Demographic and Clinical Characteristics
Demographic and clinical characteristics were compared between younger and older participants using one-way analysis of variance and chi-square statistics. No differences were found between groups with respect to gender, years of education, ethnicity, sexual orientation, relationship status, drug or alcohol abuse, length of HIV illness, or number of years on HIV medications. However, membership in the older cohort was associated with older age, F(1, 183) = 211.02, p < .01, a higher income, χ2(1, N = 184) = 4.65, p = .03, and increased likelihood of cognitive impairment, χ2(1, N = 185) = 3.18, p = .04. Broken down by cohort, mean age was 56.3 (SD = 4.8) years among individuals in the older group and 41.0 (5.0) years among individuals in the younger group. With respect to income, 66.7% of older participants reported annual earnings of over $10,000 compared with only 48.2% of younger participants. Finally, while 77.5% of elders demonstrated global cognitive impairment on neuropsychological testing, the proportion of younger subjects with similar deficits was found to be 68.9%.
The mean adherence rate across all 185 participants was 75.8%, with older participants achieving a mean adherence rate of 84.9% and younger participants achieving a mean adherence rate of 72.8%, a difference that is statistically significant, F(1, 183) = 11.78, p = .001. As can be seen in , the percentage of participants classified as poor adherers was twice as high among younger participants than older participants (68% and 33%, respectively), a difference that is again statistically significant, χ2(1, N = 185) = 16.84, p < .001.
Highly active antiretroviral therapy adherence as a function of age.
Demographic Variables and Psychiatric Status
A series of univariate analyses were then conducted to examine the impact of demographic and psychosocial variables and psychiatric status on medication adherence in both younger and older cohorts. Findings revealed that poor adherence among younger HIV-seropositive adults was associated with current drug abuse/dependence, χ2(1, N = 140) = 5.65, p = .02, and lack of independent financial resources, χ2(1, N = 140) = 3.69, p = .05. More specifically, of those younger adults meeting DSM–IV diagnostic criteria for drug abuse/dependence, only 6.3% were classified as good adherers compared with 35.8% of participants without a drug abuse/dependence diagnosis. Younger participants who were financially supported by others (e.g., family, friends, government programs) were also less likely to be adherent, with only 28.2% classified as good adherers compared with 46.7% of participants with independent financial resources. With respect to psychiatric symptomatology, higher levels of apathy, t(138) −1.92, p = .05, were found to be associated with substandard medication adherence in this cohort.
In contrast, poor adherence among older participants was found to be associated with both income level, χ2(1, N = 45) = 4.05, p = .04, and sexual orientation, χ2(1, N = 45) = 4.64, p = .03. More specifically, older adults with an annual income of at least $10,000 per year were more likely to be classified as good adherers than were those with smaller incomes (76.7% and 46.7%, respectively). In addition, elders who identified themselves as gay or bisexual were also more likely to be classified as good adherers when compared with their heterosexual counterparts (82.4% and 47.1%, respectively). Psychiatric variables (i.e., anxiety, depression, irritability, apathy) were not predictive of medication adherence among older adults.
Health Beliefs, Locus of Control, and Self-Efficacy
A series of univariate analyses was conducted to assess the relationship between medication adherence and each dimension of the HBM in both younger and older cohorts. Results revealed that lower perceived treatment utility, t(138) = 3.49, p < .01, reduced sense that family and friends support treatment compliance, t(138) = 2.29, p = .02, weaker intensions to adhere, t(138) = 2.47, p = .02, and greater perceived barriers to treatment, t(138) −1.93, p = .05, were all associated with poor medication adherence in this sample. Poor adherence among younger individuals was also found to be associated with less internal locus of control, t(138) = 2.20, p = .03, greater chance locus of control, t(138) = −1.96, p = .05, and lower levels of treatment adherence self-efficacy, t(138) = 3.08, p < .01.
Examination of these variables within the older cohort revealed that poor medication adherence was associated with less internal locus of control, t(43) = 1.39, p = .05, and the perception that family and friends support regular adherence to a treatment plan, t(43) = −2.34, p = .03. Of interest, treatment adherence self-efficacy and other dimensions of the HBM were not associated with adherence among older participants.
Independent t test analyses were conducted to examine the relationship between neurocognitive functioning and medication adherence in both younger and older participants. Neurocognitive status was not significantly associated with treatment compliance among young HIV-seropositive adults; however, a number of neuropsychological domains were found to be related to adherence among older individuals. More specifically, poor adherence in this cohort was associated with weaker performance on measures of both learning and memory, t(43) = 2.07, p = .05, and executive functioning, t(43) = 2.59, p = .01. Similarly, nonadherent elders demonstrated significantly reduced levels of global cognitive functioning, t(43) = 2.74, p < .01.
After completing the series of univariate analyses described above, a multivariate model was constructed to identify the most robust predictors of medication adherence in each cohort. Stepwise logistic binary regression was used with standard entry criteria (probability of F-to-enter ≤ 0.05 and probability of F-to-remove ≥ 0.10). Only those variables demonstrating statistical significance (p < .05) during univariate analyses were entered into each of the models. Results of multivariate analyses for the younger cohort are displayed in . The overall multivariate model was significant, χ2(2, N = 140) = 22.27, p < .001, and correctly classified 73.0% of cases. Results show that both perceived treatment utility (β = .162, p = .004) and treatment adherence self efficacy (β = .285, p = .01) made independent, and statistically significant, contributions to the prediction of medication adherence. After controlling for other predictors in the model, the odds of good medication adherence increased by 1.176 (CIor = 1.06, 1.31) with each unit increase in perceived treatment utility and by 1.330 (CIor = 1.06, 1.67) with each unit increase in treatment self-efficacy.
Multivariate Logistic Regression Analysis Predicting Adherence Versus Nonadherence Among Younger HIV-Positive Individuals
Results of multivariate analyses for the older cohort are presented in . The overall multivariate model was again significant, χ2(1, N = 45) = 5.97, p = .02, and correctly classified 76.9% of cases. Results show that neurocognitive functioning (β = .022, p = .02) remained the sole significant predictor of medication adherence among older participants. After controlling for other predictors in the model, the odds of good medication adherence decreased by 1.220 (CIor = 1.03, 1.42) with each unit decrease in global cognitive functioning.
Multivariate Logistic Regression Analysis Predicting Adherence Versus Nonadherence Among Older HIV-Positive Individuals