Psychosocial factors in the relationship between socioeconomic status and cardiometabolic risk: the HCHS/SOL Sociocultural Ancillary Study

Abstract

Background

U.S. Hispanics/Latinos display a high prevalence of metabolic syndrome (MetSyn), a group of co-occurring cardiometabolic risk factors (abdominal obesity, impaired fasting glucose, dyslipidemia, elevated blood pressure) associated with higher cardiovascular disease and mortality risk. Low socioeconomic status (SES) is associated with higher risk for MetSyn in Hispanics/Latinos, and psychosocial factors may play a role in this relationship.

Purpose

This cross-sectional study examined psychosocial factors in the association of SES and MetSyn components in 4,996 Hispanic/Latino adults from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) Sociocultural Ancillary Study.

Methods

MetSyn components were measured at the baseline examination. Participants completed interviews to determine psychosocial risks (e.g., depression) and resources (e.g., social support) within 9 months of baseline (< 4 months in 72.6% of participants). Confirmatory factor analysis (CFA) and structural equation modeling (SEM) were used to identify latent constructs and examine associations.

Results

Participant mean age was 41.7 years (SE = 0.4) and 62.7% were female. CFA identified single latent factors for SES and psychosocial indicators, and three factors for MetSyn [blood pressure, lipids, metabolic factors]. SEMs showed that lower SES was related to MetSyn factors indirectly through higher psychosocial risk/lower resources (Y-Bχ² (df = 420) = 4412.90, p < .05, RMSEA = .042, SRMR = .051). A statistically significant effect consistent with mediation was found from lower SES to higher metabolic risk (glucose/waist circumference) via psychosocial risk/resource variables (Mackinnon's 95% asymmetric CI = −0.13 to −0.02).

Conclusions

SES is related to metabolic variables indirectly through psychosocial factors in U.S. Hispanics/Latinos of diverse ancestries.

Introduction

Hispanics/Latinos (hereafter referred to as Hispanics), the largest U.S. ethnic/racial minority group (1), display high rates of metabolic syndrome (MetSyn) (2), a co-occurring set of cardiometabolic risk factors that includes elevated blood pressure, dyslipidemia, hyperglycemia, and abdominal obesity (3, 4) and that predicts risk for diabetes (5), cardiovascular disease (CVD)(6), and all-cause mortality (3). The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) reported an overall MetSyn prevalence of 34% in men and 36% in women in a large, multi-site cohort of over 16,000 Hispanic individuals, with rates climbing to 55% and 62% in men and women aged 65 to 74, respectively (7). Given the rapid population growth of the U.S. Hispanic population, which is expected to double in size by 2060 (8), improved understanding of factors associated with MetSyn prevalence is critical to reduce CVD burden and associated healthcare costs.

Low socioeconomic status (SES) has been shown to be associated with higher MetSyn risk in multiple racial/ethnic groups, including Hispanics (9-12). SES gradients in indicators of cardiometabolic risk have been variable in studies of Hispanics, however, with some studies reporting flattened or inconsistent gradients in individuals of different heritage groups or in men (11, 13-15). Factors underlying the SES-metabolic risk association are complex and may include differences in access to screening and preventative interventions, lifestyle factors, physiological stress pathways, and psychosocial risk and protective factors (16, 17). Research exploring connections between SES, MetSyn, and psychosocial factors is limited and represents an important avenue for investigation that may inform early interventions in the connection between low SES and elevated chronic disease risk (18).

According to the Reserve Capacity Model (19-21), three inter-related psychosocial pathways may contribute to associations between lower SES and worse physical health: 1) Elevated exposure to harmful or threatening situations; 2) increased negative cognitive-emotional experiences; and 3) a smaller bank of resilient social and personal resources with which to manage stressors, threats, and demands. Studies that have used the reserve capacity model as a framework for understanding SES gradients in cardiometabolic risk have found that negative cognitive-emotional factors (e.g., depression, anger, anxiety, loneliness) and psychosocial resources (e.g., self-esteem, mastery, optimism) contributed to associations of SES with MetSyn (15, 22), abdominal obesity (22), and poorer blood pressure nocturnal dipping in middle-aged Mexican American women (12). Similarly, a prospective study of predominantly non-Hispanic white women showed that lower SES predicted higher MetSyn risk across 12 years - both directly, and indirectly through a path from low SES to low psychosocial resources/high negative emotions and cognitions to increased risk of incident MetSyn (21). Other studies have found evidence for the influence of social and personal resources (e.g., social support, social integration, perceived control) in connecting SES with physical health outcomes including stroke (23), coronary heart disease (24), and all-cause mortality (25) [for a review see (26)]. In sum, the research base suggests that psychosocial resources and negative cognitive emotional factors may explain or relate in important ways to SES related cardiometabolic risk disparities (26, 27).

The current study examined the contributions of multiple psychosocial risk (i.e., negative cognitive-emotional factors) and resource factors (i.e., intrapersonal resources such as self-esteem; interpersonal resources such as social support) to associations between SES and cardiometabolic risk, defined using the components of MetSyn, in participants from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) and HCHS/SOL Sociocultural Ancillary Study. We hypothesized that lower SES (education, income) would be associated with higher levels on MetSyn factor(s), and that lower SES would be associated with higher psychosocial risk and lower resource factors. In addition, we predicted indirect effects whereby SES would relate to cardiometabolic risk in part or fully through its associations with psychosocial variables. Given the unique, heterogeneous Hispanic sample and the limited research to date examining psychosocial underpinnings of SES-cardiometabolic health disparities, this research builds significantly on the existing research base, which in Hispanic groups is comprised primarily of studies in women and Mexican Americans. By focusing on potentially modifiable, health-relevant psychosocial variables (14), the current study may highlight opportunities for primary prevention of cardiometabolic disorders in the high-risk U.S. Hispanic population.

Methods

Participants

HCHS/SOL is a large, community-based epidemiological cohort study of chronic disease prevalence, incidence, and risk and protective factors in 16,415 Hispanic adults (self-identified as Mexican, Cuban, Puerto Rican, Dominican, Central and South American, or other/more than one Hispanic heritage group) aged 18–74 years at enrollment. The HCHS/SOL methodology (28) and sampling design (29) have been presented in detail elsewhere. In brief, participants were recruited in four U.S. field centers (Bronx, NY; Chicago, IL; Miami, FL; San Diego, CA) between 2008 and 2011 using a two-stage area household probability sampling design, with oversampling in the 45- to 74-year-old age group. A comprehensive baseline clinical examination and interview assessment was conducted in participant's language of preference. Participants were asked to bring all prescribed and over-the-counter medications taken in the last 4 weeks to the baseline exam, where medications were scanned (via Universal Product Code bar codes) or manually transcribed. The study was approved by the institutional review boards at all HCHS/SOL institutions, including the coordinating center and reading centers, and informed consent was obtained from all individual participants in the study. Study methods were performed in accordance with the ethical standards as outlined in the 1964 Declaration of Helsinki and its later amendments.

This quasi-cross-sectional analysis was conducted in participants from the HCHS/SOL-Sociocultural Ancillary Study (SCAS; N=5313), who completed self-report assessments of psychosocial and sociocultural factors via an in-person interview within 9 months of their HCHS/SOL baseline clinical examination. The term quasi-cross-sectional is utilized to reflect that, similar to the Multi-Ethnic Study of Atherosclerosis (MESA) (30) and other large cohort studies, time-lagged data collection was conducted (e.g., (31)), with some data collected at the primary baseline clinical visit, and other data collected at a separate ancillary study visit. The majority of SCAS participants (72.6%) completed their interview within 4 months of the baseline clinical exam. The overall length of time for the SCAS interview was 1-2 hours, however this included additional variables not a part of the current analysis. A full description of the SCAS methods and sample has been presented (32). Participants in the SCAS sample are representative of the HCHS/SOL cohort, with the exception of lower participation of some higher SES individuals (32). Individuals with history of coronary heart disease (n=234), stroke (n=57), both conditions (n=22), or those missing these variables (n=4) were excluded from the analysis, because MetSyn is generally viewed as a useful intermediate stage in predicting these outcomes. The final analytic sample included 4996 individuals.

Measures

Socioeconomic Status

Self-reported educational attainment (3 categories: less than high school/GED, high school diploma/GED, greater than high school diploma/GED) and household income (10 categories ranging from <$10,000/year to >$100,000/year) were the observed variables used to represent SES.

Psychosocial Resource and Risk Factors

Psychosocial risk and resource variables were selected for use in the SCAS based on previous research that has supported their connection to SES, and health and well-being. Descriptions of each measure, including number of items, response scale, and reliability statistics (Cronbach's alpha) for English and Spanish versions, is presented in Appendix Table 1. Indicators of social resources included measures of perceived social support [the Interpersonal Support Evaluation List, 12-item version (33)], social integration/social network diversity [e.g., number of social roles in which the individual has contact at least every 2 weeks; Cohen's Social Network Index (34)], and family cohesion [Family Environment Scale (35)]. Indicators of psychological resources included measures of purpose in life [Life Engagement Test (36)], self-esteem [Rosenberg self-esteem scale (37)], and optimism (versus pessimism) [Life Orientation Test, Revised (38)]. Indicators of psychological risk included measures of depressive symptoms [Center for Epidemiological Studies Scale for Depression (39)], anxiety [Spielberger State-Trait Anxiety Scale (40)], anger [Spielberger State-Trait Anger Scale (40)], cynical hostility [Cook-Medley Hostility Scale, Cynicism subscale (41)], loneliness [revised UCLA Loneliness Scale (42)], and hopelessness [Hopelessness Scale (43)]. Specific measures were chosen based on prior use and evidence of validity and reliability in Hispanic and Spanish-speaking populations. When analyses of internal consistency were conducted for the SCAS sample, adequate reliability was found for all scales in both English (α = .75 to .97) and Spanish (α = .72 to .94) versions. See Appendix Table 1 for internal consistency reliability of individual measures.

Cardiometabolic Risk

The variables that comprise the metabolic syndrome according to the widely used, unified American Heart Association/National Heart, Lung, and Blood Institute (AHA/NHLBI) definition (3) were used to indicate cardiometabolic risk. These indicators included systolic and diastolic blood pressure, fasting glucose, triglycerides, high-density lipoprotein cholesterol (HDL-C), and waist circumference. Variables were measured using a standardized protocol during the HCHS/SOL baseline clinical exam. Systolic and diastolic blood pressure were measured three times at 1-minute intervals, after the participant had rested 5 minutes in the seated position, using an automatic sphygmomanometer (Omron model HEM-907 XL, Omron Healthcare Inc., Bannockburn, IL). The average of the three measurements was used in this analysis. Waist circumference was measured at the uppermost lateral border of the right ilium. A magnesium/dextran sulfate method was used to measure HDL-C, and fasting plasma glucose was measured using a hexokinase enzymatic method (Roche Diagnostics, Indianapolis, IN). Triglycerides were measured using a Roche Modular P chemistry analyzer via a glycerol blanking enzymatic method (Roche Diagnostics). All testing protocols and assay methods are described in HCHS/SOL Manual 7 (at http://www.cscc.unc.edu/hchs/).

Rather than using cut scores to define high and low risk on these variables, MetSyn risk factors were represented continuously in confirmatory factors analysis (CFA) to identify a latent factor model of MetSyn (see Analytic Plan below). This approach accounts for the fact that differences may exist in distributions of body weight and waist circumference across sociodemographic (e.g., ethnic, heritage) groups, so that current cut points may not optimally capture risk for ethnically or racially diverse groups (44-46). In addition, the relationship between MetSyn component values and CVD and mortality risk may differ across populations (6, 47, 48). This continuous variable approach was also considered a preferable approach to characterize relationships of SES and psychosocial variables with cardiometabolic factors across a gradient of risk levels, rather than only for those meeting certain cut scores. In addition, this approach is consistent with prior studies that used the Reserve Capacity Model as a framework for understanding psychosocial factors in SES gradients of cardiometabolic risk (15, 21), which fosters the ability to compare findings across studies.

Covariates

Primary analyses controlled for demographic characteristics (age, sex, language preference, nativity/immigration status, Hispanic heritage group, field center). Medications relating to metabolic outcomes were not included as controls. Their influence is difficult to capture accurately due to inability to account for specific dosage, length of use, and adherence. Instead, sensitivity analyses were conducted, repeating SEMs after the removal of 1,483 participants who were on at least one of the following medications: lipid lowering drugs/antihyperlipidemics, fibric/nicotinic acids, antidiabetics, and antihypertensives. Rather than potential confounders, dietary intake and physical activity are viewed as mechanisms through which psychosocial variables may contribute to cardiometabolic health outcomes. Thus, these factors were not included as covariates in primary models, but were adjusted for in secondary analyses.

Analytic Plan

Descriptive analyses were performed using PASW Statistics 18.0 (SPSS, Inc., Chicago IL). Model assumptions regarding linearity, normality, independence, and homoscedasticity of errors were assessed graphically and analytically. All variables were examined on a continuous scale to maximize statistical power. All analyses accounted appropriately for design effects (stratification and clustering census block groups) and sample weights (29).

Confirmatory factor analyses (CFAs) were conducted to evaluate the measurement models for latent variables representing SES (income and education), psychosocial risks and resources (e.g., negative emotions, interpersonal and intrapersonal resources), and MetSyn. Factor loadings and model fit statistics were examined to determine if a given indicator should be excluded, or if one or more indicators represented multiple constructs. For the psychosocial variables, three models were tested based on theoretical and conceptual pathways by which psychosocial variables may influence cardiometabolic health: 1) a single-factor model with all variables as indicators, representing the general shared variance in psychosocial components; 2) a two-factor model allowing risk and resource indicators to create separate latent factors; and 3) a three-factor solution with social resources, psychological resources, and psychological risk variables comprising the separate latent factors. Based on prior research (7, 49), three MetSyn models were tested: 1) a single-factor model with systolic and diastolic blood pressure (SBP, DBP), triglycerides, HDL-C, glucose, and waist circumference as indicators; 2) a two-factor model with blood pressure (SBP, DBP), and triglycerides, glucose, HDL-C, and waist circumference loading on separate factors; and 3) a three-factor model, with blood pressure (SBP, DBP), lipids (triglycerides, HDL-C), and metabolic variables (glucose, waist circumference) representing separate factors. Model testing for MetSyn was conducted solely to determine the optimal representation for subsequent structural equation modeling. Prior research has examined the factor structure of MetSyn (49) and the prevalence of MetSyn and its correlates (7) in the full HCHS/SOL cohort.

Subsequently, two competing structural equation models (SEMs) were conducted to examine the proposed pathways. Model 1 tested an indirect model with SES relating to MetSyn factor/s via the psychosocial resource and risk factor (Figure 1). Model 2 added direct effects from SES to MetSyn factors. Model fit statistics were examined to determine if Model 2 demonstrated improved descriptive and statistical fit over the indirect effects model. MPlus (50), which employs a maximum likelihood robust (MLR) estimation procedure, was used to estimate model parameters. This procedure provides a chi-square test statistic [Yuan-Bentler T₂ (Y-Bχ²; (51)] and standard errors that are adjusted for multivariate non-normality and missing data. Two recommended descriptive fit indexes (52), the root mean square error of approximation (RMSEA) and the standardized root mean square residual (SRMR), were used to evaluate model fit. Both parameters represent descriptive indices of overall model fit, with values .05 or less indicative of good fit and .08 or less indicative of acceptable fit. The two competing (nested) models were compared statistically using the chi-square difference test (ΔY-Bχ²) and descriptively using the RMSEA and SRMR. Secondary analyses included adjustment for behavioral factors (dietary intake and physical activity) potentially involved in the pathway from SES and psychosocial factors to MetSyn. Sensitivity analyses were conducted by repeating the SEMs after removing the 1,483 participants who were on lipid lowering, fibric/nicotinic acid, glucose lowering, or blood pressure lowering medications. In addition, exploratory analyses were conducted repeating the best fitting model after stratification of the full sample by Hispanic heritage group to assess if associations and model fit were consistent across these groups.

Figure 1. Results of SEM analysis testing the indirect pathways from SES to cardiometabolic factors via Psychosocial Resources/Risk (Model 1).

Standardized regression coefficients (SE) from the structural equations model representing the indirect pathways (SE) from SES to blood pressure (systolic and diastolic blood pressure), lipids (high-density lipoprotein cholesterol and serum triglycerides), and metabolic risk factors (fasting glucose and waist circumference) via the (higher) psychosocial resources/ (lower) risk latent variable. *p < .001. Analyses controlled for age, sex, language preference, nativity/immigration status, Hispanic heritage group, and field center.

Results

Demographic Characteristics

Descriptive statistics for all study variables are shown in Tables 1 and 2. Participants in the sample were between 18 and 74 years (M = 41.7, SE = 0.4) and 62.7% were female. The majority (82.4%) of the sample was born outside of the U.S. mainland, and individuals with Mexican heritage were the largest Hispanic group (1988 individuals; 39.8% of sample). Yearly household income was below $30,000 for the majority of the sample (72.1%), and 35.0% percent had less than a high school education/GED. Most individuals preferred to complete their interview in Spanish (80.8%).

Table 1.

Sample characteristics: HCHS/SOL Sociocultural Ancillary Study (N= 4996)

Characteristic	Weighted M	SE
Age (N = 4996)	41.7	0.4
	N	%	Weighted % (95% CI)
Sex (N = 4996)
Female	3134	62.7	55.2 (53.2-57.2)
Male	1862	37.3	44.8 (42.8-46.8)
Hispanic Heritage (N = 4994)
Central American	530	10.6	7.6 (6.2-9.3)
Cuban	727	14.6	20.0 (16.0-24.7)
Dominican	503	10.1	11.8 (9.9-13.9)
Mexican	1988	39.8	37.4 (33.4-41.5)
Puerto Rican	784	15.7	15.1 (13.2-17.3)
South American	332	6.6	4.7 (3.9-5.7)
More than one/Other	130	2.6	3.4 (2.5-4.6)
Household Yearly Income (N = 4583)
<$30,000	3275	72.1	69.4 (66.5-72.1)
>$30,000	1308	28.5	30.6 (27.9-33.5)
Education Completed (N = 4893)
< HS diploma or GED	1751	35.0	31.5 (29.3-33.8)
HS diploma or GED	1305	26.1	28.1 (26.4-30.0)
>HS diploma or GED	1837	36.8	38.7 (36.2-41.3)
Nativity/Immigration Status (N = 4987)
Born in the US Mainland	875	17.3	22.4 (20.0-24.9)
Immigrated ≥ 10 years ago	2899	58.1	49.6 (47.1-52.1)
Immigrated ≤ 10 years ago	1213	24.3	28.0 (25.2-31.1)
Preferred Interview Language (N = 4996)
English	959	19.2	24.8 (22.1-27.8)
Spanish	4037	80.8	75.2 (72.2-77.9)

CI = Confidence Interval; GED = General Education Development/General Equivalency Degree; HS = High School; M = Mean; SD = Standard Deviation

Table 2.

Descriptive statistics, factor loadings, and associations among latent constructs for SES, Psychosocial Resource/Risk (R/R) and Metabolic Syndrome Indicators; N = 4996.

		Factor Loadings
Variables	Mean1 (SD)	SES	Psychosocial R/R	Blood Pressure	Lipids	Metabolic Factors
Income		.50
Education		.48
Social Support	24.23 (21.77)		.59
Social Integration	3.12 (29.76)		.31
Family Cohesion	50.24 (34.86)		.51
Self-Esteem	29.16 (22.39)		.62
Optimism	11.88 (14.05)		.57
Life Engagement	25.18 (4.87)		.58
Depression	7.81 (9.29)		−.73
Anxiety	17.74 (7.63)		−.83
Anger	17.06 (7.99)		−.53
Cynical Hostility	8.17 (5.85)		−.41
Loneliness	1.49 (0.83)		−.68
Hopelessness	2.57 (1.82)		.51
SBP (mmHg)	119.52 (29.21)			.71
DBP (mmHg)	72.19 (18.26)			.99
Triglycerides (mg/dL)	132.29 (153.70)				−.51
HDL-C (mg/dL)	48.79 (19.56)				.59
Glucose (mg/dL)	99.08 (34.93)					.32
Waist Circumference (cm)	97.19 (30.07)					.59

	Latent Factor Correlations
	SES	Psychosocial R/R	Blood Pressure	Lipids	Metabolic Factors
SES	__
Psychosocial R/R	.57	__
Blood Pressure	−.11	−.05	__
Lipids	−.17	.002	.15	__
Metabolic Factors	−.23	−.20	.51	.70	__

Table Notes:

¹Weighted mean; CES-D-10 = Center for Epidemiological Studies – 10 item Depression Scale; DBP = Diastolic Blood Pressure; FACIT-SP = Functional Assessment of Chronic Illness Therapy-Spirituality Scale; HDL-C = High-Density Lipoprotein Cholesterol; ISEL-12 = Interpersonal Support Evaluation List-12; MetSyn = Metabolic Syndrome; R/R = Psychosocial Resources/Risks; SBP = Systolic Blood Pressure; SES = Socioeconomic Status; SNI = Social Network Index; STAI = State Trait Anxiety Inventory; STAS = State Trait Anger Scale.

Latent Factors: Blood Pressure, SBP and DBP; Lipids, HDL-C and triglycerides; Metabolic Factors, glucose and waist circumference. Statistically significant correlations (p<.05) are bolded.

Confirmatory Factor Analyses (CFA)

CFA models were first fit separately for A) social resources, psychological resources, and psychological risk variables (3-factor model), and B) psychosocial risk and psychosocial resource variables (2-factor model). Although the 3- and 2-factor models fit reasonably well (3-factor, Y-Bχ² (df = 32) = 323.50, p < .05, RMSEA = .041, SRMR = .036) and (2 factor, Y-Bχ² (df = 53) = 1056.70, p < .05, RMSEA = .062, SRMR = .057), the interfactor correlations between all resulting latent variables was of large magnitude, suggesting redundancy among the factors. In the three factor model, inter-factor correlations ranged from |.67 to .78|. In the two-factor model, the correlation between risk and resource latent factors was r = −.79). Thus, a single latent variable representing combined psychosocial resources/risk was specified and tested. This model also showed acceptable model fit (Y-Bχ² (df = 54) = 1512.71, p < .05, RMSEA = .068, SRMR = .059). Standardized factor loadings, shown in Table 2, were all statistically significant and ranged from |.31 to .83|. Given adequate fit of the one factor model, and because high collinearity among derived latent factors impedes estimation of differential contributions of the factors to an outcome, the more parsimonious single factor solution was utilized in subsequent analyses. This latent variable conceptually represents the aggregate, shared variance associated with psychosocial risk and resource variables, with higher factor values reflecting greater psychosocial resources and lesser psychosocial risks.

A latent variable representing SES was then tested within the context of this 1-factor psychosocial resources/risk model. This model also fit reasonably well, (Y-Bχ² (df = 74) = 1270.74, p < .05, RMSEA = .055, SRMR = .057). Standardized factor loadings for the income and education indicators of the SES latent variable were statistically significant (.48, .50, respectively).

In model testing for MetSyn variables, the three-factor solution produced the best fit (Y-Bχ²(df = 6) = 64.13, p < .05, RMSEA = .043, SRMR = .036), with blood pressure (systolic, diastolic), lipids (HDL-C, triglycerides) and metabolic risk factors (waist circumference, glucose) comprising three separate latent variables. The standardized factor loadings were all statistically significant (ranged from |.32-.99|).

Correlations among derived factors are shown in Table 2. Higher SES exhibited a strong positive correlation with higher psychosocial resources/lower risk, and a weak negative correlation with two MetSyn latent factors: lipids and metabolic factors (glucose, waist circumference). Higher psychosocial resources/lower risk was weakly negatively associated with blood pressure and metabolic factors. Interfactor correlations among the three MetSyn factors were significant and ranged from r = .15 to .70.

Structural Equation Models

Tests of the two SEMs showed that Model 1 [indirect effects only from the SES latent variable to the psychosocial resources/risk latent variable to the three MetSyn latent variables; YBχ²(df = 420) = 4412.90, p < .05, RMSEA = .042, SRMR = .051] and Model 2 [indirect effects as defined above, and direct effects from the SES latent variable to the 3 MetSyn latent variables; Y-Bχ²(df = 417) = 4374.73, p < .05, RMSEA = .042, SRMR = .051] both fit well descriptively. Although Models 1 and 2 demonstrated a statistical difference (ΔY-Bχ²(df = 3) = 51.28, p < .001), there were no differences in model fit based on the descriptive fit indices (ΔRMSEA < .001, ΔSRMR < .001). Thus, emphasizing descriptive fit, the more parsimonious Model 1 was determined to be the “better-fitting” model.

Standardized regression coefficients for Model 1 are presented in Figure 1. Higher SES was significantly related to greater psychosocial resources/lesser risk, which in turn was associated with lower values for the metabolic latent factor (glucose/waist circumference). MacKinnon's asymmetric confidence interval (53) was calculated and showed a statistically significant mediated effect from SES to the metabolic latent factor via the psychosocial resource/risk factor (95% asymmetric CI = −0.13 to −0.04).

When the above models were repeated controlling for dietary intake and physical activity, no substantive changes in results were observed. Tests of the indirect effects only and direct effects added SEMs showed no differences in model fit based on the descriptive fit indices (ΔRMSEA < .001, ΔSRMR < .001), so again the more parsimonious Model 1 - indirect effects only - was determined to be the “better-fitting” model. The mediated effect from SES to the metabolic latent factor via the psychosocial resource/risk factor remained significant (95% asymmetric CI = −0.13 to −0.04). Furthermore, when participants on lipid lowering/antihyperlipidemics, fibric/nicotinic acids, antidiabetic, and antihypertensive medications were removed from the analytic sample, findings were consistent with the primary analyses (results not shown).

As Model 1 was identified as the best-fitting model in all above analyses, this model was repeated after stratification by Hispanic heritage group. Results of these stratified analyses are presented in Table 3. The significant indirect effect from SES to metabolic factors (waist circumference/glucose) observed in the overall sample was replicated in Dominican, Central American, and Cuban groups. In individuals of Mexican heritage, an indirect effect was found in the prediction of the lipids factor, with higher psychosocial resources/lower risk relating to a better lipid profile. In Puerto Rican individuals, there were no significant indirect effects. In South American individuals, the indirect effect from SES to waist circumference/glucose remained significant, but the path from psychosocial resources/risk to waist circumference/glucose was positive instead of negative as in the overall sample. The sample size of this group is very small, however (n = 350), and the group composition is diverse (representing several South American countries), which could compromise validity and stability of this finding. Overall, stratified analyses showed that higher resources/lower risk is generally protective with regards to metabolic syndrome factors, but perhaps less so in individuals of Puerto Rican heritage.

Table 3.

Standardized regression coefficients (SE) from SEM analyses testing indirect pathways from SES to cardiometabolic factors via psychosocial resources/risk (R/R) latent variable, stratified by Hispanic heritage group.

Hispanic Heritage Group	n	R/R on SES β (SE)	p	Blood Pressure on R/R β (SE)	p	Lipids on R/R β (SE)	P	Metabolic Risk Factors1 on R/R β (SE)	p
Central American	549	0.58 (0.09)	<0.01	−0.05 (0.03)	0.17	0.09 (0.08)	0.27	−0.26 (0.10)	<0.01
Cuban	772	0.56 (0.10)	<0.01	−0.02 (0.03)	0.60	0.08 (0.05)	0.07	−0.21 (0.07)	<0.01
Dominican	526	0.35 (0.13)	<0.01	0.06 (0.05)	0.26	−0.05 (0.12)	0.69	−0.31 (0.10)	<0.01
Mexican	2072	0.66 (0.07)	<0.01	−0.03 (0.03)	0.33	0.10 (0.04)	0.02	−0.09 (0.06)	0.01
Puerto Rican	875	0.59 (0.06)	<0.01	0.01 (0.04)	0.07	−0.01 (0.06)	0.92	−0.03 (0.07)	0.71
South American	350	0.50 (0.14)	<0.01	−0.03 (0.07)	0.70	−0.01 (0.09)	0.89	0.19 (0.09)	0.03

Table Notes: SE = Standard Error; SEM= Structural Equation Model; SES = Socioeconomic Status; R/R = Psychosocial Resources/Risk

¹Glucose, Waist Circumference

Discussion

The aim of the current study was to examine the role of psychosocial risk and resource factors in the relationship of SES with MetSyn variables in a large, multi-site sample of Hispanic men and women from several heritage groups. Results showed that higher SES was significantly associated with higher psychosocial resources/lower risk, and significantly correlated with lower risk on two of the three MetSyn components. These findings are consistent with other studies in the HCHS/SOL cohort that have shown inverse relationships between SES and hypertension (28) and diabetes (54) prevalence.

The finding of a single latent factor underlying multiple psychosocial risk/resource variables has empirical precedent in Mexican American women (55, 56) and other samples (57, 58). The single, integrated factor is also consistent with conceptual models suggesting strong interrelationships among cognitive, emotional, and interpersonal domains of psychosocial functioning (59-61). It is clear that psychosocial resource and risk factors do not occur in isolation; those who have stronger social support, optimism, and other resources tend to have lower emotional distress, and vice versa, as these factors are shaped by shared biological, social, and environmental contexts (62-64). In addition, these variables likely influence cardiometabolic health via common autonomic and neuroendocrine regulatory pathways (e.g., the hypothalamic-pituitary-adrenal axis, sympathetic nervous system) (65-67).

In the overall sample, SES was found to relate indirectly to a latent factor capturing fasting glucose (glucose regulation) and waist circumference (abdominal obesity) via psychosocial resources/risk. This finding is consistent with two previous studies of Mexican American women, which found that psychosocial resources partially explained associations of SES with waist circumference (15, 22) and fasting glucose (15), but not other MetSyn components. These factors may be particularly relevant in Hispanics, who experience a disproportionally high prevalence of MetSyn and type 2 diabetes compared to non-Hispanic whites (54). Results were generally consistent when stratified by Hispanic heritage group, with similar findings in Central American, Cuban, and Dominican participants. However, in Hispanics of Mexican heritage, the indirect relationship was significant only for lipids, and not waist circumference/glucose, and no significant indirect effects were observed for Puerto Ricans. These variations underscore the importance of exploring disparities and differences in correlates of chronic disease risk within the large Hispanic pan-ethnic group.

Prior studies conducted among Mexican women in the San Diego area showed that the role of psychosocial factors in the association of SES with MetSyn variables and with nocturnal blood pressure dipping was stronger in less acculturated than in more acculturated individuals (as assessed by language preference) (15). We were unable to examine patterns of acculturation in the current study given marked confounding among field center, heritage group, and language. For example, 96% of participants enrolled in the Miami field center completed the interview in Spanish, and 96% participants in the same field center were of Cuban heritage. We have instead controlled for contextual variables including heritage group, field center, language, and time in the US (born in the US, immigrated >10 years ago, immigrated <10years ago) to better isolate the effects of SES and psychosocial variables.

The current study has several strengths and limitations that should be considered. The inclusion of a multi-site cohort of men and women from varied Hispanic heritage groups allows for improved inference compared to previous studies that have examined associations among SES, psychosocial factors, and MetSyn conducted in women (15, 21) or in Hispanics of Mexican heritage (15) only. However, the sample is not nationally representative (28) and thus findings may not be applicable to all U.S. Hispanics, such as those residing in rural areas. Given the quasi-cross-sectional design, no conclusions can be made about directionality of influence among SES, psychosocial variables, and MetSyn. Longitudinal studies are needed to better assess temporality and parse the differential contributions of SES and psychosocial variables to cardiometabolic processes over time. A limitation of the single unified factor representation of psychosocial variables is the inability to untangle the specific psychosocial pathways through which SES may influence metabolic risk. Further, due to the gap in time between assessment of metabolic syndrome components and psychosocial variables (<4 months for 72.6% of the sample), associations if variables were assessed contemporaneously or in the reverse sequence are unknown. Although years of formal education and household income are commonly used indicators of SES, they may not fully capture all elements of this multi-faceted construct (e.g., employment characteristics, neighborhood SES). Finally, as seen in the standardized path coefficients in Figure 1, all effect sizes were small. This is not surprising when considering that psychosocial factors represent only one component in the multifactorial interplay of genetic, biological, environmental, and behavioral factors that together explain disparities in health and disease (26, 68).

Nonetheless, consideration of psychosocial factors may be warranted in future efforts to reduce cardiometabolic risk among U.S. Hispanics. A large body of evidence supports the potential for psychological interventions to lessen negative emotions such as depression and anxiety, improve coping ability, and decrease the impact of negative emotions on health (e.g., (69-72)). Interventions that aim to decrease psychosocial risk and fortify resources may lessen detrimental effects of low SES environments on health, for example by increasing “reserve capacity” (73). Social factors are amenable to intervention as well. The Community Preventive Services Task Force found sufficient evidence for the health benefits of social support interventions in community settings to place them in the “strongly recommended” category for interventions (74). Such interventions may be particularly impactful in ethnic/racial minority populations who experience marked disadvantages in socioeconomic circumstances and cardiometabolic risk. Future longitudinal studies supporting the roles of psychosocial factors in SES disparities will provide additional impetus for such interventions.

Conclusion

Our findings add to an expanding body of literature supporting the potential roles of psychosocial risk (e.g., depression, anxiety) and resource (e.g., social support, self-esteem) factors in the SES-metabolic health gradient (27) for at-risk populations. Given the high prevalence of MetSyn in U.S. Hispanics, and the large and growing nature of this population, identifying modifiable aspects of the SES-MetSyn relationship is useful in improving understanding of cardiometabolic risk and identifying targets for prevention and management.

Appendix

Appendix Table 1.

Psychosocial resource and risk indicators from the HCHS/SOL Sociocultural Ancillary Study (SCAS) interview.¹

Measure	Description	Time Frame	Items	Response Scale	Cronbach's α for English/Spanish
Psychosocial Resource Indicators
Interpersonal Support Evaluation List (ISEL-12)	Perceived social support and emotional belonging	In general	12	4-point Likert-type scale, Definitely false to Definitely true	E = .86, S = .80
Social Network Integration (SNI)	Number of roles involving social contact at least every 2 weeks	In general	12	Count	N/A
Family Cohesion subscale of the Family Environment Scale (FES)	Family support, teamwork (vs. conflict, criticism)	In general	18	True/False	E = .73, S = .72
Rosenburg Self-Esteem Scale	Personal worth, respect for self	In general	6	4-point Likert-type scale, Strongly disagree to Strongly agree	E = .87, S = .79
Life Orientation Test-Revised (LOT-R)	Optimism (vs. pessimism)	In general	9	5-point Likert-type scale, I disagree a lot to I agree a lot	E = .97, S = .89
Life Engagement Test (LET)	Purpose in life, valued activities	In general	6	5-point Likert-type scale, Strongly disagree to Strongly agree	E = .80, S = .72
Psychosocial Risk Indicators
Center for Epidemiological Studies Scale for Depression (CES-D)	Symptoms of depression (e.g., depressed mood, changes in energy level)	Past week	10	4-point Likert-type scale, Rarely/none to All of the time	E = .82, S = .82
State-Trait Anxiety Inventory (STAI)	Anxiety-related characteristics (e.g., restlessness, worry)	In general	10	4-point Likert-type scale, Almost never to Almost always	E = .92, S = .94
State-Trait Anger Scale (STAS)	Anger-related characteristics (e.g, quick temper, irritability)	In general	10	4-point Likert-type scale, Almost never to Almost always	E = .87, S = .85
Cook Medley Cynicism Scale	Cynicism characteristics (e.g., mistrustful of others’ intentions)	In general	6	True/False	E = .82, S = .77
Revised UCLA Loneliness Scale	Loneliness (e.g. isolation, lacking companionship)	In general	3	3-point Likert-type scale, Hardly ever to Often	E = .78, S = .76
Hopelessness Scale	Hopelessness about goals, the future	In general	2	5-point Likert-type scale, Absolutely disagree to Absolutely agree	N/A

¹Not an exhaustive list of variables assessed in the SCAS interview; only indictors included in the current analysis are listed

N/A = not applicable; UCLA = University of California, Los Angeles.