on reduced HIV/AIDS concern and, thus, on recent risk behavior. This attenuation is reasonable among MSM who perceive themselves at very low risk for HIV.
Our finding on the significant indirect effect of HAART-related mitigate-susceptibility belief on risk behavior is supported by seven of 13 studies that found significant direct effects between similar constructs and risk behavior [16, 17, 18, 19, 20, 21, 22]. None of these studies assessed mediation via reduced HIV/AIDS concern, and thus it is unknown whether these observed direct effects would attenuate under mediation analyses.
Our model may also help to explain some of the inconsistent findings. Of the six remaining studies, four reported statistically significant univariate associations between risk behavior and similar mitigate-susceptibility belief constructs, but statistically non-significant associations after adjustment for reduced HIV/AIDS concern [14, 28, 29, 30]. Given our findings, it is reasonable to expect that the association between mitigate-susceptibility belief (exposure) and risk behavior (outcome) will attenuate when evaluated in the presence of its mediator (reduced HIV/AIDS concern) [52, 53].
Mitigate HIV/AIDS Belief
Our findings stand in contrast to six studies that did not observe an association between the belief that HAART mitigates HIV/AIDS severity and risk behavior [15, 16, 18, 20, 21, 27]. Two reasons might account for these contrasting findings. First, we found that the effect of mitigate-HIV/AIDS belief on recent risk behavior was strongly mediated by reduced HIV/AIDS concern. None of the six studies assessed whether similarly measured mitigate-HIV/AIDS beliefs were associated with reduced HIV/AIDS concern, and thus could not address plausible indirect effects on risk behavior. The absence of a statistically significant direct effect between an exposure (mitigate-HIV/AIDS belief) and outcome (risk behavior), does not rule out important mediated effects from that exposure on that outcome .
Second, our measurement items required subjects to appraise their personal likelihood of a quality life taking HAART, assuming they had acquired HIV. In contrast, all six studies assessed how subjects perceived HAART in curing or reducing the severity of HIV/AIDS in other persons or without explicit regard to self [15, 16, 18, 20, 21, 27]. Thus, it is possible that our items were able to measure a belief construct more salient to reduced personal concerns about HIV/AIDS and risk behavior.
Implications for Research and Prevention
Consistent with the literature on HAART optimism, we found that only a minority of MSM strongly endorsed HAART-efficacy belief and complacency constructs [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 27, 28, 29, 30, 31]. Thus, the population-attributable risk of HAART-related HIV/AIDS complacency on the incidence of risk behavior and HIV infection among young MSM was probably very low at the time of our survey, as pointed out by Elford and colleagues of MSM in the Netherlands .
However, since the efficacy of HAART has improved substantially since the time of our survey and that estimated annual HIV incidence among MSM has continued to increase [1, 6], our findings underscore the importance in evaluating the current prevalence of HAART-efficacy beliefs and HIV/AIDS complacency among MSM, and their effects on current HIV-acquisition risks. It is plausible that beliefs and attitudes about the efficacy of HAART to mitigate HIV/AIDS severity or HIV susceptibility have strengthened since the time of our survey, as suggested in two reports [21, 22].
Our findings also suggest three recommendations for intervention research and prevention programs that target HAART-related HIV/AIDS complacency among MSM. First, new research is needed to evaluate the content, delivery, and efficacy of messages designed to counter HAART-efficacy beliefs that are now recognized as accurate. At a minimum, these messages should underscore the fact that transmission from HIV-infected persons on HAART occurs, and that HIV/AIDS remains a severely disabling, costly, and fatal disease [39, 40, 41, 55].
Second, among young MSM who perceive themselves at some risk for HIV, our findings suggest that messages that counter beliefs that HAART mitigates HIV susceptibility may be more effective at reducing HIV/AIDS complacency and risk behavior than messages that counter beliefs that HIV/AIDS is no longer a severe disease. Among young MSM who perceive themselves at very low risk, however, the effects of these messages may be equally effective at reducing HIV/AIDS complacency.
Third, our findings suggest that prevention programs that target HIV/AIDS complacency consider strategies that address both HBM/PMT and CDT causal pathways. HBM/PMT-based interventions designed to heighten uncertainty of risk and perceived vulnerability might reduce HIV/AIDS complacency and risk behavior among some MSM; however, these messages may also threaten self-image and induce defensiveness [56, 57]. CDT-based interventions designed to preserve self-image and interventions that incorporate affective outcomes such as anticipated regret have been effective in reducing risk behaviors and should also be considered [56, 57, 58, 59].
The findings in this report are subject to several important limitations. First, since our survey was restricted to 23–29 year-old men who attended MSM-identified venues in six cities, our findings may not generalize to MSM who are younger or older, who reside in other cities, and who do not attend MSM-identified venues. Also, because our survey was conducted 2–4 years after HAART became widely available, the extent to which our findings are relevant among young MSM today is unknown.
Second, because our definition of HAART required participants to recognize the term “protease inhibitors,” our analyses may have excluded some MSM who had formulated beliefs about new combination therapies even though they were unaware of the highly active component of these therapies. We were not surprised that many men reported being unaware of protease-inhibitor-based HAART considering that our survey was conducted of young MSM shortly after protease inhibitors became widely available. Notably, of identified HAART-optimism studies, only two assessed awareness of protease-inhibitor-based HAART [9, 15]. Our finding among 23–29 year old MSM (66.4% aware of protease inhibitors) is consistent with findings from these two surveys composed of younger MSM 15–22 years of age (45% aware of protease inhibitors) and older MSM (mean age 35.1 years; 86% aware of protease inhibitors) [9, 15].
Third, because YMS was cross-sectional, evidence for the directionality of causal pathways could only be inferred from contrasting theoretical expectations. Our findings do not rule out the possibility, for example, that some or all recent risk behaviors occurred before the development of HAART-related complacent attitudes. Thus, our final models and posited causal pathways are only “plausible” and require verification with prospective studies in which temporal relationships are evaluated explicitly.
Fourth, our risk-behavior indices are not presumed to be 100% reliable and valid as assumed by SEM, and thus our path coefficients are subject to bias of unknown direction and magnitude. Additionally, only two items were retained to measure reduced HIV susceptibility concern. Thus, adequate measurement of this construct is questionable and additional items should be developed and validated for this construct. Finally, YMS was not designed to evaluate any one particular behavioral theory and did not assess other potentially important determinants of risk. Additional research is needed to evaluate whether our model might explain meaningful variance in risk behavior and HIV-acquisition risk when evaluated in the presence of other theoretically important determinants of risk.
Despite these limitations, our consistent findings in multiple geographic samples, in accord with considerable theoretical and empirical expectations, suggest that only a few years after the widespread availability of HAART, young MSM who held stronger HAART-related efficacy beliefs and complacency constructs were more likely to engage in risk behavior and acquire HIV. Whether HAART-related HIV/AIDS complacency and risk behavior have a reciprocal relationship remains to be clarified. We hope that our plausible model might spur new research to clarify this relationship and discern effects on current HIV acquisition risks among MSM, and thus help to inform prevention programs designed to reduce those risks.
We are grateful to the young men who volunteered for this research project and to the dedicated staff who contributed to its success. We are especially grateful to the YMS Phase II coordinators: John Hylton and Karen Yen (Baltimore); Santiago Pedraza (Dallas); Denise Fearman-Johnson and Bobby Gatson (Los Angeles); David Forest and Henry Artiguez (Miami); Vincent Guilin (New York); and Tom Perdue (Seattle).
- 3.Centers for Disease Control and Prevention. HIV prevalence, unrecognized infection, and HIV testing among men who have sex with men—five U.S. cities, June 2004–April 2005. MMWR. 2005;54:597–601.Google Scholar
- 4.Centers for Disease Control and Prevention. Subpopulation estimates from the HIV incidence surveillance system—United States, 2006. MMWR. 2008;57:985–9.Google Scholar
- 5.Centers for Disease Control and Prevention. Act against AIDS campaign. http://www.cdc.gov/hiv/aaa/index.htm.
- 23.Becker MH. The health belief model and personal health behavior. Health Educ Monogr. 1974;2:324–508.Google Scholar
- 32.Aronson E. The theory of cognitive dissonance: a current perspective. In: Berkowitz L, editor. Advances in experimental social psychology. San Diego: Academic Press; 1969.Google Scholar
- 34.Bingham TA, Harawa NT, Johnson DF, et al. The effect of partner characteristics on African American men who have sex with men in the young men’s survey, Los Angeles, 1999–2000. AIDS Educ Prev. 2002;15(Supplement A):39–52.Google Scholar
- 40.Sullivan P, Kayitenkore K, Chomba E, et al. Is the reduction of HIV transmission risk while prescribed antiretroviral therapy (ARVT) different for men and women? Results from discordant couples in Rwanda and Zambia (paper WEAC101). Presented at 5th international AIDS society conference on HIV pathogenesis, treatment, and prevention, July 22, 2009, Cape Town.Google Scholar
- 41.Donnell D, Kiarie J, Thomas K, Baeten J, et al. ART and risk of heterosexual HIV-1 transmission in HIV-1 serodiscordant African couples: a multinational prospective study (paper 136). Presented at 17th conference on retroviruses and opportunistic infections, February 19, 2010, San Francisco.Google Scholar
- 42.Ajzen I, Fishbein M. Understanding attitudes and predicting social behavior. Englewood Cliffs: Prentice Hall; 1980.Google Scholar
- 46.Kline RB. Principles and practice of structural equation modeling. New York: Guilford Press; 2005.Google Scholar
- 52.Szklo M, Nieto FJ. Epidemiology: beyond the basics. Sudbury: Jones and Bartlett Publishers; 2007.Google Scholar