Unfavorable Social Determinants of Health Are Associated with Lower Cognitive Performance in Non-Hispanic Black Men - The HABS-HD Study

Abstract

Introduction

Non-Hispanic Black (NHB) men experience disproportionate exposure to adverse social determinants of health (SDoH). The cumulative impact on cardiometabolic risk factors, cognitive and neuroimaging outcomes is not well understood.

Methods

We analyzed data from 460 NHB men. Cumulative SDoH index spanning multiple domains was examined both categorically and continuously. Regression models were performed to estimate associations of cumulative SDoH index with cardiometabolic risk factors, cognitive performance, and neuroimaging biomarkers.

Results

Higher cumulative SDoH burden was associated with a greater prevalence of diabetes, tobacco use, and cognitive impairment. Cumulative SDoH was not associated with neuroimaging biomarkers (p>0.05). Greater cumulative SDoH burden showed an independent association with poor memory, executive function, and processing speed (p <0.01)

Discussion

Our findings identified social and structural factors that were linked to cardiometabolic risk and cognitive health among NHB men; however, cumulative SDoH burden was not independently associated with white matter hyperintensities or hippocampal volume.

Introduction:

Social determinants of health (SDoH), including socioeconomic conditions, neighborhood and physical environments, access to important resources, education, and community, are established indicators of health inequities and biological aging [1]. These determinants are associated with disease risk across the life course, and have been linked to cardiometabolic outcomes (e.g., hypertension and diabetes) and markers of vascular injury and neurodegeneration [2–6]. It is well known that there exists a social gradient in health, suggesting that marginalized groups experience worse health outcomes which are driven by structural inequities, discrimination, and unequal distribution of resources [7].

Understanding racial inequities in the United States necessitates a focus on the role of SDoH in perpetuating disparities in health. SDoH are connected and adverse determinants tend to cluster in marginalized groups and rarely compared within a single sex and marginalized racial ethnic group. These determinants may have a differential impact on health, which necessitates a more holistic framework in the assessment and requires a within group comparison for the appropriate contextual framing [8–10]. However, evidence in the association between SDoH and many physiological outcomes is based on single assessments. Few indices of cumulative disadvantage capture the burden of adverse determinants across an exhaustive range of domains. A comprehensive framework may provide more nuanced insights to better identify at-risk, marginalized populations or those who are responsive to social interventions [11].

Non-Hispanic Black (NHB) men face some of the most pronounced and persistent SDoH-related risks due to structural racisms, socioeconomic marginalization, social isolation, and limited access to high-quality healthcare [8, 12]. Black men in the United States are at an increased risk of cardiometabolic conditions compared with other demographic populations, except for American Indians [13]. These disparities are compounded by chronic exposure to discrimination and cumulative stressors that are known to exacerbate cardiometabolic disease pathways [14]. Disparities in cognitive performance also emerge from inequities, as adverse SDoH are associated with lower cognitive performance and an increased risk of cognitive impairment [15, 16]. The prevalence of cognitive decline is higher among Black Americans [17]; however, very little data exists on the cognitive decline in NHB men, a group that experiences undue burden of adverse social determinants of health [18].

Socioeconomic status has been associated with differences in cognitive health with individuals from lower socioeconomics often experiencing higher rates of cognitive decline [19]. Educational attainment also plays a role, such that individuals with higher education often have lower risk of decline and a slower rate of decline over time [20, 21] Neighborhood disadvantages and other environmental factors also play a role in cognitive performance [22]. These factors may interact to impact cognitive performance in NHB men, and investigating their complex, synergistic effects can help to inform intervention development at different levels.

The Stress Process Model provides a useful framework for understanding how SDoH can influence cardiometabolic, cognitive and neuroimaging outcomes in NHB men [23, 24]. This model posits that social and structural disadvantage shapes exposure to stressors, access to coping resources, and subsequent health outcomes over time [23, 24]. Stress is socially patterned and embedded in everyday roles, lived experiences, and inequitable social conditions [23, 25]. For NHB men exposed to adverse SDoH, chronic financial hardship, neighborhood disadvantage, constrained access to care, and social marginalization may function as persistent stressors that accumulate and proliferate across multiple domains of life [23, 24]. Additionally, limited access to psychosocial and material resources may reduce the capacity to buffer these exposures [23, 24]. Applied to the present study, the Stress Process Model suggests that cumulative SDoH burden may influence cardiometabolic risk, brain health, and cognition in NHB men through both structural and stress-related pathways. Specifically, adverse SDoH may function as chronic stressors that accumulate over time, contribute to physiological wear and tear, and increase vulnerability to cardiometabolic dysregulation, vascular injury, and neurological dysfunction. At the same time, these exposures may limit access to psychosocial and material resources that could buffer their harmful effects. This framework therefore supports our focus on cumulative SDoH burden as a multidimensional exposure that may shape neuroimaging and cognitive outcomes in NHB men through linked social, behavioral and biological, and clinical pathways [25].

NHB men remain underrepresented in neuroepidemiologic research, limiting the understanding of population-specific mechanisms. The analytic focus on NHB men was intentional and supported by literature documenting structural inequities and disproportionate exposure to adverse social conditions across the life course [26]. NHB men experience a higher burden of cardiometabolic disease, earlier onset of vascular risk factors, greater exposure to social and economic disadvantage, and elevated risk of cognitive impairment and dementia compared with other demographic groups [3, 4, 18]. By restricting analyses to NHB men, the present study aims to reduce residual confounding by race-sex interactions, enhance etiologic specificity, and provide targeted evidence relevant to addressing health disparities in cognitive aging and dementia risk. The objective of this study was to examine associations between cumulative SDoH burden and cardiometabolic risk factors, neuroimaging markers, and cognitive outcomes among NHB men. Guided by the Stress Process Model, we conceptualized cumulative SDoH burden as an exposure that may increase vulnerability to adverse health outcomes. We hypothesized that greater cumulative SDoH burden would be associated with a higher prevalence of cardiometabolic risk factors and cognitive impairment, as well as with worse neuroimaging and cognitive outcomes, exhibiting graded associations across increasing levels of SDoH exposure. We further hypothesized that these associations would persist after accounting for traditional cardiometabolic risk factors and cognitive status. This present study contributes to the existing literature by identifying the syndemic social and structural factors that are linked to cardiometabolic risk, neuroimaging, and cognitive health among NHB men.

Methods

Cohort Study

The Health and Aging Brain Study – Health Disparities (HABS-HD), originally established as the Health & Aging Brain Among Latino Elders (HABLE) project, is a longitudinal, community-based cohort housed at the University of North Texas Health Science Center’s Institute for Translational Research [27, 28]. HABS-HD was purposefully designed to recruit NHW, NHB, and Hispanic adults and is the largest Alzheimer’s disease cohort focused on the Amyloid, Tau, and Neurodegeneration Framework. The HABS-HD study design and participant characteristics have been previously described [27, 28]. Oversight and ethical approval were provided by the University of North Texas Health Science Center Institutional Review Board, and all participants provided written informed consent in accordance with the Declaration of Helsinki.

Exposure: Construction of the Cumulative SDoH Index

The cumulative SDoH index was constructed using a data-driven and domain-informed approach across five domains (adapted from the Kaiser SDoH Framework and Hagan et al. [11]): community and social context, neighborhood and physical environment, education, economic stability, and healthcare access and quality. Individual components included a combination of binary indicators (e.g., presence of chronic stressors, insurance status, financial strain) and Likert-scale measures (e.g., social support items), which were recoded into favorable (0) and unfavorable (1) categories based on established conceptual frameworks of social disadvantage. Continuous composite variables, including total social support and chronic stress scores, were dichotomized using cohort-specific median thresholds to reflect relative disadvantage [11]. The cumulative SDOH burden index is an integer-valued score derived from binary indicators. Quartile boundaries were applied to balance group sizes (n=115 per quartile). Because the index is discrete, some integer values appear at adjacent quartile boundaries. Ties at boundaries were resolved by forced equal-n assignment. Quartile rank was subsequently entered as a continuous variable in linear trend tests to evaluate dose-response relationships. Detailed definitions of all SDoH variables, including survey items, response options, and coding schemes, are available in the HABS-HD data dictionary provided to investigators upon data request. To enhance transparency and reproducibility, Supplemental Table 1 provides comprehensive details for each variable, including source code, question/description, raw response options, binary coding rules, and rationale for classification. All variable definitions and descriptions were directly derived from the HABS-HD data dictionary.

All binary indicators were summed to create a continuous cumulative SDoH index (range = 0–37), with higher scores indicating greater disadvantage. The index was additionally categorized into quartiles to evaluate graded associations across increasing levels of disadvantage. This approach allows for harmonization across heterogeneous domains while maintaining interpretability of cumulative disadvantage burden (Supplemental Table 1).

Within the community and social context domain, indicators included marital status (i.e., marriage or partnership associated with greater social integration), perceived social support, and chronic stress exposures. Social support items were scored according to established direct and reverse-coding conventions, such that lower levels of perceived support indicated greater disadvantage. Individual items were summed to create a composite social support score, which was dichotomized at the sample median, with values below the median classified as unfavorable. Chronic stress exposures, including caregiving strain, financial stress, and job strain, were aggregated into a total stress score and similarly dichotomized at the median, with higher stress burden coded as 1 for unfavorable.

The neighborhood and physical environment domain captured residential stability, housing tenure, and neighborhood socioeconomic disadvantage. Indicators included years at current residence, home ownership status, and Area Deprivation Index (ADI) measures. ADI was linked to each participant at the census block group level using geocoded residential addresses matched to Neighborhood Atlas data. Both the national ADI percentile (1–100) and state ADI decile (1–10) were included in the neighborhood and physical environment domain. The national percentile situates each block group relative to all U.S. block groups, while the state decile captures relative disadvantage within Texas, providing complementary reference frames for neighborhood-level exposure. A threshold of ≥51 was applied to reflect neighborhoods at or above the national median level of disadvantage, providing an empirically grounded and symmetric designation of unfavorable neighborhood conditions. For the state ADI decile, a threshold of ≥6 was applied, placing neighborhoods within the top 40–50% of most disadvantaged areas statewide. These thresholds were selected to capture broad exposure to neighborhood disadvantage rather than to isolate extreme deprivation, and each indicator was coded as unfavorable (1) or favorable (0). To our knowledge, this study is the first to aggregate these SDOH indicators into a cumulative burden index within the HABS-HD cohort. In both metrics, higher values reflect greater disadvantage, which is consistent with existing HABS-HD studies [29, 30].

The education domain reflected educational attainment based on total years of education and completion of high school. Fewer than 12 years of education or absence of a high-school diploma or GED was classified as 1 for unfavorable.

Indicators in the economic stability domain included household income, employment status, and stress related to financial hardship or job strain. Annual household income below $50,000, unemployment, and endorsement of job-related or financial stress were coded as 1 for unfavorable conditions.

Healthcare access and quality domain incorporated insurance status, access to a personal healthcare provider, affordability of care, and use of preventive services. Lack of health insurance, absence of a regular provider, inability to afford medical care within the past 12 months or delayed routine check-ups (greater than two years) were classified as 1 for unfavorable.

Index Construction and Quartile Classification

Binary indicators of favorable ( = 0) or unfavorable ( = 1) across all domains were summed to generate a continuous cumulative SDoH index, with higher values indicating greater cumulative social disadvantage. For descriptive purposes, the index was examined as a continuous measure and subsequently categorized into quartiles based on the sample distribution. Quartile 1 represented the lowest SDoH burden (most favorable conditions), whereas Quartile 4 represented the highest burden (most unfavorable conditions). See Table 1 for overall categories and Supplemental Table 1 for HABS-HD specific variable details.

Table 1.

SDoH Table

Domain	Example Items (from HABS-HD Question/Description field)
Economic Stability	• Total annual household income • Retired or not working • Job strain • Financial strain
Neighborhood & Physical Environment	• Years at current residence • Homeownership status • Area Deprivation Index (state) • Area Deprivation Index (national)
Education	• Highest grade completed • High school diploma or GED
Food Security	(No variable in dataset — placeholder for future measures)
Community & Social Context	• Marital status • Hard to find someone to go on a trip • No one to share worries/fears • Someone to help if sick • Someone to talk to about personal problems • Not often invited to do things • Hard to find help when moving • Presence of chronic stressors • Total chronic stress score
Healthcare System / Access & Quality	• Private insurance coverage • Medicare/Medicaid coverage • No insurance • Has personal doctor/provider • Could not afford doctor visit (past 12 months) • Time since last routine check-up

Age

Participant age was recorded in years based on self-report responses.

Cardiometabolic Risk Factors

Cardiometabolic risk factors included diabetes, hypertension, dyslipidemia, obesity, and tobacco use, each modeled as a dichotomous indicator. Detailed operational definitions for these variables have been described previously in HABS-HD methodological reports [3–5]. In brief, hypertension was defined by documented diagnosis or antihypertensive medication use; diabetes was determined by clinical diagnosis or treatment; dyslipidemia reflected medical history or lipid-lowering therapy; obesity was defined as body mass index (BMI) ≥ 30 kg/m²; and tobacco use captured current or prior dependence.

Neuroimaging for Hippocampal Volume

Hippocampal volumes were derived from T1-weighted structural magnetic resonance imaging using HippoDeep [31, 32]. This automated deep-learning–based pipeline generated lateralized left and right hippocampal volume estimates, along with total intracranial volume (ICV), for each participant. Hippocampal atrophy is a well-established neuroimaging marker associated with mild cognitive impairment (MCI) and Alzheimer’s disease and was therefore examined as an indicator of neurodegeneration [33]. For this study, the left and right hippocampal volumes were summed to create a total hippocampal volume measure. Total hippocampal volume and ICV were standardized to z-scores prior to analysis to facilitate comparability across participants.

Neuroimaging for WMH Volume

White matter hyperintensities (WMH) were quantified from fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) using the harmonized processing pipeline established for the HABS-HD cohort [27]. WMH measures were derived from FLAIR scans acquired between 2017 and 2025 on Siemens Skyra and Vida scanners. Each scan was linked to detailed scanner metadata and the visit-specific MRI acquisition date to allow analytic adjustment for scanner-related variability and to support harmonization across four primary acquisition periods.

WMH segmentation was conducted using the Lesion Growth Algorithm implemented in the Lesion Segmentation Toolbox (LST) for SPM. For each imaging session, the processing pipeline generated a raw WMH volume expressed in milliliters (mL). Automated quality-control procedures were applied to evaluate data completeness, the frequency of zero-valued WMH measurements, agreement between raw and log-transformed WMH values, distributional irregularities, and the presence of placeholder or invalid codes. These checks indicated acceptable levels of missingness (<2.5%) and zero values (<7%) and confirmed internal consistency between raw and transformed WMH measures across scans. Quality-control (QC) flags generated by the imaging pipeline, including FLAIR acquisition QC and WMH segmentation QC indicators, were retained and used to exclude images that did not meet predefined structural or lesion-mapping quality criteria. Thus, there exists a small percentage of missingness for these neuroimaging markers shown in Table 2.

Table 2.

Characteristics of Non-Hispanic Black Men Stratified by Social Determinants of Health (SDoH) Quartile

Higher SDoH index and quartile indicate greater SDoH burden.
Characteristic	Overall (N=460)	Q1 (N=115)	Q2 (N=115)	Q3 (N=115)	Q4 (N=115)
SDoH Index, Mean (SD)	10.1 (5.2)	4.4 (1.3)	7.7 (0.9)	11.0 (1.1)	17.3 (3.5)
Median [Min, Max]	9.0 [1.0, 29.0]	5.0 [1.0, 6.0]	8.0 [6.0, 9.0]	11.0 [9.0, 13.0]	17.0 [13.0, 29.0]
SDoH – Education, Mean (SD)	0.2 (0.5)	0.0 (0.1)	0.1 (0.4)	0.1 (0.5)	0.5 (0.7)
Median [Min, Max]	0 [0, 2.0]	0 [0, 1.0]	0 [0, 2.0]	0 [0, 2.0]	0 [0, 2.0]
SDoH – Economic, Mean (SD)	1.3 (1.0)	0.5 (0.5)	1.0 (0.8)	1.4 (0.9)	2.3 (1.0)
Median [Min, Max]	1.0 [0, 4.0]	1.0 [0, 2.0]	1.0 [0, 3.0]	1.0 [0, 4.0]	2.0 [0, 4.0]
SDoH – Neighborhood, Mean (SD)	1.4 (1.2)	0.5 (0.7)	1.1 (1.0)	1.8 (1.1)	2.4 (1.1)
Median [Min, Max]	1.0 [0, 4.0]	0 [0, 3.0]	1.0 [0, 4.0]	2.0 [0, 4.0]	2.0 [0, 4.0]
SDoH – Healthcare, Mean (SD)	2.4 (1.2)	1.9 (0.5)	2.1 (0.9)	2.5 (1.1)	3.1 (1.5)
Median [Min, Max]	2.0 [0, 7.0]	2.0 [1.0, 3.0]	2.0 [0, 6.0]	2.0 [1.0, 6.0]	3.0 [1.0, 7.0]
SDoH – Social, Mean (SD)	4.8 (3.7)	1.5 (1.3)	3.3 (1.6)	5.3 (2.1)	9.1 (3.6)
Median [Min, Max]	4.0 [0, 17.0]	1.0 [0, 5.0]	3.0 [0, 8.0]	5.0 [1.0, 10.0]	9.0 [1.0, 17.0]
Age (years), Mean (SD)	63.4 (7.2)	64.6 (7.7)	64.2 (7.6)	63.8 (6.7)	61.2 (6.3)
Diabetes, n (%)	124 (27.0)	22 (19.1)	35 (30.4)	29 (25.2)	38 (33.0)
Hypertension, n (%)	362 (78.7)	91 (79.1)	97 (84.3)	88 (76.5)	86 (74.8)
Dyslipidemia, n (%)	262 (57.0)	71 (61.7)	62 (53.9)	66 (57.4)	63 (54.8)
Obesity, n (%)	231 (50.2)	50 (43.5)	60 (52.2)	64 (55.7)	57 (49.6)
Tobacco Dependence, n (%)	93 (20.2)	9 (7.8)	16 (13.9)	22 (19.1)	46 (40.0)
WMH (mL), Mean (SD)	4.9 (9.4)	4.0 (8.3)	4.3 (6.5)	5.9 (12.0)	5.6 (10.1)
Median [Min, Max]	1.36 [0, 65.3]	1.3 [0, 65.3]	1.5 [0, 33.2]	1.5 [0, 61.6]	1.16 [0, 57.8]
Missing, n (%)	27 (5.9)	4 (3.5)	8 (7.0)	10 (8.7)	5 (4.3)
Intracranial Volume (mm3), Mean (SD)	1530000 (116000)	1550000 (123000)	1520000 (119000)	1530000 (98200)	1510000 (122000)
Missing, n (%)	82 (17.8)	21 (18.3)	20 (17.4)	22 (19.1)	19 (16.5)
Total Hippocampal Volume (mm3), Mean (SD)	6470 (809)	6540 (827)	6500 (813)	6480 (853)	6380 (747)
Missing, n (%)	119 (25.9)	27 (23.5)	32 (27.8)	33 (28.7)	27 (23.5)
Normalized WMH, Mean (SD)	−0.0 (2.70)	0.259 (1.32)	−0.128 (3.04)	0.169 (2.60)	−0.290 (3.36)
Missing, n (%)	82 (17.8)	21 (18.3)	20 (17.4)	22 (19.1)	19 (16.5)
MRI Scanner
Vida1	369 (80.2)	92 (80.0)	94 (81.7)	92 (80.0)	91 (79.1)
Skyra	1 (0.2)	0 (0)	0 (0)	0 (0)	1 (0.9)
Vida1a	4 (0.9)	0 (0)	1 (0.9)	0 (0)	3 (2.6)
Vida2	4 (0.9)	2 (1.7)	0 (0)	1 (0.9)	1 (0.9)
Missing, n (%)	82 (17.8)	21 (18.3)	20 (17.4)	22 (19.1)	19 (16.5)
Cognitive Impairment, n (%)	195 (42.4)	37 (32.2)	42 (36.5)	49 (42.6)	67 (58.3)
Memory (z), Mean (SD)	−0.0 (0.9)	0.2 (0.9)	0.1 (0.8)	−0.0 (0.9)	−0.3 (0.9)
Executive Function (z), Mean (SD)	−0.0 (0.9)	0.3 (0.8)	0.1 (0.8)	−0.1 (0.9)	−0.3 (0.9)
Processing Speed (z), Mean (SD)	−0.0 (0.9)	0.3 (0.6)	−0.0 (0.9)	−0.1 (1.1)	−0.2 (1.0)
Language (z), Mean (SD)	−0.0 (0.9)	0.1 (0.8)	0.1 (0.9)	−0.1 (0.9)	−0.1 (0.9)

Following our previous methodology [3–5], raw WMH volumes were log-transformed following the addition of a small constant (1e-6) to account for zero values. To adjust for interindividual differences in head size, WMH volumes were normalized using residuals from a linear regression model with ICV as the predictor and log-transformed WMH volume as the outcome. Potential outliers were identified using the interquartile range (IQR) method applied to the ICV-adjusted WMH residuals[3–5]. Values falling below the 25th percentile minus 1.5 times the IQR or above the 75th percentile plus 1.5 times the IQR were excluded. The resulting ICV-adjusted WMH residuals were used as the primary WMH outcome measure in subsequent analyses.

Clinical Cognitive Status

Clinical cognitive classification was determined using a standardized adjudication process previously established in the HABS-HD cohort[27, 28]. Diagnostic determinations incorporated participant- and informant-reported changes in everyday functioning in conjunction with performance on the comprehensive neuropsychological assessment battery. Neuroimaging measures were not used in the diagnostic adjudication. For analytic purposes, participants were categorized as cognitively impaired if they met criteria for mild cognitive impairment (determined by neuropsychological z-scores ≤ −1.5 on at least one cognitive test and Clinical Dementia Rating sum boxes score ≥0.5 but <2) or dementia (determined by neuropsychological z-scores ≤ −2 on two or more cognitive tests and Clinical Dementia Rating sum boxes score ≥ 2)[27, 28].

Neuropsychological Evaluation and Cognitive Domains

The neuropsychological testing protocol for the HABS-HD has been described in detail elsewhere[27, 28]. In brief, participants completed a comprehensive battery assessing multiple cognitive domains, including the Mini–Mental State Examination (MMSE); Digit Span and Logical Memory subtests from the Wechsler Memory Scale–Third Edition (WMS-III); Digit Symbol Substitution; Trail Making Test Parts A and B; the Spanish-English Verbal Learning Test (SEVLT); Animal Naming (semantic fluency); and the FAS test of phonemic verbal fluency.

Cognitive domain construction followed previously established HABS-HD procedures [34]. Individual test scores were first standardized to z-scores, after which domain-specific composite scores were calculated by averaging the relevant standardized measures.

The episodic memory composite was derived from four tasks: SEVLT Immediate Recall, SEVLT Delayed Recall, Logical Memory I, and Logical Memory II. Executive function was represented by the mean of z-scores from the Digit Span and Digit Symbol Substitution tests. Processing speed was indexed using a composite of Trails A and Trails B z-scores, with values inverted so that higher scores reflected better performance. Language ability was assessed using the average z-score from Animal Naming and FAS verbal fluency tasks.

Statistical Analysis

From the November 2025 HABS-HD Release 7 dataset, participant data were available for 4,134 individuals across all sex (male & female) and racial/ethnic groups. Of these, 1,592 were male. The analytic sample was restricted to non-Hispanic Black (NHB) male participants, resulting in 465 individuals based on self-identified race/ethnicity. Five participants were excluded due to missing education data, yielding a final analytic sample of 460 NHB male participants (Table 1). Analyses were conducted using complete-case data, and no imputation was performed for missing covariates or other variables included in the models.

Missingness across the 38 SDoH source variables was generally low in the analytic sample of 460 NHB males. The two variables with the highest missingness were ADI state decile and national percentile rankings, both missing for 63 participants (13.7%), likely reflecting geocoding failures or residence outside of ranked census block groups. Job-related chronic stress was missing for 47 participants (10.2%), and household income was missing for 21 participants (4.6%). All remaining 34 variables had missingness at or below 0.4% (n≤2), and eight variables: years of education, high school completion, private insurance, Medicare, no insurance, primary care provider, cost-related care avoidance, and cost of care in the past 12 months had no missing data (Supplemental Table 2). Because the cumulative index was constructed using listwise summation with missing values excluded, missing indicators were assigned a value of zero, conservatively underestimating cumulative SDoH burden for affected participants. Given the minimal missingness across most indicators, this assumption is unlikely to meaningfully bias the overall index distribution or study conclusions.

Descriptive Analyses

We summarized participant characteristics across SDoH quartiles and presented them for descriptive purposes only; no formal statistical testing was performed.

Modified Poisson Regression

We evaluated associations between cumulative SDoH burden and the prevalence of cardiometabolic conditions (hypertension, diabetes, dyslipidemia, obesity, and tobacco use), as well as cognitive impairment, using modified Poisson regression with robust standard errors [35]. Each outcome was analyzed in a separate model as a binary dependent variable. SDoH burden was entered as a categorical exposure defined by quartiles, with the lowest quartile serving as the reference group. All models included age, parameterized per 5-year increment. Statistical significance was assessed using a two-sided α level of 0.05, and the prevalence ratio (PR), 95% confidence intervals (CI), and p-value were reported

Multivariable Linear Regression Analyses

Associations between cumulative SDoH burden and neuroimaging and cognitive outcomes were examined using multivariable linear regression models. SDoH burden was modeled categorically by quartiles, with Quartile 1 (lowest burden) designated as the reference category.

Neuroimaging Outcomes

Neuroimaging models were adjusted for age (per 5-year increment), MRI scanner, and cardiometabolic risk factors, including diabetes, hypertension, dyslipidemia, obesity, and tobacco use. WMH volume was quantified as intracranial volume (ICV)–adjusted residuals obtained by regressing log-transformed WMH volume on ICV. Hippocampal volume was analyzed as standardized (z-scored) bilateral hippocampal volume, with additional adjustment for ICV included in all models. Results are presented as regression coefficients (β) with 95% confidence intervals (CI), interpreted as adjusted differences in WMH or hippocampal volume relative to Quartile 1.

Cognitive Outcomes

Associations between SDoH quartiles and cognitive performance were assessed separately for memory, executive function, processing speed, and language domains using multivariable linear regression. All cognitive models were adjusted for age and cardiometabolic risk factors (diabetes, hypertension, dyslipidemia, obesity, and tobacco use). Effect estimates are reported as β coefficients with corresponding 95% CI and two-sided p values.

Cardiometabolic risk factors were included as covariates to account for potential confounding given their established associations with cognitive outcomes [6]. Although these variables may lie along the causal pathway between SDoH and cognition, their inclusion allows for estimation of associations independent of traditional vascular risk factors.

Sensitivity Analyses

To evaluate whether observed associations were independent of baseline cognitive impairment, fully adjusted cognitive models were re-estimated with additional adjustment for cognitive status, categorized as cognitively unimpaired versus cognitively impaired (mild cognitive impairment or dementia). Analyses were guided by a priori directional hypothesis grounded in prior literature. Given the single primary exposure and theoretically motivated outcome selection, p-values were not adjusted for multiple comparisons.

Trend Analyses

We evaluated linear trends across SDoH quartiles using multivariable linear regression models with SDoH quartile modeled as an ordinal numeric variable (coded 1–4), such that the regression coefficient represented the estimated change in the outcome per 1-quartile increase in SDoH quartile. Separate models were fit for each outcome with covariate adjustment as previously described. For visualization, we additionally fit parallel models treating SDoH quartile as a categorical factor (Q1–Q4) and estimated covariate-adjusted marginal means with 95% CI. Regression coefficients, 95% CI, and p-values for trend were extracted from the ordinal trend models.

Statistical analyses were conducted in R v4.2.3. P-values less than 0.05 were considered to be statistically significant. Data Availability: HABS-HD data can be requested at the following: https://loni.usc.edu/. The following study was conducted using data available at release 7.

Results

NHB Men Cohort Characteristics

The analytic sample included 460 NHB men (Table 2). The mean age was 63.4 years (SD = 7.2). Higher SDoH quartiles reflected progressively greater cumulative SDoH burden across domains. The prevalence of cardiometabolic conditions increased across SDoH quartiles, including diabetes (19.1% in Quartile 1 to 33.0% in Quartile 4) and current tobacco use (7.8% to 40.0%). Hypertension and dyslipidemia were similar across all quartiles. Cognitive impairment prevalence increased with higher SDoH burden, from 32.2% in Quartile 1 to 58.3% in Quartile 4. Mean cognitive domain scores showed graded declines across SDoH quartiles for memory, executive function, processing speed, and language.

Associations between SDoH Quartiles and Cardiometabolic Risk Factors and Clinical Cognitive Impairment

Higher cumulative SDoH burden was differentially associated with the prevalence of cardiometabolic conditions and cognitive impairment (Table 3). Higher SDoH burden was associated with a greater prevalence of diabetes, with elevated prevalence observed in Quartile 2 (PR = 1.10 [95% CI 1.01–1.20]; p = 0.034) and Quartile 4 (PR = 1.15 [95% CI 1.05–1.26]; p = 0.002) compared with Quartile 1. Higher SDoH burden was associated with increased prevalence of tobacco use, with significant associations in Quartile 3 (PR = 1.10 [95% CI 1.02–1.19]; p = 0.012) and Quartile 4 (PR = 1.29 [95% CI 1.19–1.39]; p < 0.001). No significant associations were observed for hypertension, dyslipidemia, or obesity. Participants in the highest SDoH quartile exhibited a higher prevalence of cognitive impairment compared with those in the lowest quartile (PR = 1.23 [95% CI 1.13–1.34]; p < 0.001).

Table 3.

Associations Between Cumulative Social Determinants of Health (SDoH) Burden and Prevalent Cardiometabolic Conditions and Cognitive Impairment Among Non-Hispanic Black Men

Outcome	Predictor	Prevalence Ratio (95% CI)	p value
Hypertension	Age (per 5-year increase)	1.01 (1.00–1.03)	0.102
	SDoH Quartile 2	1.03 (0.98–1.09)	0.285
	SDoH Quartile 3	0.99 (0.93–1.05)	0.684
	SDoH Quartile 4	0.98 (0.93–1.05)	0.606
Diabetes	Age (per 5-year increase)	1.04 (1.02–1.07)	<0.001
	SDoH Quartile 2	1.10 (1.01–1.20)	0.034
	SDoH Quartile 3	1.06 (0.97–1.15)	0.204
	SDoH Quartile 4	1.15 (1.05–1.26)	0.002
Dyslipidemia	Age (per 5-year increase)	1.03 (1.01–1.05)	0.001
	SDoH Quartile 2	0.95 (0.88–1.03)	0.218
	SDoH Quartile 3	0.98 (0.90–1.06)	0.538
	SDoH Quartile 4	0.98 (0.90–1.06)	0.554
Obesity	Age (per 5-year increase)	0.97 (0.95–0.99)	0.013
	SDoH Quartile 2	1.06 (0.97–1.16)	0.171
	SDoH Quartile 3	1.08 (0.99–1.18)	0.076
	SDoH Quartile 4	1.02 (0.94–1.12)	0.602
Tobacco Use	Age (per 5-year increase)	0.99 (0.97–1.01)	0.217
	SDoH Quartile 2	1.05 (0.98–1.13)	0.150
	SDoH Quartile 3	1.10 (1.02–1.19)	0.012
	SDoH Quartile 4	1.29 (1.19–1.39)	<0.001
Cognitive Impairment	Age (per 5-year increase)	1.04 (1.02–1.07)	<0.001
	SDoH Quartile 2	1.03 (0.94–1.13)	0.519
	SDoH Quartile 3	1.08 (0.99–1.18)	0.098
	SDoH Quartile 4	1.23 (1.13–1.34)	<0.001

Modified Poisson regression with robust variance estimation was used to examine associations between cumulative social determinants of health (SDoH) burden and the prevalence of hypertension, diabetes, dyslipidemia, obesity, tobacco use, and cognitive impairment among non-Hispanic Black men. SDoH burden was modeled categorically as quartiles, with Quartile 1 (lowest burden) serving as the reference group. All models were adjusted for age, modeled per 5-year increment. Results are presented as prevalence ratios (PRs) with 95% confidence intervals (CIs). Statistical significance was defined as a two-sided p < 0.05.

Associations of Cumulative SDoH Quartiles with Neuroimaging Outcomes

We examined associations between cumulative SDoH quartiles (Q2–Q4 vs. Q1) and two neuroimaging outcomes: WMH volume and hippocampal volume. Results from fully adjusted linear models are presented in Figure 1. For WMH volume (ICV-adjusted residual), higher SDoH quartiles were not significantly associated with WMH burden compared to the lowest quartile. For hippocampal volume (z-score), associations were also non-significant across all quartile comparisons.

Figure 1. Associations of Cumulative Social Determinants of Health (SDoH) Quartiles With Neuroimaging Outcomes in Non-Hispanic Black Men.

Forest plot showing adjusted linear regression coefficients (β) and 95% confidence intervals (CIs) for associations between cumulative SDoH burden (quartiles) and neuroimaging outcomes among non-Hispanic Black men. Estimates represent contrasts for SDoH Quartiles 2–4 relative to Quartile 1 (lowest burden; reference). Outcomes include intracranial volume–adjusted white matter hyperintensity residuals (WMH ICV-adjusted residual) and standardized bilateral hippocampal volume (HCV z). The WMH model was adjusted for MRI scanner, age (per 5-year increment), diabetes, hypertension, dyslipidemia, obesity, and tobacco use. The hippocampal volume model included the same covariates and additionally adjusted for intracranial volume (ICV z) and scanner. The vertical dashed line at β = 0 indicates no association; points denote adjusted β estimates and horizontal bars indicate 95% CIs.

Associations between SDoH Quartiles and Cognitive Domain Scores

Higher cumulative SDoH burden was associated with worse performance across multiple cognitive domains (Figure 2). Relative to the lowest SDoH quartile, higher SDoH burden was associated with progressively lower memory performance in Quartile 3 (β = −0.27; 95% CI −0.48 to −0.06; p = 0.01) and Quartile 4 (β = −0.59; 95% CI −0.82 to −0.37; p < 0.01). Similar graded associations were observed for executive function, with lower performance in Quartile 3 (β = −0.44; 95% CI −0.64 to −0.23; p < 0.01) and Quartile 4 (β = −0.60; 95% CI −0.82 to −0.39; p < 0.01). Lower processing speed was also associated with increasing SDoH burden, including significant reductions in Quartile 2 (β = −0.27; 95% CI −0.49 to −0.05; p = 0.02), Quartile 3 (β = −0.42; 95% CI −0.64 to −0.20; p < 0.01), and Quartile 4 (β = −0.58; 95% CI −0.81 to −0.35; p < 0.01). For language performance, higher SDoH burden was more modestly associated with lower scores, with a borderline association observed for Quartile 4 relative to Quartile 1 (β = −0.23; 95% CI −0.45 to 0.00; p = 0.05), while Quartiles 2 and 3 were not significantly different from Quartile 1.

Figure 2. Associations of Cumulative Social Determinants of Health (SDoH) Quartiles With Cognitive Domain Performance in Non-Hispanic Black Men.

Forest plot showing adjusted linear regression coefficients (β) and 95% confidence intervals (CIs) for associations between cumulative social determinants of health (SDoH) burden and cognitive domain performance among non-Hispanic Black men. SDoH burden was modeled categorically as quartiles, with Quartile 1 (lowest burden) serving as the reference group. Cognitive outcomes include memory, executive function, processing speed, and language domain z scores. All models were adjusted for age (per 5-year increment), diabetes, hypertension, dyslipidemia, obesity, and tobacco use. The vertical dashed line at β = 0 indicates no association. Points represent adjusted β estimates and horizontal bars denote 95% CIs.

Associations between SDoH Quartiles and Cognitive Domain Scores Adjusting for Clinical Cognitive Impairment

In sensitivity analyses additionally adjusting for baseline cognitive impairment, associations between higher cumulative SDoH burden and poorer cognitive performance remained significant for memory, executive function, and processing speed, with Quartiles 3 and 4 consistently associated with lower scores relative to Quartile 1 (all p < 0.05), whereas associations with language performance were no longer significant (Supplemental Figure 1).

Associations between SDoH Quartiles and Cognitive Domain Scores Using Linear Trends

In multivariable models evaluating linear trends across SDoH quartiles, no significant associations were observed for WMH (ICV-adjusted residuals: β = 0.03 per quartile increase, 95% CI −0.20 to 0.27; p = 0.79) or hippocampal volume (β = −0.05, 95% CI −0.14 to 0.03; p = 0.23) (Figure 3). In contrast, higher SDoH quartile (greater burden) was associated with significantly lower performance across all cognitive domains. Each 1-quartile increase in SDoH was associated with lower memory (β = −0.19, 95% CI −0.26 to −0.12; p < 0.001), executive function (β = −0.21, 95% CI −0.27 to −0.14; p < 0.001), processing speed (β = −0.19, 95% CI −0.26 to −0.12; p < 0.001), and language scores (β = −0.08, 95% CI −0.16 to −0.01; p = 0.02).

Figure 3. Adjusted associations of SDoH quartiles with MRI and cognitive outcomes using trend analyses.

Panels show covariate-adjusted marginal means (points) and 95% confidence intervals (error bars) across SDoH quartiles (Q1–Q4) for the following outcomes, in order: WMH (ICV-adjusted residuals), hippocampal volumes (z), memory (z), executive function (z), processing speed (z), and language (z). Panel annotations report the linear trend estimate from separate regression models in which SDoH quartile was modeled as an ordinal numeric term (1–4): beta per 1-quartile increase, 95% confidence interval, and p-trend. Adjusted means were estimated from models treating SDoH quartile as a categorical factor; p-trend values were derived from the corresponding ordinal trend models.

Discussion:

The objective of this study was to evaluate associations between cumulative SDoH burden and cardiometabolic risk factors, neuroimaging markers, and cognitive outcomes among NHB men, a population disproportionately exposed to adverse social conditions and at elevated risk for cognitive impairment. Two primary findings were observed. First, higher SDoH burden was associated with a greater prevalence of diabetes, tobacco use, and cognitive impairment, and poorer cognitive performance across memory, executive function, and processing speed domains. These associations were robust across categorical (quartiles) and continuous specifications of SDoH burden and largely persisted after adjustment for cardiometabolic risk factors and baseline cognitive impairment. Second, cumulative SDoH burden was not independently associated with WMH or hippocampal volume. Together, these findings suggest that social disadvantage may exert a stronger influence on cognitive performance and impairment than on structural neuroimaging markers in NHB men, which is consistent with existing literature[30, 36].

Higher SDoH burden was associated with a greater prevalence of cardiometabolic risk, including diabetes and tobacco use. This is consistent with the existing literature highlighting disparities in cardiometabolic risk factors observed among NHB men[37, 38]. This specific population in the United States experiences disproportionately higher rates of diabetes and related complications, which co-occurs with structural inequities, such as neighborhood disadvantage, limited access to healthcare, economic instability, and chronic stress[39, 40]. Previous work has shown that diabetes risk is shaped by SDoH with racial and ethnic minority populations facing persistent disparities[41]. In the present study, higher SDoH burden was associated with greater diabetes prevalence, and higher tobacco use reflects broad structural conditions, which reinforces the need for multilevel and tailored interventions to reduce cardiometabolic risk among NHB men. It is important to understand whether these findings persist over time.

The null associations observed between SDoH quartiles and WMH volume in NHB males likely reflect the explanatory role of modifiable cardiometabolic risk factors that lie on the pathway between social disadvantage and white matter injury [2–4]. When models were fully adjusted for established WMH risk factors including age, hypertension, diabetes, dyslipidemia, obesity, and tobacco use, the association between SDoH quartiles and WMH was null across all comparisons. This pattern suggests that SDoH may not operate independently on white matter pathology but rather exerts its influence through downstream cardiometabolic conditions that are themselves more proximal drivers of cerebrovascular injury [3, 4]. Hypertension and related conditions may therefore mediate, rather than confound, the relationship between social determinants and WMH burden in this population. This interpretation is consistent with a growing body of literature demonstrating that cardiovascular risk factors, which are disproportionately prevalent among Black males due in part to structural social disadvantage, account for a substantial proportion of racial disparities in WMH volume [3, 4]. Taken together, these findings underscore the importance of upstream intervention on modifiable cardiometabolic risk factors as a mechanism through which reducing social disadvantage may ultimately protect against white matter injury in this high-risk group.

The robust associations observed between cumulative SDoH burden and cognitive performance are consistent with a growing body of evidence linking social adversity with accelerated cognitive aging among Black adults [30, 36]. Domain-specific effects, specifically memory, executive functioning, and processing speed, suggest that cumulative disadvantage may disproportionately affect neural systems sensitive to chronic stress, reduced cognitive reserve, and vascular dysfunction. Prior studies from MESA and REGARDS similarly demonstrate that educational attainment, neighborhood disadvantage, and healthcare access predict cognitive outcomes independent of traditional factors [42–44]. These findings likely support a pathway through which sustained exposure to social disadvantage contributes to cognitive vulnerabilities via inflammatory and vascular mechanisms, which are established contributors to cognitive dysfunction particularly in the memory, executive function, visuospatial/processing speed domains [45–47].

Notably, the relatively high prevalence of cognitive impairment observed even among participants in the lowest SDoH burden quartile warrants careful interpretation. This likely reflect key characteristics of the HABS-HD cohort and the age (mean 63.4 years) and demographic structure of the sample (NHB men), which may capture individuals already at elevated risk for cognitive decline. Additionally, low SDoH burden represents a relative categorization within a cohort of NHB men and does not indicate the complete absence of social or structural disadvantage. As such, the reference group (low SDoH quartile) may carry some baseline vulnerability not captured by the SDoH index.

These findings may also be framed through the Stress Process Model concepts of stressor proliferation and resource depletion, which are consistent with the pathways described earlier [24, 25]. For NHB men, adverse SDoH exposures may not occur in isolation; rather, one disadvantage, such as economic strain or neighborhood disadvantage, may trigger additional stressors across other domains, including healthcare access, social relationships, and opportunities for health-promoting behaviors [8, 9, 14, 26]. Over time, this proliferation of stressors may coincide with the erosion of psychosocial and material resources, such as financial stability, social support, and access to consistent, high-quality care, that might otherwise buffer their effects. In this way, cumulative SDoH burden may contribute to worse cognitive outcomes not only because of the presence of multiple disadvantages, but because these disadvantages can compound one another while simultaneously depleting the resources needed to maintain brain health and resilience.

Strengths and Limitations

This study has several notable strengths. First, it focuses explicitly on NHB men, a population that remains underrepresented in neuroepidemiology research despite disproportionate exposure to adverse social conditions and elevated risk for cardiometabolic disease and cognitive impairment. By centering this group, the analysis addresses a critical gap in the literature and avoids extrapolating inferences from predominantly White cohorts. Second, the use of a cumulative SDoH burden index allowed for a multidimensional assessment of social adversity that more closely reflects real-world clustering of social risks than single-domain indicators. The consistency of associations across both categorical and continuous specifications of SDoH burden further supports the robustness of the findings.

Several limitations should also be considered. The study was cross-sectional; therefore, we are not able to make causal inferences related to the predictor and outcomes, and it is important to determine whether these findings persist over time. Although we used all available data from HABS-HD to construct the SDoH index, some important factors were not collected or obtainable including nutrition and physical activity, which may be relevant for cognitive and brain health. Additionally, simply adding individual factors into a cumulative index does not capture the weighted effects of each individual factor; however, this approach is widely accepted [11, 48]. Neuroimaging data were available for a subset of participants, which may have reduced power to detect modest associations with WMH volume or hippocampal volume and increases selection bias. A limitation of the present study is the inability to formally stratify analyses by cognitive status. Adjusting cognitive domain models for clinical cognitive impairment risks over-adjustment, as the diagnostic classification is derived in part from the same neuropsychological battery underlying the outcome z-scores. Stratification was not feasible given sample size constraints. Future work in larger, diverse cohorts should evaluate whether the observed SDOH-cognition associations are modified by cognitive status. Additionally, the SDoH burden index captures self-reported exposure but not timing, duration, or severity of social adversity, which may differentially influence brain structure versus cognitive performance. We also acknowledge that dichotomization and median-based splits can introduce potential limitations, including loss of information and reduced statistical power. P-values are unadjusted for multiple comparisons. Findings near α=0.05 should be interpreted cautiously, and replication in independent cohorts is warranted. Additionally, forced quartiling of a discrete integer index results in some boundary values appearing across adjacent quartiles, which may modestly influence between-quartile effect estimates. Readers should interpret adjacent quartile comparisons with this in mind. Lastly, these findings may not be applicable beyond the NHB male community-based cohort from HABS-HD. Participants were recruited from Texas and within a defined age range, which may limit geographic and age-related representation. In addition, HABS-HD employed a community-based recruitment strategy, and the voluntary participation may introduce selection bias (e.g., healthier or more motivated), affecting the extent to which these findings extend to broader NHB population or other demographic groups.

Conclusion and Implications

Findings from the present study underscore the importance of SDoH as key correlates of inequities in cardiometabolic risk factors and cognitive health among NHB men. Specific interventions targeting cardiometabolic risk have been shown to improve or preserve cognitive outcomes [34]; however, specific strategies targeting education systems, economic instability, neighborhood environments, and exposure to discrimination may represent promising leverage points for future intervention and policy efforts [49, 50]. Public health frameworks, such as the Centers for Disease Control and Prevention & Alzheimer’s Association Healthy Brain Initiative, emphasizes the importance of structural and community-level interventions as critical components of dementia risk in vulnerable populations. Further, research suggests that enhancing social resources, such as social cohesion and support, can mitigate the harmful cognitive effects of cumulative disadvantage. The findings from the present study reinforce the importance of understanding the cumulative burden of SDoH to clarify the pathways through which social adversity impacts brain health in NHB men. Our study and findings highlight a critical need for prevention strategies that integrate social policy with traditional biomedical approaches to reduce inequities in dementia risk with a particular focus on interventions for those with the greatest prevalence of not only the outcomes but modifiable associative risk factors.

Highlights.

• Cumulative social disadvantage was strongly associated with poor cognitive performance across multiple domains in non-Hispanic Black men.

• SDoH burden was not associated with hippocampal volume or white matter hyperintensities, suggesting cognitive impacts may arise through pathways not captured by neuroimaging markers.

• Future work should evaluate longitudinal SDoH trajectories and mechanistic pathways to identify prevention targets that may mitigate cognitive decline in non-Hispanic Black men.

Research in Context.

Systematic Review:

The authors reviewed the literature using trusted search engines, including PubMed, Google Scholar, to find articles investigating the associations of cumulative social determinant of health (SDoH) burden on cardiometabolic risk factors, cognitive performance, and neuroimaging outcomes in non-Hispanic Black (NHB) men. However, few studies have investigated cumulative SDoH burden associations on the outcomes, particularly in NHB men.

Interpretation:

These findings advance our understanding of how the accumulation of social determinants of health impact cardiometabolic risk factors, cognitive and neuroimaging outcomes on NHB men in middle to late life.

Future Directions:

Key next steps should include understanding the pathways by which SDoH and impact these outcomes. Also understanding whether the observed relationships vary by social context is important for improving brain health of NHB men.