Description of the Cohort
The target enrollment for the study was 200 participants. Two hundred thirty-nine persons were approached for enrollment before 200 agreed to participate. The 39 persons who declined enrollment were similar in age, sex, race/ethnicity, and site of diagnosis to the 200 enrolled participants (data not shown). Of the 200 enrolled participants: 12 were excluded because they were found to have a false positive rapid HIV test (n = 5); had a diagnosis of HIV older than 90 days (n = 4) upon medical record review; had already linked to care (n = 2); or were transferred outside of Houston for medical care (n = 1). Of the 188 eligible participants, three withdrew consent and one died before completing the baseline survey. Four in the group that completed the baseline survey did not complete the CES-D, leaving 180 participants in the present analyses; 81% were enrolled within 1 month of notification of HIV serostatus.
Participants ranged in age from 18 to 70 years (mean 38, SD, 11). Females comprised 32% of the cohort (). The majority of participants were from racial/ethnic minority backgrounds, with 51% non-Hispanic Black, 39% Hispanic, and 10% non-Hispanic White. Approximately half of the respondents had not completed high school and was unemployed, while two-thirds earned <$15,000 per year. The majority of the cohort was uninsured. The pre-dominant HIV risk factor was heterosexual sex (60%), followed by men who had sex with men (MSM) (32%) and IDU (8%).
Baseline characteristics and characteristics associated with depression in univariate analyses in the Steps Study cohort of persons newly diagnosed with HIV infection in Houston, TX (n = 180)
Forty-one percent of participants were recruited while hospitalized, while the remainder was recruited from out-patient facilities or the ER. At baseline, approximately half had CD4 cell counts of 200 × 106/l or less. A substantial proportion (39%) of participants reported some illicit drug use in the last 6 months, primarily use of marijuana (26%), cocaine (21%), and/or pain medication or sedatives (15%). The CAGE alcoholism screen was positive in 24% of the cohort.
Depression Prevalence and Correlates in Univariate Analyses
Using the 20-item CES-D scale, 67% of participants screened positive for depression. Eliminating the five somatic items from the CES-D and retaining the same cut-point at 16 resulted in a prevalence of 45%.
In the univariate analyses, depressed participants were more likely to be women (χ2 = 4.2, P = 0.04) and report unemployment (χ2 = 4.9, P = 0.03) and lower yearly income (χ2 = 9.7, P = 0.008; ). There were no differences in age, race, education, insurance status, place of diagnosis, baseline CD4 cell count, or HIV risk factor between the depressed and not depressed groups.
Any drug use in the preceding 6 months was highly associated with a positive depression screen (χ2 = 11.8, P = 0.0006). Specifically, depressed participants were over twice as likely to report recent marijuana use (χ2 = 9.7, P = 0.002), four times as likely to report recent cocaine use (χ2 = 11.3, P = 0.0008), and almost three times as likely to report recent sedative or pain medication abuse (χ2 = 4.7, P = 0.03), compared to non-depressed participants. Significantly more depressed than non-depressed participants screened positive for alcoholism (χ2 = 3.7, P = 0.05).
At the baseline survey, self-reported low access to care was strongly associated with depressive symptoms (χ2 = 14.7, P = 0.0001), as was a report of needing medical care in the preceding 6 months but being unable to obtain it (χ2 = 8.8, P = 0.003; ). Depressed participants were also significantly more likely to report barriers to appointment adherence (χ2 = 7.5, P = 0.006). Low self-efficacy (χ2 = 10.7, P = 0.001) and social support (χ2 = 4.3, P = 0.04) were twice as prevalent in the depressed cohort compared to the non-depressed cohort. Levels of trust in the health care system and physicians and perceived benefits of care were similarly high in both groups. Univariate analyses of depression correlates using the abbreviated CES-D yielded similar results (data not shown).
Multivariate Model of Depression
The multivariate logistic regression model of depression included age, gender, education, income, employment status, HIV risk factor, CD4 cell count, alcohol use, substance abuse, unmet need for medical care in the last 6 months, access to care, barriers to adherence to appointments, self-efficacy, and social support (). Strong associations between depression and female sex (adjusted odds ratio [OR] 5.71, 95% CI 1.76, 18.5; Wald χ2 = 8.4, P = 0.004), any substance abuse in the last 6 months (adjusted OR 3.93, 95% CI 1.49, 10.3, Wald χ2 = 7.7, P = 0.006), low access to care (adjusted OR 4.69, 95% CI 1.48, 14.9, Wald χ2 = 6.9, P = 0.009), and low self-efficacy (adjusted OR 3.05, 95% CI 1.22, 7.63, Wald χ2 = 5.7, P = 0.02), were seen. Income ≥ $25,000 per year (adjusted OR 0.28, 95% CI 0.09, 0.88, Wald χ2 = 4.7, P = 0.03), and baseline CD4 cell count between 201 and 350 cells × 106/l (adjusted OR 0.26, 95% CI 0.08, 0.79, Wald χ2 = 5.6, P = 0.02, compared to ≤200), were protective of a positive screen for depression. Analyses of the correlates of depression in the multivariate model using the CES-D 15 yielded similar results for gender, barriers to appointment adherence, and self-efficacy, though in the multivariate analysis income was no longer statistically significant (adjusted OR 0.69, 95% CI 0.21, 2.26, Wald χ2 = 0.4, P = 0.54). CD4 cell count between 201 and 350 cells × 106/l (adjusted OR 0.37, 95% CI 0.12, 1.17, Wald χ2 = 2.8, P = 0.09, compared to ≤200), alcohol abuse (adjusted OR 2.39, 95% CI 0.85, 6.74, Wald χ2 = 2.7, P = 0.10), substance abuse (adjusted OR 2.10, 95% CI 0.87, 5.06, Wald χ2 = 2.7, P = 0.10), and low social support (adjusted OR 2.33, 95% CI 0.88, 6.25, Wald χ2 = 2.9, P = 0.09), were of borderline significance.
Variables associated with depression in multivariate analysis in the Steps Study cohort of persons newly diagnosed with HIV infection in Houston, TX (n = 158)
Linkage to Care Analyses
During the first 180 days after diagnosis, 60% of the original cohort of 180 participants linked into HIV care (). There was a trend toward poorer linkage in the depressed group, as only 56% of the depressed participants were linked to care compared to 68% of the non-depressed participants, though this result was not statistically significant (χ2 = 2.6, P = 0.11). In the univariate analysis, variables significantly associated (P < 0.05) with successful linkage to care included employment and poor self-efficacy, while the other factors listed in were not significant. In a multivariate model that also included age, gender, employment status, and self-efficacy, depression remained a borderline predictor of unsuccessful linkage to care (adjusted OR 2.00, 95% CI 0.96, 4.14, Wald χ2 = 3.5, P = 0.06), as did unemployment (adjusted OR 1.78, 95% CI 0.92, 3.43, Wald χ2 = 3.0, P = 0.09), while high self-efficacy (adjusted OR 2.72, 95% CI 1.39, 5.31, Wald χ2 = 8.6, P = 0.003) also predicted unsuccessful linkage. Thirty-six of the 180 participants were unable to be contacted or verified as alive and still in the Houston area at 180 days after enrollment, including 11 participants who were known to be deceased, two who withdrew consent after completing the baseline survey, and 23 who were lost to follow-up. Analyses using data from the remaining 144 persons revealed a similar relationship, with 68% of the depressed group versus 78% of the non-depressed group successfully linking into care (χ2 = 1.8, P = 0.17). The multivariate model in this sub-population was similar: depression was a borderline predictor of poor linkage to care (adjusted OR 2.20, 95% CI 0.87, 5.59, Wald χ2 = 2.7, P = 0.10), while unemployment (adjusted OR 2.46, 95% CI 1.02, 5.97, Wald χ2 = 4.0, P = 0.05), and high self-efficacy (adjusted OR 4.07, 95% CI 1.67, 9.91, Wald χ2 = 9.5, P = 0.002) were significant predictors of unsuccessful linkage.
Linkage to care by depression status in the Steps Study cohort of persons newly diagnosed with HIV infection in Houston, TX (n = 180)