Social inequalities contribute to racial/ethnic disparities in depressive symptomology among men who have sex with men

Abstract

Purpose

Racial/ethnic minorities experience disproportionate rates of depressive symptoms in the United States. The magnitude that underlying factors—such as social inequalities—contribute to these symptoms is unknown. We sought to identify exposures that explain racial/ethnic differences in clinically significant depressive symptomology among men who have sex with men (MSM).

Methods

Data from the Multicenter AIDS Cohort Study (MACS), a prospective cohort study, were used to examine clinically significant symptoms of depression (Center for Epidemiologic Studies Depression Scale score ≥ 20) among non-Latinx White, non-Latinx Black, and Latinx MSM. We included 44,823 person-visits by 1729 MSM seen in the study sites of Baltimore/Washington, DC; Chicago; Pittsburgh/Columbus; and Los Angeles from 2000 to 2017. Regression models estimated the percentage of depressive symptom risk explained by social, treatment, and health-related variables related to race/ethnicity. Machine-learning methods were used to predict the impact of mitigating differences in determinants of depressive symptoms by race/ethnicity.

Results

At the most recent non-missing MACS visit, 16% of non-Latinx White MSM reported clinically significant depressive symptoms, compared to 22% of non-Latinx Black and 25% of Latinx men. We found that income and social-environmental stress were the largest contributors to racial/ethnic disparities in risk for depressive symptoms. Similarly, setting the prevalence of these two exposures to be equal across racial/ethnic groups was estimated to be most effective at reducing levels of clinically significant depressive symptoms.

Conclusion

Results suggested that reducing socioeconomic inequalities and stressful experiences may be effective public health targets to decrease racial/ethnic disparities in depressive symptoms among MSM.

Appendix

Baseline educational attainment was categorized as high school or less, some college, college graduate, and at least some postgraduate education. Categories were modeled after [66]. Recruitment center location was categorized into the four active Multicenter AIDS Cohort Study (MACS) sites: Baltimore, Maryland/Washington, DC; Chicago, Illinois; Pittsburgh, Pennsylvania/Columbus, Ohio; and Los Angeles, California. Recruitment cohort was coded into categories based on the MACS recruitment waves captured in this study: 1984, 1987, and 2002. The latest MACS recruitment wave, initiated in 2010, was not included in this study, because the social-environmental stress and social mobility MACS questions were asked before the 2010 recruitment began and individuals who did not respond to the social-environmental stress and social mobility questions were excluded.

Illicit drug use since the last MACS visit was coded as binary, with an individual classified as having taken illicit drugs if they had at least weekly use of poppers, crack, methamphetamine, cocaine, heroin, speedball, and/or ecstasy since the last MACS visit or used two or more of the above listed drugs since the last MACS visit. This definition came from [66].

Self-reported annual income at a MACS visit was categorized as less than $20,000; $20,000 to $39,999; $40,000 to $59,999; and $60,000 or more, with categories adapted from [66]. Nearest-neighbor interpolation was used to fill in missing incomes with values within 6 months of the missing data. Possession of health insurance during a MACS visit was coded as binary. Social-environmental stress was treated as binary and adapted from the Urban Life Stress Scale [67]. A participant was coded as positive for stress if they reported that the neighborhood environment, crime/violence, racism/discrimination, or police relations had ever caused a lot or extreme stress.

Sum parental educational attainment was calculated by summing a participant’s father’s and mother’s highest level of education, while the participant was in high school. The participant’s father’s and mother’s education levels were coded as 0 = high school or less; 1 = junior college or trade school; 2 = college graduate; 3 = postgraduate education; and missing = missing visit, no father/mother while participant was in high school, participant does not know, or participant refused to answer. Therefore, the generated sum parental educational attainment variable ranged from zero to six.

Social mobility served as a proxy for the social opportunities taken advantage of by a participant and was calculated by first averaging the highest level of education of a participant’s father and mother, while the participant was in high school, and then subtracting this value from the participant’s education level reported at their baseline visit. Because both the average of the highest level of education of a participant’s father and mother and a participant’s baseline education ranged from zero to three, the generated social mobility variable ranged from negative three to three, with negative numbers indicating reduced mobility and positive numbers indicating increased mobility.

Use of antidepressant medication at a MACS visit was coded as binary. Adherence to highly active antiretroviral therapy (HAART) since last MACS visit among persons living with HIV was defined as less than 75%, 75% to 94%, 95% to 99%, and 100% adherence, as determined through self-report. This definition was adapted from [68]. Reported use of HAART by an individual living with HIV at a MACS visit was treated as binary, with any type of HAART (protease inhibitors, integrase inhibitors, etc.) being coded as positive.

Family history of depression, a non-modifiable risk factor for depression [69], was coded as binary, with a positive value being assigned if a participant reported that a relative had ever been diagnosed as having depression. Relatives were defined as the biological mother, father, brothers, and sisters.

Burden of comorbidities reported at a MACS visit was categorized as zero comorbidities, one comorbidity, and two or more comorbidities. Six possible comorbid conditions were included: hepatitis C infection (defined as detectable hepatitis C RNA in serum); high blood pressure (defined as systolic pressure > 140 mmHg or diastolic pressure > 90 mmHg); diabetes (defined as fasting glucose ≥ 126 mg/dL or a self-report of previous clinical diabetes diagnosis with use of medication); dyslipidemia (defined as either a fasting total cholesterol ≥ 200 mg/dL, low-density lipoprotein ≥ 130 mg/dL, high-density lipoprotein < 40 mg/dL, triglycerides ≥ 150 mg/dL, or use of lipid-lowering medications with self-report of a previous clinical diagnosis); kidney disease (defined as an estimated glomerular filtration rate < 60 mL/min/1.73 m² body surface area or a urine protein-to-creatinine ratio ≥ 200 mg protein/1 g creatinine); or cancer (defined as diagnosis of any cancer type at, or within a year of, the MACS study visit). These comorbid conditions were selected and defined in the study [70].

Insomnia, a risk factor for depression [71], was coded as binary. An individual who reported insomnia or problems sleeping since the last MACS visit was treated as a positive response. HIV serostatus was treated as binary, with an individual being coded as seropositive or seronegative at a MACS visit. HIV serostatus was determined through enzyme-linked immunosorbent assay (ELISA), confirmed by Western blot [24].

CD4 cell count at HAART initiation captured the number of CD4 cells, as determined by flow cytometry, an individual living with HIV had when they initiated HAART (or the earliest MACS visit post-HAART initiation). Viral copy-years following HAART initiation served as a cumulative measure of a person living with HIV’s viral load after they began taking HAART. Viral load was measured via Roche assays. The definition and equation of viral copy-years came from the study [72], and the variable is conceptually similar to pack-years for smoking. Before calculation of viral copy-years, missing viral load values were filled in using multiple imputation with the Markov chain Monte Carlo method [73]. Age, current CD4 cell count, and HAART treatment history at a MACS visit were used as predictor variables.