Association of Social Adversity with Comorbid Diabetes and Depression Symptoms in the Hispanic Community Health Study/Study of Latinos Sociocultural Ancillary Study: A Syndemic Framework

Abstract

Background

U.S. Hispanics/Latinos experience high lifetime risk for Type 2 diabetes and concurrent psychological depression. This comorbidity is associated with poorer self-management, worse disease outcomes, and higher mortality. Syndemic theory is a novel social epidemiological framework that emphasizes the role of economic and social adversity in promoting disease comorbidity and health disparities.

Purpose

Informed by the syndemic framework, this study explored associations of socioeconomic and psychosocial adversity (low income/education, trauma history, adverse childhood experiences, ethnic discrimination, neighborhood problems [e.g., violence]) with comorbidity of diabetes and depression symptoms in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) and Sociocultural Ancillary Study.

Methods

Participants were 5,247 Latino adults, aged 18–74, enrolled in four U.S. cities from 2008 to 2011. Participants completed a baseline physical exam and measures of depression symptoms and psychosocial adversity. Multinomial logistic regression analyses were conducted to examine associations of adversity variables with comorbid diabetes and high depression symptoms.

Results

Household income below $30,000/year was associated with higher odds of diabetes/depression comorbidity (odds ratio [OR] = 4.61; 95% confidence interval [CI]: 2.89, 7.33) compared to having neither condition, as was each standard deviation increase in adverse childhood experiences (OR = 1.41; 95% CI: 1.16, 1.71), ethnic discrimination (OR = 1.23; 95% CI: 1.01, 1.50), and neighborhood problems (OR = 1.53; 95% CI: 1.30, 1.80).

Conclusion

Low household income, adverse childhood experiences, ethnic discrimination, and neighborhood problems are related to comorbid diabetes and depression in U.S. Latinos. Future studies should explore these relationships longitudinally.

Low household income is associated with comorbidity of diabetes and high depression symptoms in U.S. Latinos. Adverse childhood experiences, discrimination, and neighborhood problems are associated with comorbidity compared to having neither condition.

Introduction

Hispanics/Latinos (hereafter, Latinos) are the largest minority group in the USA and have the highest lifetime risk of Type 2 diabetes mellitus (T2DM) among all major racial/ethnic groups [1–3]. Compared to non-Hispanic whites, Latino individuals with T2DM are more likely to have poorer diabetes outcomes [4–6]. While comorbid depression among adults with diabetes is common across populations [7, 8], the risk of this comorbidity is twice as high in Latinos than in the general population [4, 9). As many as 33% of Latino individuals with T2DM experience concurrent depression [4]. Even when unaware of their diabetes diagnosis, Mexican Americans with diabetes are more likely to experience clinically significant depression symptoms than those without diabetes [10].

The term syndemic, a combination of synergy and epidemic, refers to a synergistic co-occurrence of two or more diseases that is precipitated or exacerbated by social and economic inequality and results in an increased burden of disease for a particular population [11–13). Adverse conditions across socio-ecological levels, from upstream socioeconomic adversity (e.g., poverty/low socioeconomic status [SES]) to proximal psychosocial and neighborhood environmental conditions (e.g., psychological trauma, family disruption, discrimination, unstable housing) are theorized to drive the development of syndemics. As a result, health vulnerability and risky health behaviors increase, which promotes the occurrence and clustering of diseases [11, 14, 15]. The syndemic framework offers a novel approach for the investigation of disease clustering that is of high interest, as evidenced by the March 2017 special series on syndemics in The Lancet [16, 17]. Three key phenomena comprise a syndemic: (a) two or more diseases cluster or are frequently comorbid in a given population; (b) there is bidirectionality and/or interaction between the diseases; and (c) social and economic adversity promotes this co-occurrence and bidirectionality/interaction [11, 14, 18]. Beyond recognizing the presence of syndemics when they occur and exploring biological pathways of disease interaction, the Lancet series urges researchers to identify specific adverse social conditions that are associated with and may increase the likelihood of the syndemic in specific populations, as a critical step toward improving prevention and intervention efforts, informing health policy, and reducing health inequity [14, 17, 18].

A substantial body of evidence documents bidirectionality and synergistic associations between depression and T2DM [19–22], with particularly detrimental effects (poorer self-management, more complications, and higher disability) in Latinos [4, 6, 23], supporting the existence of a diabetes and depression syndemic in this population. Greater mortality and higher health care costs associated with comorbid diabetes and depression have also been documented across diverse populations, including Latinos [19, 24–27]. The risk of these adverse sequelae of T2DM suggests an urgent need to identify factors contributing to high rates of diabetes and depression in U.S. Latino individuals.

The conceptualization of comorbid T2DM and depression as a syndemic may be helpful in identifying risk factors that could be targets of interventions to reduce comorbidity and poor health outcomes. Both upstream socioeconomic disadvantage and proximal psychosocial adversity (e.g., trauma/abuse, adverse childhood experiences) are associated with increased depressive symptoms and risk of diabetes and metabolic syndrome [28–31]. Associations between perceived ethnic/racial discrimination and depression are also well documented [32–34]. Neighborhood quality (e.g., safety, walkability, green space) is increasingly a focus of public health investigation [35] and may affect risk for diabetes both directly and indirectly through influence on physical activity and diet. The Multi-Ethnic Study of Atherosclerosis (MESA), a large, population-based longitudinal cohort study that includes Latino adults, showed that individuals in the bottom tertile of neighborhood physical activity resources (e.g., fewest number of sport facilities, gyms, green spaces) develop diabetes at nearly double the rate of those in the top tertile [36]. Additional studies support these observations and describe an association of neighborhood factors with diabetes, hypertension, obesity, smoking, physical activity, and emotional well-being [37–39]. However, few studies have explored these associations with comorbid diabetes mellitus and depression, and none has systematically done so in a large, diverse sample of U.S. Latinos.

The primary objective of this analysis was to explore the hypothesis that an array of social adversity variables (low income, low education, trauma history, adverse childhood experiences, discrimination, and neighborhood problems) are associated with the comorbidity of diabetes and high self-reported depressive symptoms in a large, population-based sample of U.S. Latino individuals. Given the volume of prior research demonstrating bidirectionality and interaction of depression and diabetes across populations, rather than attempting to demonstrate a diabetes-depression syndemic in this population, we explore whether or not a set of adversity variables with documented relationships to health may be associated with the comorbidity in a diverse Latino sample.

Materials and Methods

The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) is a population-based epidemiological cohort study of chronic disease prevalence and risk and protective factors in 16,415 Latino adults aged 18–74 years at enrollment. Participants self-identified as Hispanic or Latino and of Mexican, Cuban, Puerto Rican, Dominican, Central/South American, or other or more than one Latino heritage. Recruitment was conducted via a stratified two-stage probability sampling approach in four U.S. field centers (Bronx, NY; Chicago, IL; Miami, FL; San Diego, CA) between 2008 and 2011, with oversampling of the 45–74-year-old age group. Full descriptions of the HCHS/SOL sampling approach [40] and methods [41] have been previously published. A baseline clinical examination and interview were conducted in participants’ preferred language (Spanish or English) and included demographic and biological assessments (e.g., anthropometrics, blood draw). The study was approved by the Institutional Review Boards at all HCHS/SOL institutions and written informed consent was obtained from all participants.

To more thoroughly examine psychological and sociocultural factors related to chronic disease, the Sociocultural Ancillary Study (SCAS) recruited a subset of 5,312 participants from the HCHS/SOL cohort (~1320 per field center) [42]. All participants of the HCHS/SOL cohort who consented to being contacted for future research and were willing to attend a separate study visit within 9 months were eligible. Study recruiters attempted to reach 7,321 HCHS/SOL participants for participation: 6,246 (85.3%) were successfully contacted (1,075 could not be reached) and 5,313 participated between February 2010 and June 2011. The SCAS sample is representative of the HCHS/SOL cohort with the exception of slightly lower participation by individuals in higher SES groups [42]. SCAS participants attended a study visit within 9 months of their HCHS/SOL baseline clinical exam to complete in-person interviews in which a variety of psychosocial risk and protective variables were self-reported. Most (72.6%) completed the SCAS assessment within 4 months of their baseline exam. A full description of the SCAS methods and sample is available elsewhere [42].

Measures

Adversity variables

Household yearly income and highest education level completed were assessed in the HCHS/SOL baseline clinical interview and were included in models categorically: three categories for education level (<high school diploma/General Education Development test [GED], high school diploma/GED, >high school diploma/GED) and 10 categories for household income ranging from less than $10,000 to more than $100,000.

All psychosocial adversity variables were self-reported during the SCAS interview via measures previously validated in both Spanish and English. Item responses were summed to create total scores in accordance with scoring protocols for each measure. Total scores were standardized into z-scores prior to analyses to facilitate interpretation of model coefficients. Lifetime trauma history was measured with the 10-item Traumatic Stress Schedule (TSS) [43], which asks participants to report lifetime exposure to a variety of potentially traumatic experiences (e.g., “Did anyone ever take something from you by force or threat of force, such as in a robbery, mugging or hold up?”). Scores range from 0 to 10 with higher scores indicating a higher number of potentially traumatic experiences. Adverse childhood experiences were assessed using the 10-item Adverse Childhood Exposure (ACE) scale [44]. This measure surveys participant exposure to adverse events in childhood/adolescence that are known to be associated with later adult health (e.g., parental abuse: “Did a parent or other adult in the household often or very often push, grab, slap, or throw something at you?”). Scores range from 0 to 10 with higher scores indicating a higher number of adverse experiences. Internal consistency and other psychometric data are not reported for trauma history and adverse childhood experiences given the count format and the fact that they are not intended to capture single underlying latent constructs. Ethnic/racial discrimination was measured using the 17-item brief Perceived Ethnic Discrimination Questionnaire—Community Version (PEDQ-CV) [45]. Items on this scale assess the frequency with which participants have experienced discrimination related to race/ethnicity (e.g., “Have others made you feel like an outsider who doesn’t fit in because of your dress, speech, or other characteristics related to your ethnicity?”). Responses are scored on a five-point Likert scale from “Never” to “Very Often” and scores range from 17 to 85, with higher scores reflecting more discrimination. Internal consistency reliability (Cronbach’s alpha) was very good for both language versions (English = .91, Spanish = .87). Neighborhood problems were measured by asking participants to assess seven dimensions of their neighborhoods (noise, traffic, lack of access to adequate food, proximity of parks or playgrounds, trash/litter, availability and maintenance of sidewalks, violence) on a four-point Likert scale [46] from “Not a problem” to “Very Serious Problem.” Participants were instructed to think of their neighborhood as “the area around where you live and around your house.” Scores range from 7 to 28, with higher scores indicating more neighborhood problems. Internal consistency reliability for both language versions in this sample was acceptable (English = .79; Spanish = .78).

Depression symptoms and diabetes

Depression symptoms were assessed by the 10-item Center for Epidemiological Studies Depression Scale (CES-D-10) [47], a subset of the original 20-item CES-D scale [48]. Participants rate the frequency with which they have experienced various common symptoms of depression in the past week along a Likert-type scale ranging from “none of the time” to “most of the time.” Scores range from 0 to 30 and were dichotomized for analysis. Scores ≥10 were categorized as “high depression symptoms,” indicating the likely presence of clinically significant depression but not equivalent to a clinical diagnosis of major depression. The ≥10 cut point has shown good sensitivity and specificity when measured against the 20-item CES-D cut point of ≥16 [47], which was validated using Diagnostic and Statistics Manual of Mental Disorders - Third Edition criteria for clinical depression. Internal consistency reliability of the CES-D-10 in the SCAS cohort was .82 for both English and Spanish versions [49]. The presence of diabetes was determined according to American Diabetes Association criteria [50] (fasting plasma glucose ≥126 mg/dL [7 mmol/L], 2 hr postload glucose [2 hr oral glucose tolerance test] ≥200 mg/dL [11.1 mmol/L], or hemoglobin A1c ≥6.5%) and/or by self-report of diabetes diagnosis from a physician, and/or prescribed glucose-lowering medication. To explore associations with the comorbidity of diabetes and elevated depression symptoms, four categories based on diabetes status and depression symptoms were created: (a) individuals with no diabetes and low depression symptoms; (b) individuals with diabetes and low depression symptoms; (c) individuals with no diabetes and high depression symptoms; (d) individuals with both diabetes and high depression symptoms (comorbidity).

Covariates

Demographic covariates included age (in years), sex, Latino heritage group (e.g., Mexican, Cuban, Puerto Rican), language preference (Spanish or English), field center (Bronx, Chicago, Miami, or San Diego), nativity/immigration status (immigrated <10 years ago, immigrated ≥10 years ago, or born in U.S. mainland), and health insurance status (any or none). Models did not control for lifestyle factors (e.g., smoking, physical activity) as these factors are posited as a mechanism or pathway through which structural and psychosocial risk factors relate to diabetes, depression symptoms, and their comorbidity.

Statistical Analyses

Prevalence and means of adversity variables by four-level group membership ((a) no diabetes/low depression symptoms; (b) diabetes/low depression symptoms; (c) no diabetes/high depression symptoms; (d) diabetes/high depression symptoms [comorbidity]) were calculated with 95% confidence intervals (CIs). Both omnibus and post hoc tests were conducted to examine group differences. For this comparison, household income and education level variables were dichotomized to display categories potentially associated with higher health risks (household income ≤30,000/year; education <high school degree). Tests of group differences were conducted using nonweighed group means. Multinomial logistic regression was then conducted to explore the association of adversity variables with group membership. First, the low depression symptoms/no diabetes group was used as the reference group to examine the magnitude of odds associated with each adversity variable of having depression or diabetes alone or of being in the high depression symptoms/diabetes (comorbid) group. Then, the comorbid group was used as the reference group to explore which adversity variables were associated with odds of comorbidity compared to all other groups. To explore associations with upstream socioeconomic adversity, the first logistic regression model (Model 1) included household income and education level (10- and 3-level variables, respectively, modeled as continuous variables) with adjustment for demographic variables (age, sex, nativity/immigration status, Latino heritage group, language preference, field center, health insurance coverage). Model 2 replaced continuous income and education with dichotomized variables to explore associations with high-risk categories (household income <$30,000/year; education <high school diploma/GED). Sensitivity analyses were conducted to examine potential differences between these associations in the SCAS and full HCHS/SOL samples. Specifically, Models 1 and 2 were repeated in the full HCHS/SOL sample, with results presented in Supplementary Appendix Tables 2 and 3. Model 2 was then repeated in the SCAS sample with inclusion of the remaining block of proximal psychosocial and neighborhood adversity factors (lifetime trauma, adverse childhood experiences, perceived discrimination, and neighborhood problems; Model 3). A small number of participants (N = 72) reported previous diabetes diagnosis but were not taking medications for diabetes and diabetes was not detected by clinical exam. Model 3 was repeated in a sensitivity excluding these 72 cases.

All analyses accounted for design effects (stratification and clustering of census block groups) and sample weights, which are adjusted for nonresponse and calibrated to the 2010 U.S. census populations according to age, sex, and Latino heritage [40]. Descriptive analyses were calculated using complex survey procedures in IBM SPSS Statistics 22.0 (IPM, Armonk, NY). Regression models were estimated using the full information maximum likelihood robust estimation procedure in MPlus Version 7.0 [51], which adjusts parameters and standard errors for missing data and nonnormality. Model assumptions regarding linearity, independence, and homoscedasticity of errors were assessed prior to conducting inferential analyses.

Results

Demographics and descriptive characteristics are reported in Table 1. The SCAS population age ranged from 18 to 74 years (mean [M] = 44.5, standard deviation [SD] = 13.6) and 54.9% were women. The majority of the population were born outside of the U.S. mainland (78%), preferred Spanish over English (75.4%), and had a household income under $30,000/year (70.1%). Mexican and Mexican American individuals comprised the largest Latino heritage group (36.5%). An average of 2.1 (standard error [SE] = 0.05, range = 0–9) lifetime traumatic events and 2.5 (SE = 0.05, range = 0–10) adverse childhood experiences were reported. Average perceived discrimination score was 25.1 (SE = 0.20, range = 17–85) and average neighborhood problems score was 12.0 (SE = 0.11, range = 7–28). Unadjusted bivariate correlations among all adversity variables are reported in Supplementary Appendix Table 1. Prevalence of diabetes was 16.9% (95% CI: 15.5, 18.4) and 29.3% (95% CI: 27.5, 31.1) reported high (i.e., clinically significant) depression symptoms. Overall prevalence of the comorbidity of diabetes and high depression symptoms was 6.8% (95% CI: 6.0, 7.8).

Table 1.

Descriptive statistics for Hispanic Community Health Study/Study of Latinos (HCHS/SOL) Sociocultural Ancillary Study (SCAS) sample (N = 5,313)

	N	Weighted % (95% CI)
Age
18–44	2,035	56.5 (54.2, 58.7)
44+	3,278	43.5 (41.3, 45.8)
Female	3,299	54.9 (52.9, 56.8)
Male	2,014	45.1 (43.2, 47.1)
Hispanic/Latino Background
Central/South American	903	12.4 (10.7, 14.3)
Cuban	775	20.3 (16.3, 25.1)
Dominican	534	11.7 (9.9, 13.8)
Mexican	2,080	36.5 (32.6, 40.6)
Puerto Rican	880	15.8 (13.8, 18.0)
More than one/Other	137	3.3 (2.4, 4.5)
Field center
Bronx, NY	1,342	30.2 (26.4, 34.4)
Chicago, IL	1,329	15.6 (13.2, 18.4)
Miami, FL	1,315	29.1 (23.8, 35.1)
San Diego, CA	1,327	25.0 (21.2, 29.2)
Education completed
<High school or GED	1,923	32.5 (30.4, 34.8)
High school or GED	1,383	28.0 (26.3, 29.8)
>High school or GED	1,998	39.4 (37.0, 42.0)
Household income
<$30,000/year	3,515	70.1 (67.3, 72.8)
>$30,000/year	1,357	29.9 (27.2, 32.7)
Nativity/immigration status
Born in the U.S. mainland	917	22.0 (19.7, 24.4)
Immigrated ≥10 years ago	3,137	50.9 (48.4, 30.1)
Immigrated <10 years ago	1,247	27.2 (24.4, 38.1)
Spanish language interview	4,296	75.4 (72.5, 78.1)
Health insurance coverage	2,670	52.3 (49.7, 54.8)
Diabetesa	1,136	16.9 (15.5, 18.4)
Objectively diagnosedb	1,064	16.0 (14.7, 17.4)
High depression symptoms (CESD-10 >10)	1,618	29.3 (27.5, 31.3)
	N	Weighted M (SE)
Lifetime trauma history (TSS)	5,264	2.1 (0.04)
Adverse childhood experiences (ACES)	5,244	2.5 (0.05)
Perceived discrimination (PED-Q)	5,174	25.1 (0.20)
Neighborhood problems	5,258	12.0 (0.11)

ACES Adverse Childhood Exposures Scale; CESD-10 Center for Epidemiologic Studies Depression Scale-10-item version (scores ≥10 = clinically significant depression symptoms); GED General Education Development test; PEDQ Perceived Ethnic Discrimination Questionnaire; SE standard error; TSS Traumatic Stress Schedule (count of lifetime burden/exposure to traumatic events); 95% CI 95% confidence interval.

^aBased on diagnostic criteria recommended by the American Diabetes Association (ADA): [fasting plasma glucose ≥126 mg/dL (7 mmol/L), 2 hr postload glucose (2 hr oral glucose tolerance test [OGTT]) ≥200 mg/dL (11.1 mmol/L), or HbA1c ≥6.5%] and/or by self-report of diabetes diagnosis from a physician or prescribed glucose-lowering medication; self-report only (1.4%); objectively confirmed (98.6%).

^bBased on ADA criteria or glucose-lowering medication use (excluding 72 cases of self-report only).

Prevalence and means of adversity variables by group membership (no diabetes/low depression symptoms, diabetes/low depression symptoms, no diabetes/high depression symptoms, diabetes/high depression symptoms [comorbidity]) are presented in Table 2. Statistically significant differences by group membership are evident, with individuals in the comorbid diabetes/high depression symptoms group and no diabetes/high depression symptoms group experiencing the highest prevalence and means for all adversity variables. Notably, the prevalence of lower household income (<$30,000/year) was almost 90% in the comorbid group. Post hoc comparisons of individual group differences revealed that prevalence and means of all adversity variables except lower education were significantly higher for those in the comorbid group compared to individuals with neither condition (no diabetes/low depression symptoms group) or diabetes only (diabetes/low depression symptoms group). When sensitivity analyses were conducted to compare diabetes/depression group differences in adversity variables in the SCAS to the full HCHS/SOL sample, no substantive differences in results were found. Results for the full sample are presented in Supplementary Appendix Table 2.

Table 2.

Structural and psychosocial risk factors by group membership in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) Sociocultural Ancillary Study (SCAS) (N = 5,247)^a

	No diabetesb/ low depression symptomsc (n = 2,946)	Diabetes/low depression symptoms (n = 683)	No diabetes/ high depression symptoms (n = 1,182)	Diabetes/ high depression symptoms (n = 436)	Omnibus ANOVA/ General Linear Model (p-value)
Socioeconomic risk factors	Weighted % (95% CI)
Household income <$30,000/year	63.3 (59.6, 66.8)d	71.1 (65.4, 76.2)d,e	82.7 (79.1, 85.9)e	89.9 (85.7, 92.9)e	p < .01
Education <high school/GED	26.6 (23.6, 30.0)d	42.7 (37.2, 48.4)e	39.3 (35.3, 43.5)d,e	46.3 (40.0, 52.7)e	p < .01
Psychosocial risk factors	M (SE)
Lifetime trauma history (TSS)	2.0 (0.05)d	1.9 (0.09)d	2.5 (0.10)e	2.2 (0.13)e	p < .01
Adverse childhood experiences (ACES)	2.2 (0.06)d	2.1 (0.12)d	3.3 (0.12)e	3.0 (0.17)e	p < .01
Perceived discrimination (PED-Q)	24.4 (0.24)d	24.0 (0.47)d	27.4 (0.43)e	25.6 (0.66)e	p < .01
Neighborhood problems	11.5 (0.15)d	11.3 (0.20)d	13.1 (0.20)e	13.0 (0.27)e	p < .01

ACES Adverse Childhood Exposures Scale; GED General Education Development test; PEDQ Perceived Ethnic Discrimination Questionnaire; TSS Traumatic Stress Schedule.

^aIndividuals with data for both diabetes status and depression symptoms.

^bBased on diagnostic criteria of the American Diabetes Association: [fasting plasma glucose ≥126 mg/dL (7 mmol/L), 2 hr postload glucose (2 hr OGTT) ≥200 mg/dL (11.1 mmol/L), or HbA1c ≥6.5%] and/or by self-report of diabetes diagnosis from a physician, and/or prescribed glucose-lowering medication.

^cDepression symptoms measured by the Center for Epidemiological Studies Depression Scale-10-item version (CES-D-10), “high depression” = CES-D-10 score >10.

^dStatistically significant difference compared to diabetes/high depression symptoms group.

^eStatistically significant difference compared to no diabetes/low depression symptoms group.

Multinomial logistic regression analyses are presented in Table 3 (no diabetes/low depression as the reference group) and Table 4 (comorbid group as the reference group). In Table 3, Model 1, which included continuous socioeconomic variables, after adjustment for covariates, a one-category increase in household yearly income was associated with 38% lower odds of diabetes/high depression symptom comorbidity (OR = 0.72; 95% CI: 0.66, 0.77) compared to having neither condition. Education level was not associated with the odds of diabetes/high depression symptoms comorbidity versus neither condition. In Model 2, categorical analyses showed that low household income (<$30,000/year) was associated with a more than fourfold increase in odds of diabetes/high depression symptom comorbidity (OR = 4.61; 95% CI: 2.89, 7.33) compared to the odds of having neither condition. Having less than a high school degree or GED was also associated with increased odds of no diabetes/high depression symptoms (OR = 1.42; 95% CI: 1.09, 1.84) but not of diabetes/high depression symptoms compared to neither condition. Sensitivity analyses again showed a similar pattern of results in the HCHS/SOL full sample, with the exception that in Model 1, continuous analyses showed that higher education was significantly associated with lower odds of diabetes/high depression symptom comorbidity (OR = 0.79; 95% CI: 0.70, 0.90; Supplementary Appendix Table 3).

Table 3.

Multinomial logistic regression analysis with no diabetes^a/low depression symptoms^b as reference group

	Diabetes/low depression symptoms			No diabetes/high depression symptoms			Diabetes/high depression symptoms
	OR	95% CI	p-value	OR	95% CI	p-value	OR	95% CI	p-value
Model 1: Continuous Socioeconomic Variables + Covariatesc
Household incomed	0.95	0.89, 1.00	.07	0.81	0.76, 0.85	.00*	0.72	0.66, 0.77	.00*
Education level completede	0.85	0.71, 1.01	.07	0.84	0.74, 0.96	.00*	0.87	0.72, 1.06	.16
Model 2: Socioeconomic Risk Categories + Covariates
Low household income (<$30,000/year)	1.25	0.91, 1.73	.17	2.66	1.93, 3.66	.00*	4.61	2.89, 7.33	.00*
Low education level (<High school diploma/GED)	1.40	1.03, 1.91	.03*	1.42	1.09, 1.84	.02*	1.28	0.89, 1.82	.19
Model 3: Socioeconomic + Psychosocial Variables + Covariates
Low household income (<$30,000/year)	1.17	0.85, 1.62	.33	2.07	1.54, 2.77	.00*	3.28	2.08, 5.17	.00*
Low education level (<High school diploma/GED)	1.38	1.02, 1.88	.03*	1.36	1.04, 1.77	.02*	1.22	0.86, 1.71	.27
Lifetime trauma history (TSS)	0.97	0.83, 1.12	.65	1.17	1.03, 1.34	.02*	0.96	0.79, 1.16	.67
Adverse childhood experiences (ACES)	1.01	0.86, 1.19	.87	1.26	1.12, 1.42	.00*	1.41	1.16, 1.71	.00*
Perceived discrimination (PEDQ)	1.06	0.89, 1.27	.49	1.26	1.12, 1.41	.00*	1.23	1.01, 1.50	.04*
Neighborhood problems	1.12	0.95, 1.30	.18	1.37	1.20, 1.55	.00*	1.53	1.30, 1.80	.00*

ACES Adverse Childhood Exposures Scale; GED General Education Development test; OR odds ratio; PEDQ Perceived Ethnic Discrimination Questionnaire; TSS Traumatic Stress Schedule; 95% CI 95% confidence interval.

^aBased on diagnostic criteria recommended by the American Diabetes Association: [fasting plasma glucose ≥126 mg/dL (7 mmol/L), 2 hr postload glucose (2 hr OGTT) ≥200 mg/dL (11.1 mmol/L), or HbA1c ≥ 6.5%] and/or by self-report of diabetes diagnosis from a physician, and/or prescribed glucose-lowering medication.

^bDepression symptoms as measured by the Center for Epidemiological Studies Depression Scale-10-item version (CES-D), “high depression” = CES-D score >10.

^cCovariates = age, sex, nativity/immigration, ancestry, language preference, field center, health insurance.

^dTen-level variable categorized as <$10,000, $10,000–15,000, >$15,000–20,000, >$20,000–25,000, >$25,000–30,000, $30,000–40,000, >$40,000–50,000, >$50,000–75,000, >$75,000–100,000, or >$100,000.

^eThree-level variable categorized as <high school diploma/GED, high school diploma/GED, or >high school diploma/GED.

*Statistically significant at p < .05.

Table 4.

Multinomial logistic regression analysis with comorbid diabetes^a/high depression symptoms^b as reference group

	No diabetes/low depression symptoms			Diabetes/low depression symptoms			No diabetes/high depression symptoms
	OR	95% CI	p-value	OR	95% CI	p-value	OR	95% CI	p-value
Model 1: Continuous Socioeconomic Variables + Covariatesc
Household income4	0.72	0.66, 0.77	.00*	0.76	0.70, 0.82	.00*	0.89	0.82, 0.97	.01*
Education level completed5	0.87	0.72, 1.06	.16	1.03	0.83, 1.27	.81	1.04	0.86, 1.26	.81
Model 2: Socioeconomic Risk Categories + Covariates
Low household income (<$30,000/year)	0.22	0.14, 0.35	.00*	0.27	0.17, 0.43	.00*	0.58	0.37, 0.91	.02*
Low education level (<High school diploma/GED)	0.78	0.55, 1.12	.18	1.10	0.75, 1.62	.63	1.11	0.78, 1.57	.56
Model 3: Socioeconomic + Psychosocial Variables + Covariates
Low household income (<$30,000/year)	0.23	0.16, 0.41	.00*	0.31	0.20, 0.50	.00*	0.59	0.37, 0.94	.03*
Low education level (<High school diploma/GED)	0.77	0.54, 1.09	.14	1.08	0.74, 1.58	.69	1.09	0.78, 1.53	.62
Lifetime trauma history (TSS)	1.04	0.85, 1.27	.71	1.00	0.82, 1.22	.99	1.21	0.98, 1.50	.07
Adverse childhood experiences (ACES)	0.70	0.58, 0.85	.00*	0.72	0.58, 0.89	.00*	0.90	0.73, 1.10	.30
Perceived discrimination (PEDQ)	0.83	0.68, 1.03	.09	0.89	0.72, 1.09	.26	1.04	0.85, 1.27	.70
Neighborhood problems	0.63	0.53, 0.75	.00*	0.71	0.59, 0.86	.00*	0.88	0.75, 1.03	.12

ACES Adverse Childhood Exposures Scale; GED General Education Development test; OR odds ratio; PEDQ Perceived Ethnic Discrimination Questionnaire; TSS Traumatic Stress Schedule; 95% CI 95% confidence interval.

^aBased on diagnostic criteria recommended by the American Diabetes Association: [fasting plasma glucose ≥126 mg/dL (7 mmol/L), 2 hr postload glucose (2-hr OGTT) ≥200 mg/dL (11.1 mmol/L), or HbA1c ≥6.5%] and/or by self-report of diabetes diagnosis from a physician, and/or prescribed glucose-lowering medication.

^bDepression symptoms as measured by the Center for Epidemiological Studies Depression Scale-10-item version (CES-D), “high depression” = CES-D score >10.

^cCovariates = age, sex, nativity/immigration, ancestry, language preference, field center, health insurance.

^dTen-level variable categorized as <$10,000, $10,000–15,000, >$15,000–20,000, >$20,000–25,000, >$25,000–30,000, $30,000–40,000, >$40,000–50,000, >$50,000–75,000, >$75,000–100,000, or >$100,000.

^eThree-level variable categorized as <high school diploma/GED, high school diploma/GED, or >high school diploma/GED.

*Statistically significant at p < .05.

In Table 3, Model 3, psychosocial adversity variables were added to Model 1 (10-level household income and 3-level education variables with demographic covariates). Adverse childhood experiences, perceived discrimination, and neighborhood problems were significantly associated with increased odds of diabetes/high depression symptom comorbidity versus having no diabetes/low depression symptoms. Each SD increase in adverse childhood experiences (SD = 2.34 adverse experiences) was associated with 41% higher odds (OR = 1.41; 95% CI: 1.16, 1.71) of diabetes/high depression symptoms. One-SD increases in perceived discrimination and neighborhood problems were associated with 23% (OR = 1.23; 95% CI: 1.01, 1.50) and 53% (OR = 1.53; 95% CI: 1.30, 1.80) higher odds for diabetes/high depression symptom comorbidity, respectively. Adverse childhood experiences, discrimination, and neighborhood problems were also associated with increased odds of depression symptoms without diabetes (no diabetes/high depression symptoms group; see Table 3), relative to being in the no diabetes/low depression symptom group. Lifetime trauma history was significantly associated with increased odds of having no diabetes/high depression symptoms (OR = 1.17; 95% CI: 1.03, 1.34) but not comorbidity, relative to having no diabetes/low depression symptoms. When the 72 cases where diabetes diagnosis was determined by self-report alone (e.g., not additionally detected by lab values or medication use) were excluded from models, no changes in results were observed.

When models were repeated using the diabetes/high depression symptom (i.e., cormorbid) group as the reference group (Table 4), household income <$30,000/year, but not low education, was significantly associated with lower odds of no diabetes/high depression symptoms (OR = 0.59; 95% CI: 0.37 – 0.94), diabetes/low depression symptoms (OR = 0.27; 95% CI: 0.17, 0.43), and no diabetes/low depression symptoms (OR = 0.31; 95% CI: 0.20, 0.50) compared to the comorbid group. Adverse childhood experiences were associated with lower odds of having diabetes/low depression symptoms (OR = 0.72; 95% CI: 0.58, 0.89) and neither condition (OR = 0.70; CI: 0.58, 0.85) but not having no diabetes/high depression symptoms. The same was true for neighborhood problems (OR: 0.71; 95% CI: 0.59, 0.86: and OR: 0.63; 95% CI: 0.53, 0.75, respectively). Perceived discrimination was not significantly associated with odds of membership in any other group, when compared to the comorbid group.

Discussion

The high prevalence of T2DM in Latinos and the increase in adverse disease outcomes related to comorbid T2DM and depression [4, 6, 23] underscore the need for identification of correlates of that comorbidity in Latinos. Both upstream socioeconomic and proximal psychosocial adversity variables were linked to the comorbidity in our sample. When interpreting results of our analyses, however, careful consideration of the comparison group is critical. The prevalence of each adversity variable examined was higher in individuals with comorbidity compared to individuals with neither condition or with diabetes and low depression symptoms. Lower household income (<$30,000/year), adverse childhood experiences, perceived ethnic/racial discrimination, and perceived neighborhood problems were associated with increased odds of comorbidity compared to those with neither condition. Low household income was the only variable that distinguished the diabetes/high depression symptom group from all other groups in analyses that used this comorbid group as the reference. Given the known medical and financial burden and established increase in mortality risk conferred by comorbidity of diabetes and depression [4, 6, 23], these results are clinically meaningful and have implications for both clinical and public health practice.

Lower household income was associated with the highest risk for comorbidity (OR = 4.61; 95% CI: 2.89, 7.33) compared to odds of neither condition and was uniquely associated with lower odds of membership in all other groups compared to the comorbid group. The income-related findings align with a large body of quantitative evidence documenting the health risks, including increased risk for both diabetes and psychological depression, associated with low SES [28–30, 52]. The magnitude of the OR, even after adjustment for health insurance coverage and other socio-demographic variables, supports the profound influence of income on health. As widely reported, low socioeconomic status affects health in a multifactorial fashion, through differences in access to healthful foods, screening and prevention services, safe neighborhoods, and stable housing, as well as through lifestyle factors, physiological stress pathways, and psychosocial risk and protective factors [53, 54]. Given that poverty increases risk for exposure to violence and trauma exposure and is associated with poorer mental health, poorer nutrition, and worse healthcare access, it may be considered as one of the drivers of the systemic vulnerability leading to the depression and diabetes syndemic.

The association of adverse childhood experiences with higher odds of comorbidity compared to odds of having neither condition—but not with lower odds of being in the no diabetes/high depression group compared to the comorbid group—indicates that the association of adverse childhood experiences and comorbidity may be largely driven by the association of adverse childhood experiences with depression. These results add to a growing body of literature on linkages between childhood adversity and adult mental health outcomes [44, 55–58]. Adverse childhood experiences were prevalent in this sample; the mean number of events experienced was 2.3, and 28.4% of participants reported four or more events [59]. In a recent systematic review of 37 studies internationally that measured adverse childhood events (N = 253,719), only 13% of the meta-analytic sample reported four or more events [60]. In a prior publication on adverse childhood event prevalence in the HCHS/SOL SCAS, the events most often reported by participants were parental separation or divorce, emotional/psychological abuse, physical abuse, household drug/alcohol abuse, and having a household member in prison [59]. Though no items on the ACE scale ask about migration, the high prevalence of migration experiences in this sample (78% emigrated to the USA from another country) suggests that immigration-related family disruptions may be an important, unmeasured driver of early life adversity.

Perceived ethnic/racial discrimination was prevalent in this sample with 79.5% reporting exposure to ethnicity-related discrimination [61]. While associated with increased odds of comorbidity compared to neither condition, discrimination was not associated with lower odds of membership in other groups compared to the comorbid reference group. Therefore, it was not an adversity variable that uniquely distinguished comorbid group membership. Prevalence of perceived discrimination in this sample was similar to other samples of community-dwelling U.S. Latinos, including in the validation sample of the PEDQ-CV [62–64], and slightly lower than in U.S. black individuals [62, 63]. Substantive prior research has demonstrated linkages between perceived ethnic/racial discrimination and a variety of poor health outcomes, with generally larger effect sizes for mental health outcomes [32, 33]. In Latino individuals with diabetes, more frequent perceived discrimination has been related to increased depression symptoms [65, 66] and higher diabetes-related distress [66]. Though not uniquely associated with comorbid depression symptoms and diabetes, the high prevalence of discrimination in this sample emphasizes the importance of its consideration as a social determinant of health.

In our sample, a one-SD increase in neighborhood problems was associated with a 53% increase in odds for having diabetes and high depression symptoms compared to neither condition. However, similar to findings for adverse childhood experiences, there was no significant difference in odds of being in the diabetes/high depression group compared to the comorbid group, indicating that the association of neighborhood problems and comorbidity may be fueled by the association of neighborhood problems with depression. Neighborhood problems tend to cluster in disadvantaged neighborhoods [67] and may impact health through multiple systems and pathways—directly, as with pollution and violence, and indirectly through their influences on physical activity, dietary intake, and psychological stress and mood. Low-resource neighborhoods often have less structural support for health and safety (e.g., poorer public infrastructure, less or less effective police presence, fewer mental health services) and thus may contribute to systemic and individual vulnerability of residents. Low SES neighborhoods present a prime opportunity for multilevel interventions that address both structural (e.g., improvement of sidewalks and green spaces, violence reduction) and individual (e.g., increased health care connection and outreach services) interventions to reduce the additive health risks of psychological stress and neighborhood problems in Latino communities.

The finding that higher lifetime trauma history was not associated with increased odds of diabetes/depression symptom comorbidity in this sample was unexpected, given the documented relationship between posttraumatic stress and T2DM [68, 69]. Similar results were reported in a previous publication on stress and cardiovascular disease risk factors in the HCHS/SOL SCAS [70]: participants who reported a greater number of traumatic events had a lower prevalence of diabetes. Lifetime trauma exposure was prevalent in the HCHS/SOL cohort, with 2.1 traumatic events experienced on average, and 80% of participants reporting at least one traumatic event in their lifetime. This prevalence is higher than rates reported in a national probability sample of the U.S. population (61% among men, and 51% among women) [71] and a large, community-based sample with over 50% Latino participants (57%) [72]. Our measure of traumatic events assessed only event occurrence, however, not appraisal or response, and was not an evaluation of posttraumatic stress. Social and community protective factors, culture-based normative appraisals, and individual appraisal style [73] strongly affect interpretation, coping, and development and maintenance of a stress response [74, 75] and may protect against adverse health sequalae. Importantly, the trauma scale used does not capture major stressors and traumatic events specific to immigration, such as traumatic events during the immigration journey or associated with legal status (e.g., deportation of a parent or relative) that are associated with depression among Latinos not born in the USA. [76, 77]. Without consideration of experiences specific to immigration, traumatic event measures could be a less useful correlate of diabetes and depression symptoms outcomes.

Interestingly, none of the adversity variables associated with comorbidity were significantly associated with diabetes/low depression symptoms group, while all were associated with no diabetes/high depression symptoms. While no conclusions can be made about directionality or etiologic mechanisms of disease development from these cross-sectional results, this pattern of findings may suggest that depression symptoms could be a pathway through which social adversity increases risk for the diabetes-depression syndemic, given the known contribution of depression to risk for diabetes [20, 78, 79]. Examining CIs of adversity variable ORs between the no diabetes/high depression symptoms and the comorbid group also shows that the CIs do in fact overlap, indicating lack of statistically significant difference. However, the magnitudes of ORs, which are substantially larger for comorbidity than for no diabetes/high depression symptoms, especially for low household income (OR = 4.61 vs. 2.66), suggest possible involvement beyond greater depression symptoms alone.

Our findings should be interpreted in light of some limitations. Given the cross-sectional design, no conclusions can be made regarding temporality or causality in these relationships. The adversity variables examined are not comprehensive and represent only a subset of risk factors that may be related to a potential diabetes-depression syndemic. Adversity factors were self-reported, and in the case of lifetime trauma and adverse experiences in childhood, reporting was retrospective. Thus, these variables are subject to biases inherent in all self-reported measures (e.g., social desirability) and in measures of past experience (e.g., recall bias). Perceptions of experiences and environmental surroundings, as assessed in our perceived discrimination and neighborhood problem scales, can be influenced by psychological well-being and mood. Those with higher depression symptoms in our sample may have reported more discrimination and neighborhood problems due to the negative cognitive-emotional biases known to accompany depression. The brief depression screening measure we employed generates scores that indicate likely presence or absence of clinically significant depression but are not equivalent to a formal diagnosis of major depression. Finally, there is no widespread agreement about how to best analyze possible syndemic relationships in an epidemiologic data set. Our analysis represents an initial attempt to examine syndemic correlates in a large, diverse Latino cohort but will need to be confirmed and expanded upon in future studies.

In conclusion, the current study provides epidemiological evidence of the correlations between low SES, early life interpersonal adversity, discrimination experiences, and neighborhood disadvantage and the burdensome and costly syndemic of diabetes and depression in Latino individuals. Longitudinal studies are needed to examine the contributions of these and other adversity experiences to diabetes and depression comorbidity over time. Our results suggest that interventions on both individual and neighborhood levels, such as integrated care approaches implemented in community-based clinics and targeted improvement of neighborhood characteristics (e.g., safety, healthy food availability) may reduce syndemic-related health inequities throughout the lifespan.

Acknowledgements

Funding The Hispanic Community Health Study/Study of Latinos was supported by contracts from the National Heart, Lung, and Blood Institute to the University of North Carolina (N01-HC65233), University of Miami (N01-HC65234), Albert Einstein College of Medicine (N01-HC65235), Northwestern University (N01-HC65236), and San Diego State University (N01-HC65237). The following institutes/centers/offices contribute to the HCHS/SOL through a transfer of funds to the National Heart, Lung, and Blood Institute: the National Center on Minority Health and Health Disparities, the National Institute of Deafness and Other Communications Disorders, the National Institute of Dental and Craniofacial Research, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Neurological Disorders and Stroke, and the Office of Dietary Supplements. The Hispanic Community Health Study/Study of Latinos Sociocultural Ancillary Study was supported by Grant 1 RC2 HL101649 from the National Institutes of Health/National Heart, Lung, and Blood Institute (Gallo/Penedo PIs). J. L. McCurley was additionally supported by a National Institutes of Health T32 training grant in Cardiovascular Epidemiology from the National Heart, Lung, and Blood Institute and University of California San Diego (5T32HL079891–06) and a GloCal Health Fellowship funded by the Fogarty International Center, National Heart, Lung, and Blood Institute, and the University of California Global Health Institute (R25 TW009343).

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Jessica L. McCurley, Angela P. Gutierrez, Julia I. Bravin, Neil Schneiderman, Samantha A. Reina, Tasneem Khambaty, Sheila F. Castañeda, Sylvia Smoller, Martha L. Daviglus, Matthew J. O’Brien, Mercedes R. Carnethon, Carmen R. Isasi, Krista M. Perreira, Greg A. Talavera, Mingan Yang, Linda C. Gallo declare that they have no conflict of interest.

Authors’ contributions Neil Schneiderman, Sylvia Smoller, Martha L. Daviglus, Mercedes R. Carnethon, Carmen R. Isasi, Krista M. Perreira, Greg A. Talavera, Linda C. Gallo contributed to study conception, design, and implementation, data acquisition, and manuscript writing. Jessica L. McCurley, Greg A. Talavera, Mingan Yang, and Linda C. Gallo contributed to data analyses and manuscript writing. Angela P. Gutierrez, Julia I. Bravin, Samantha A. Reina, Tasneem Khambaty, Sheila F. Castaneda, and Matthew J. O’Brien provided critical guidance in interpretations of results and manuscript development. All authors read and approved the final manuscript. All authors agree to be accountable for their contributions, the accuracy, and integrity of the research.

Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The study was approved by the Institutional Review Boards at all HCHS/SOL institutions.

Informed Consent Informed consent was obtained from all individual participants included in the study.