Independent and joint associations of key social exposome components with cognitive aging: triangulating evidence through cross-national data.

Independent and Joint Associations of Key Social Exposome Components with Cognitive Aging: Triangulating Evidence Through Cross-National Data - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Am J Epidemiol . Author manuscript; available in PMC: 2026 May 31. Published in final edited form as: Am J Epidemiol. 2026 Feb 5;195(2):398–406. doi: 10.1093/aje/kwaf189 Independent and Joint Associations of Key Social Exposome Components with Cognitive Aging: Triangulating Evidence Through Cross-National Data Ruijia Chen 1 , Harold Lee 2 , Jingxuan Wang 3 , Yulin Yang 3 , Sakurako S Okuzono 4 , Kristen Nishimi 5, 6 , Lindsay Kobayashi 7 , M Maria Glymour 1 , Laura D Kubzansky 3 Author information Copyright and License information 1. Department of Epidemiology, Boston University School of Public Health 2. Department of Biobehavioral Health, Penn State University 3. Department of Epidemiology and Biostatistics, University of California, San Francisco 4. Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health 5. Department of Psychiatry and Behavioral Sciences, University of California, San Francisco 6. Mental Health Service, San Francisco Veterans Affairs Health Care System 7. Department of Epidemiology, University of Michigan ✉ Corresponding Author: Ruijia Chen, ScD, Department of Epidemiology, Boston University School of Public Health, Rchen3@bu.edu PMC Copyright notice PMCID: PMC13221738  NIHMSID: NIHMS2162625  PMID: 40878756 The publisher's version of this article is available at Am J Epidemiol Abstract We examined the independent and joint associations of five key social exposome components, including financial strain, neighborhood disorder, perceived discrimination, social strain, and traumatic life events, with cognitive function levels and decline. Data were from adults aged > 50 in the US Health and Retirement Study (HRS; n=13,795; 2008–2020) and the English Longitudinal Study of Ageing (ELSA; n=9,469; 2006–2019), and adults aged ≥ 65 in their Harmonized Cognitive Assessment Protocol (HCAP) subsamples (HRS-HCAP: n=2,749; 2016; ELSA-HCAP: n=955; 2018). Using linear mixed-effects models and quantile-based g-computation, we found that all components, except traumatic life events, were associated with lower cognitive function. Simultaneously lowering all components by one quartile could improve cognitive function by 0.09 SD units (95% CI: 0.07–0.11) in the HRS, 0.13 SD units (95% CI: 0.10–0.16) in the ELSA, and 0.07 SD units (95% CI: 0.02–0.13) in the HRS-HCAP. Neighborhood disorder had the strongest negative association with cognitive function in the U.S., while financial strain had the strongest association in England. No social exposome components were associated with faster cognitive decline. The associations of key social exposome components with cognitive function were consistent across countries, although the magnitude of the joint association was greater in England. Keywords: Social exposome, cross national, cognitive function, cognitive decline, quantile g-computation INTRODUCTION Social and environmental adversities, including financial strain, perceived discrimination, neighborhood disorder, social strain, and traumatic life events, have been associated with lower cognitive function, with some linked to faster cognitive decline. 1 – 4 However, findings from studies based on single countries have been inconsistent; cross-national comparisons offer valuable insights to better understand these associations across heterogeneous contexts. 5 , 6 Triangulating findings across countries can strengthen the evidence on how social exposures contribute to brain health. It can also help us better understand whether these social exposures are critical to the general aging process or whether their effects are dependent on a country’s unique social and economic context. 7 Social exposures often co-occur and accumulate, which potentially amplifies their negative health impacts. 8 , 9 Individuals experiencing multiple adversities may exhaust their coping resources, exacerbating the harmful effects of any single domain of adversity. 13 – 15 Moreover, as social adversities are often rooted in structural forces (e.g., structural racism) and tend to cluster by socioeconomic status and race and ethnicity, failing to consider their combined effects may obscure their contribution to health inequities. 8 , 10 Recent research has introduced the concept of the social exposome. 11 While not yet well-defined, the social exposome generally refers to the totality of social and economic exposures experienced over the life course. The social exposome framework emphasizes examining the combined health impacts of social exposures, rather than assessing them in isolation. The availability of rich psychosocial data from large, population-based surveys has spurred efforts to capture the complex effects of multiple social exposures. 9 , 12 – 14 One common method is through creating cumulative risk scores, which aggregate various components of the social exposome into a single metric. This approach aligns with the accumulation of risk model, which posits that social and environmental risk factors accumulate over time, resulting in long-term negative health outcomes. 15 While the cumulative risk composite scores are practical and intuitive, they assume risks stack up in an additive manner and often fail to capture interactions between components. 9 , 16 Recent advances in quantile-based g-computation offer greater flexibility in modelling the complex health effects of clustered exposures. The quantile-based g-computation estimates the effects on the outcome of simultaneously shifting multiple exposures by one quantile and allows for interactions among the components. 17 – 19 Compared to other “mixture analysis” approaches, quantile-based g-computation is more computationally efficient, more robust to unmeasured confounding, and does not assume that all exposures have effects in the same direction. 19 While the associations of key components of the social exposome with cognitive function have been documented, few studies have examined both their independent and joint associations across countries. Using data from the HRS, the English Longitudinal Study of Ageing (ELSA), and their Harmonized Cognitive Assessment Protocol (HCAP) subsamples, we quantified the independent and joint associations of five key social exposome components, including financial strain, social strain, neighborhood disorder, perceived discrimination, and traumatic life events, with cognitive function and decline. By focusing on two English-speaking countries with harmonized measures, this design provides a unique opportunity to triangulate evidence on the associations of social exposome components with cognitive aging. METHODS Study populations We used data from the HRS, HRS-HCAP, ELSA, and ELSA-HCAP. 20 – 22 The HRS is an ongoing, nationally representative cohort of U.S. adults aged >50, initiated in 1992. 20 Similarly, the ELSA is a nationally representative cohort of English adults aged 50+, initiated in 2002. 21 Both studies began collecting cognitive function data as part of the biennial core interviews from enrollment. In 2016 (HRS) and 2018 (ELSA), as part of the HCAP Network, 22 both studies (HRS-HCAP: n=3,347; ELSA-HCAP: n = 1,273) collected novel measures of cognitive function on subsets of participants aged 65 years or older using more rigorous methods to facilitate cross-national comparisons. As only one wave of HCAP data was available per study at the time of analysis, we used HCAP data solely for cross-sectional analyses of cognitive function levels. In the HRS, key social exposome components were assessed via self-administered leave-behind questionnaires starting in 2006, with a random half of the sample completing the questionnaire in each wave. We pooled data from the 2006 and 2008 waves to obtain complete data for the full sample. 23 In the ELSA, social exposome components were assessed via self-report questionnaires administered at each wave, except perceived discrimination (measured only in 2010). To ensure comparability, we drew social exposome data primarily from the 2006 waves of HRS (pooled 2006/2008) and ELSA, with the exception of perceived discrimination, for which the 2010 wave data were used. For the HCAP analyses, these social exposome assessments were linked to the HCAP cognitive assessments. We analyzed cognitive function data collected at the same time as or after the social exposome data. Accordingly, our baseline years were 2008 for HRS samples and 2006 for ELSA samples. Cognitive function data were available from 2008 through 2020 in the HRS and from 2006 through 2018/2019 in the ELSA (see Figure S1 for timelines). Participants missing baseline cognitive function data were excluded, and this resulted in analytic sample sizes of 13,795 (HRS), 9,469 (ELSA), 2,749 (HRS-HCAP), and 955 (ELSA -HCAP) (see Figure 1 for sample flowcharts). Figure 1. Open in a new tab Sample flowchart Measures Social exposome components. We included five key domains of the social exposome: financial strain, neighborhood disorder, perceived discrimination, social strain, and traumatic life events. These domains were selected because they have been shown to influence cognitive function, 2 – 5 , 12 are common experiences among older adults, and have comparable measures between HRS and ELSA. Financial strain. In the HRS, participants rated their satisfaction with their financial situation and difficulty meeting monthly payments. Responses were averaged to create a composite score, with higher scores indicating greater strain. 24 In the ELSA, financial strain was rated by self-rated ability to manage financially, with a higher score indicating greater strain. Neighborhood disorder. Both HRS and ELSA assessed neighborhood disorder using validated items assessing physical disorder, social cohesion, and safety. 25 Responses were averaged to create a summary score, with higher scores indicating more disorder. Perceived discrimination. Both cohorts used the Everyday Discrimination Scale to assess perceived discrimination. 26 In the ELSA, the item ‘people act as if they are afraid of you’ was replaced with ‘you are treated unfairly at hospitals or by healthcare providers.’ Scores were averaged, with a higher score indicating greater discrimination. Social strain. Social strain was measured via questions on the relationships with partners, children, family members, and friends. 27 – 29 Scores were averaged for each type of strain and summed to create a total social strain score. Following previous research, 30 participants who lacked a given relationship (e.g., no children) were assigned the lowest score for that domain. Traumatic life events. In the HRS, participants reported exposure to seven lifetime traumatic events (i.e., death of a child, major disaster, combat or war zone exposure, a close contact with alcohol or drug problems, physical assault or abuse, life-threatening illness, or having a relative with a life-threatening illness) and four childhood traumatic events (i.e., trouble with police, parental substance use, physical abuse, or repeating a grade in school). 31 In the ELSA, participants reported exposure to twelve lifetime traumatic events (i.e., major disaster, life-threatening illness, physical attack or assault, sexual abuse or assault, unemployment, close friend with drug addiction, combat or war zone exposure, witnessing war or armed conflict, witnessing a death or serious injury, loss of a loved one due to war or conflict, friend’s accidental or violent death, or severe financial crisis) and three childhood traumatic events (i.e., parental conflict, parental alcohol or drug use, or physical abuse by a parent). We summed the responses within each cohort to create a traumatic life event score. To facilitate comparison, all social exposome components were z-standardized by subtracting their mean and dividing by the standard deviation (see Table S1 for the social exposome assessments). Cognitive function. In the HRS, cognitive function was assessed using the Telephone Interview for Cognitive Status (TICS), a validated protocol that includes immediate and delayed recall, backward counting, and serial sevens. The total score ranges from 0 to 27, with a higher score indicating better cognitive function. 32 , 33 In the ELSA, a similar cognitive battery was administered, but some domains (e.g., backward counting) were only assessed in certain waves. We therefore analyzed the domains that were consistently assessed across waves: 1) immediate recall (10 words; range: 0–10); 2) delayed recall (10 words; range 0–10), and 3) orientation, where participants answered questions about the day, month, year, and day of the week (range 0–4). These three domains were summed into a total score ranging from 0 to 24. Cognitive function scores were standardized to the baseline mean and standard deviation in each study. The HCAP neuropsychological assessment. The HCAP cognitive assessments were administered via a 1-hour in-person battery of neuropsychological tests. Domains assessed included word list learning and memory, story recall, object naming, comprehension, semantic fluency, attention, speed, set shifting, reasoning, constructional praxis, and orientation. 22 , 34 A harmonized General Cognitive Function (GCF) factor score was derived using rigorous pre-statistical and statistical harmonization procedures, with higher scores indicating better function. 35 To facilitate cross-national comparisons, the factor scores in both HCAP samples were standardized to the HRS-HCAP distribution, which itself was standardized to a mean of 0 and standard deviation of 1. Covariates. Covariates were assessed at baseline and were selected based on prior literature. 36 These included: age (in years), sex (men vs. women), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and Other in the HRS; White vs. non-White in the ELSA), education, father’s and mother’s education (in years; included in the HRS and HRS-HCAP only), urbanicity (urban, suburban, ex-urban; included in the HRS and HRS-HCAP only), and marital status (married, never married, partnered, separated/divorced, or widowed). For analyses involving social strain, we additionally adjusted for parental status (has vs. does not have children), presence of close relatives (yes vs. no), and presence of friends (yes vs. no). Across datasets, education was categorized into four groups: less than high school, high school graduate or General Educational Development (GED), some college, and college degree or higher. UK educational qualifications were converted to U.S. equivalents (see Table S2 ). In additional analyses, we adjusted for behavioral and health-related factors that could be confounders or mediators, including body mass index (BMI), depression, smoking status (current smoker or not), and frequency of vigorous physical activity. Statistical analysis All analyses were conducted separately for each sample. For the HRS and ELSA datasets, we pooled data across waves and estimated linear mixed-effects models to examine the independent associations of each exposome component with cognitive function levels, while adjusting for covariates. Each component was modelled separately without adjustment for the other components. The primary parameter of interest was the coefficient for each component, interpreted as the difference in cognitive function scores associated with a 1-unit difference in that component. As the HCAP samples only had one wave of available data, we fit linear regression models to examine associations of each exposome component with cognitive function levels. To quantify the associations of each exposome component with cognitive decline, we fitted linear mixed-effects models using time since baseline as the time scale. These models included the individual component, covariates, and their respective interactions with time. The primary parameter of interest was the coefficient of the interaction between time and each exposome component, interpreted as the annual difference in the rate of decline associated with a 1-unit difference in baseline exposure. To quantify the joint associations of all five social exposome components with cognitive outcomes and their relative contributions, we applied quantile-based g-computation with Gaussian link functions. 19 In quantile-based g-computation, each component was transformed into quantiles, and the model estimates the expected cognitive function under a hypothetical shift in all exposures by one quartile: E Y X q Z , ψ , η = g ψ 0 + ψ 1 ⋅ S q + η ⋅ Z where g is the link function, E Y X q is the expected cognitive function, Z denotes the covariates. ψ 0 is the intercept, ψ 1 is the coefficient for the joint effect, and S q is the weighted quantile sum of all components. Weights represent its proportional contribution to the joint effects, relative to other exposures in the same direction, summing to 1 for positive and −1 for negative associations. Confidence intervals were estimated using 200 bootstrap iterations. To address potential bias due to attrition in the Core samples, we calculated stabilized inverse probability weights for loss to follow-up due to death and other reasons, trimming both weights at the 99 th percentile. Missing data were handled using multiple imputations with chained equations (see Table S3 for participant characteristics by missingness status). These imputation models included all covariates, components of the social exposome, and baseline cognitive function. We generated ten imputed datasets per sample and pooled estimates using Rubin’s rules. As behavioral and health-related factors such as BMI, smoking, physical activity, and depression could act as mediators or confounders, we included them as additional covariates in sensitivity analyses. 37 – 40 We also fit models including 10 multiplicative interaction terms between exposome components in the quantile-based g-computation models. As weights are not well-defined in the presence of interaction terms, we did not interpret weights from these models. 19 Additionally, because the HCAP samples included only older adults aged 65 years or older, we conducted a sensitivity analysis that restricted the HRS and ELSA samples to participants aged 65 years or older at baseline. All analyses were conducted in R version 4.2.1, and quantile-based g-computation was implemented using the qgcomp package. 41 RESULTS Table 1 summarizes the demographic characteristics for each sample. On average, participants in the HRS and HRS-HCAP samples were older than those in their ELSA counterparts. The proportion of women was higher in the ELSA samples, while racial and ethnic diversity was greater in the HRS, with a substantially higher percentage of White participants in the ELSA. Participants in the HRS were more likely to be widowed compared to those in the ELSA. Pearson correlation coefficients among the exposome components ranged from 0.09 to 0.37 in the HRS, 0.06 to 0.37 in HRS-HCAP, 0.11 to 0.28 in the ELSA, and 0.04 to 0.29 in the ELSA-HCAP ( Figure S2 ). Table 1. Baseline demographic characteristics of each analytical sample HRS (N=13,795) ELSA (N=9,469) HRS-HCAP (N=2,749) ELSA-HCAP (N=955) Age, Mean (SD) 69.1 (10.3) 64.3 (11) 68.5 (7.3) 65.4 (7.4) Sex, N (%) Men 5496 (39.8) 4149 (43.8) 1058 (38.5) 418 (43.8) Women 8299 (60.2) 5320 (56.2) 1691 (61.5) 537 (56.2) Race/ethnicity (US), N (%) Non-Hispanic White 10597 (76.8) N/A 2124 (77.3) N/A Non-Hispanic Black 1746 (12.7) N/A 332 (12.1) N/A Hispanic 1149 (8.3) N/A 236 (8.6) N/A Other 304 (2.2) N/A 57 (2.1) N/A Race (UK), N (%) White N/A 9191 (97.1) N/A 932 (97.6) Non-White N/A 278 (2.9) N/A 23 (2.4) Education, N (%) Less than high school 2626 (19.0) 3881 (41.0) 460 (16.7) 438 (45.8) High-school graduate/GED 5038 (36.5) 1936 (20.4) 1009 (36.7) 200 (21.0) Some college 3127 (22.7) 2105 (22.2) 616 (22.4) 210 (22.0) College and above 3004 (21.8) 1547 (16.3) 664 (24.2) 106 (11.1) Father’s education, Mean (SD) 9.2 (3.9) NA 9.3 (4) NA Mother’s education, Mean (SD) 9.6 (3.5) NA 9.7 (3.7) NA Had children, N (%) No 1276 (9.2) 1284 (13.6) 220 (8.0) 114 (11.9) Yes 12519 (90.8) 8185 (86.4) 2528 (92.0) 842 (88.1) Had family members, N (%) No 806 (5.8) 704 (7.4) 127 (4.6) 57 (6.0) Yes 12989 (94.2) 8765 (92.6) 2622 (95.4) 898 (94.0) Had friends, N (%) No 1007 (7.3) 492 (5.2) 166 (6.1) 40 (4.2) Yes 12788 (92.7) 8977 (94.8) 2582 (93.9) 915 (95.8) Marital status, N (%) Married/partnered 9019 (65.4) 6705 (70.8) 1877 (68.3) 698 (73.1) Never married 400 (2.9) 452 (4.8) 76 (2.8) 47 (4.9) Separated/divorced 1536 (11.1) 855 (9.0) 314 (11.4) 78 (8.2) Widowed 2840 (20.6) 1457 (15.4) 483 (17.6) 132 (13.8) Urban residence, N (%) Urban 6365 (46.1) NA 1251 (45.5) NA Suburban 3268 (23.7) NA 668 (24.3) NA Ex-urban 4162 (30.2) NA 830 (30.2) NA BMI, Mean (SD) 28.1 (5.9) 28.4 (5.4) 28.6 (5.6) 28.6 (5.2) Depression, Mean (SD) 1.4 (1.9) 1.5 (2) 1.2 (1.8) 1.4 (1.8) Current smoking, N (%) No 12089 (87.6) 8003 (84.5) 2478 (90.2) 832 (87.1) Yes 1706 (12.4) 1466 (15.5) 270 (9.8) 123 (12.9) Physical activity, N (%) Everyday 370 (2.7) NA 67 (2.4) NA > 1 per week 2831 (20.5) 1873 (19.8) 661 (24.1) 204 (21.4) 1 per week 1164 (8.4) 876 (9.3) 286 (10.4) 115 (12.0) 1–3 per month 1042 (7.6) 918 (9.7) 222 (8.1) 80 (8.4) Never 8388 (60.8) 5802 (61.3) 1512 (55.0) 556 (58.2) Open in a new tab Note. HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; HCAP = Harmonized Cognitive Assessment Protocol. Statistics were pooled from imputed datasets. Percentages are shown next to counts for categorical variables. NA = not applicable. The characteristics of the HCAP samples were assessed during the social exposome assessment. In the HRS, higher levels of financial strain (β = −0.03, 95% CI −0.04 to −0.02), social strain (β = −0.05, 95% CI −0.06 to −0.04), perceived discrimination (β = −0.05, 95% CI −0.06 to −0.04), and neighborhood disorder ( β = −0.05, 95% CI −0.06 to −0.04) were associated with lower levels of cognitive function. Similar patterns were observed in the ELSA, where these components were also negatively associated with cognitive function ( β financial strain = −0.06, 95% CI −0.08 to −0.04, β neighborhood disorder = −0.02, 95% CI −0.04 to −0.01, β discrimination = −0.02, 95% CI −0.04 to −0.001, β social strain = −0.02, 95% CI −0.04 to −0.01) ( Figure 2 ). Unexpectedly, traumatic life events were associated with better cognitive function in both the HRS ( β traumatic life events = 0.002, 95% CI −0.01 to 0.01) and ELSA ( β traumatic life events = 0.01, 95% CI −0.01 to 0.02), although confidence intervals crossed the null. None of the social exposome components were associated with the rate of cognitive decline ( Table S4 ) Figure 2. Estimated differences in cognitive function levels associated with five key domains of the social exposome, including financial strain, neighborhood disorder, perceived discrimination, social strain, and traumatic life events, in the Health and Retirement Study (HRS), English Longitudinal Study of Ageing (ELSA), HRS-Harmonized Cognitive Assessment Protocol (HRS-HCAP), and ELSA-HCAP. Open in a new tab The bars represent the estimated differences in cognitive function for each component of social exposome, color-coded by sample. Error bars represent 95% confidence intervals. Negative values indicate that higher levels of the social exposome component are associated with lower cognitive function. Models were adjusted for age, sex, race and ethnicity, education, marital status, and the presence of children, friends, or close relatives when domains of the social exposome were reported. HRS and HRS-HCAP models were additionally adjusted for parental education and urbanicity. Analyses were conducted separately for each sample and social exposome component. Across all samples, cognitive function was standardized to baseline mean and standard deviation. For HRS and ELSA, analyses were conducted using data pooled from multiple waves. For each sample, results were pooled from ten imputed datasets. In the HRS-HCAP, consistent with the HRS, all components except traumatic life events were associated with lower levels of cognitive function ( β financial strain = −0.03, 95% CI −0.06 to −0.00001, β neighborhood disorder = −0.05, 95% CI −0.08 to −0.02, β discrimination = −0.05, 95% CI −0.08 to −0.02, β social strain = −0.04, 95% CI −0.07 to −0.01). In the ELSA-HCAP, social strain ( β social strain = −0.09, 95% CI −0.15 to −0.03) and financial strain ( β financial strain = −0.08, 95% CI −0.14 to −0.002) were associated with lower levels of cognitive function ( Figure 2 ). As we did not observe independent associations between any of the social exposome components and rate of cognitive decline, we performed quantile-based g-computation for pooled cognitive function levels only. A simultaneous one-quartile reduction in all social exposome components was associated with a 0.09 (95% CI: 0.07– 0.11) SD increase in cognitive function in the HRS, 0.13 SD in the ELSA (95% CI: 0.10–0.16), and 0.07 (95% CI: 0.02–0.13) SD in the HRS-HCAP. In terms of relative contribution, neighborhood disorder carried the largest negative weight in the HRS samples (−0.46 in the HRS, −0.34 in the HRS-HCAP), while financial strain was most influential in the ELSA (−0.66 in the ELSA, −0.52 in the ELSA-HCAP) ( Figure 3 ). Figure 3. Joint associations of key components of the social exposome with levels of cognitive function. Open in a new tab Panel A illustrates the differences in levels of cognitive function associated with changing all domains of the social exposome simultaneously by one quartile stress in the Health and Retirement Study (HRS), English Longitudinal Study of Ageing (ELSA), HRS-Harmonized Cognitive Assessment Protocol (HRS-HCAP), and ELSA-HCAP. The y-axis represents the differences in cognitive function. Error bars show 95% confidence intervals. Panel B shows the weight values representing the relative contribution of each component of the social exposome (financial strain, neighborhood disorder, perceived discrimination, social strain, and traumatic life events) to the overall joint effect on cognitive function for each cohort. Models were adjusted for age, sex, race, education, marital status, and having any children, friends, or close relatives when reporting the social exposome. The HRS and HRS-HCAP models were additionally adjusted for parental education and urbanicity. Analyses were conducted separately in each sample. For each sample, results were pooled from ten imputed datasets. The left and right sides of the weight plot should not be compared with each other because the length of the bars corresponds to the effect size only relative to other effects in the same direction. Sensitivity analyses In sensitivity analyses adjusting for behavioral and health-related factors, the independent associations between social exposome components and cognitive function were generally attenuated. However, in the HRS and ELSA, traumatic life events were positively associated with cognitive function following these adjustments ( Figure S3 ). Associations with cognitive decline were mostly null, except for neighborhood disorder in the HRS, which was associated with slower rates of decline after these additional adjustments ( Table S4 ). The magnitude of the joint association was also attenuated ( Figure S4 ). Including interaction terms between the exposome components slightly altered the estimated joint association to 0.11 SD units (95% CI: 0.07 to 0.14) in the HRS and 0.12 SD units (95% CI: 0.05 to 0.18) in the ELSA ( Figure S5 ). Finally, the findings remained largely consistent after restricting the HRS and ELSA samples to participants aged 65 years or older ( Figure S6 - 7 ). DISCUSSION This study examined the independent and joint associations of five key components of the social exposome with cognitive function across nationally representative samples in the United States (HRS) and England (ELSA), including their HCAP neuropsychological subsamples. Across samples and countries, most social exposome components were consistently associated with worse cognitive function. Effect estimates were less precise in the HCAP subsamples, and associations were attenuated after adjusting for behavioral and health-related covariates. Somewhat unexpectedly, higher levels of traumatic life events were associated with better cognitive function in most samples, though estimates were imprecise. Quantile-based g-computation estimates suggested that jointly reducing all social exposome components by one quartile was associated with better cognitive performance, with larger estimated effects in England than in the US. Neighborhood disorder contributed most to the negative joint association in the U.S., while financial strain played the strongest role in England. However, no individual component was associated with cognitive decline over time. The inverse associations of financial strain, neighborhood disorder, perceived discrimination, and social strain with level of cognitive function were generally consistent across countries and cognitive assessments. These social exposome components represent common stress exposures that may contribute to poorer cognitive function and increased risk of dementia through biological pathways, including elevated inflammatory markers (e.g., C-reactive protein (CRP) or interleukin-6 (IL-6)). 42 , 43 Additionally, they may contribute to adverse health behaviors such as poor diet, physical inactivity, and disrupted sleep, which in turn exacerbate cardiovascular conditions (e.g., diabetes and hypertension) that are established risk factors for poor cognitive function. 37 Importantly, the associations of these social exposome components with cognitive function are largely consistent across the Core and the HCAP samples. The fact that these findings hold using harmonized outcomes and in different contexts highlights the critical roles of these social exposures in the aging process. Traumatic life events were associated with higher cognitive function in the HRS, ELSA, and ELSA-HCAP, although estimates were imprecise. These findings align with prior research, including analyses from the Midlife Development in the United States Study (MIDUS) and the HRS, which observed similar associations with executive function. 3 , 44 Unlike chronic stressors such as financial strain, which tend to have sustained negative effects on cognitive function, some traumatic life events may act as acute experiences that foster post-traumatic growth and improved psychological resilience, which would break the link between traumatic experiences and subsequent poor health. 45 , 46 Alternatively, these results may reflect selection bias. Specifically, participants with greater trauma exposure and poorer cognition may have been less likely to survive to the observation period or to participate in the study. As a result, our analytic sample may overrepresent individuals who experienced fewer traumatic events or who are cognitively resilient despite such exposures. This differential selection could induce a spurious positive association between traumatic life events and cognitive function. Nevertheless, the consistency of findings across multiple samples highlights the need for further investigation into the long-term cognitive implications of traumatic experiences. We found no associations between individual social exposome components and cognitive decline in primary analyses. This aligns with recent studies using HRS and MIDUS data, which found no independent associations of stressful life events, financial strain, neighborhood cohesion, or perceived discrimination with rate of cognitive decline. 47 , 48 , 44 Despite these null associations, we believe it is too soon to dismiss the possibility of causal relationships; further research should investigate potential methodological or substantive explanations, including the possibility that social exposures may have stronger effects on cognitive decline during certain periods in the life course. Notably, after adjusting for behavioral and health-related variables, neighborhood disorder was associated with slower rates of cognitive decline. As some of these factors could lie in the causal pathways between neighborhood disorder and cognitive function, adjusting for them may account for the indirect, detrimental effect of neighborhood disorder operating through health and behaviors and result in biased estimates. Formal mediation analyses are warranted to test these pathways in future research. Although modest (range: 0.07–0.13 SD), the joint effect sizes are comparable to those of an additional year of education, which typically confers a 0.1–0.15 SD benefit to cognitive outcomes. 49 Despite the modest effect sizes, most components of the social exposome that we examined, such as financial strain and social strain, are midlife or later-life exposures. If further research does suggest exposome components are causal contributors to cognitive function, our findings would suggest that interventions could be effective in promoting cognitive health even after early childhood and formal education have concluded. The stronger influence of neighborhood disorder in the US may reflect deeper structural inequities, such as residential segregation and systemic racism, that shape cognitive health disparities. 50 Despite England’s generally more generous and comprehensive social safety nets compared to the US, 51 experiencing financial strain may be more likely to be seen as a personal failure or give rise to a deeper sense of injustice when expectations of robust social protection are unmet, which lead to greater health threats. These interpretations warrant further investigation that incorporates macro-level data. Interestingly, including interaction terms between components of the social exposome in the g-computation models modestly increased the estimated joint effect in the HRS. Although confidence intervals overlapped, this finding raises the possibility that co-occurring social exposures may interact synergistically to amplify their effects on cognitive health. This study has several limitations. First, the relatively small sample size of ELSA-HCAP limited power to detect some associations. Nonetheless, the harmonized HCAP assessments strengthened the validity of cross-national comparisons. Second, we cannot rule out the possibility of reverse causality. If lower cognitive function increases vulnerability to adverse social exposures, then reverse causation may introduce bias away from the null. Third, recall bias could affect the reporting of stressful life events, especially among participants with poorer cognition, which may lead to non-differential misclassification and bias toward the null. Additionally, we lacked precise timing of marital or parental status changes, which may have occurred after exposure to social stressors and lie on the causal pathway. Moreover, although we selected comparable measures across HRS and ELSA, some measure differences, especially in financial strain and traumatic life events, may contribute to variation in findings across countries. We used quartiles as the cut-off point, consistent with common practice in studies employing quantile-based g-computation models. However, we recognize that this choice is somewhat arbitrary. Finally, while we included key components of the social exposome comparable across studies, others such as caregiver stress and social isolation were not assessed in this analysis. Despite these limitations, this study has notable strengths. To our knowledge, it is the first to use prospective, cross-national data to estimate the joint effects of key social exposome components on cognitive outcomes. The inclusion of the HCAP samples with detailed neuropsychological assessments further strengthens the robustness of cross-national comparisons. Moreover, quantile-based g-computation offers a novel approach to estimating the joint effects of multiple, co-occurring exposures. In conclusion, most components of the social exposome were consistently associated with lower cognitive function across the US and England, with stronger joint associations observed in England. This study demonstrates the value of cross-national data and quantile-based g-computation for deepening our understanding of the complex relationships between social exposures and cognitive aging. Supplementary Material Supplemental materials NIHMS2162625-supplement-Supplemental_materials.docx (1.3MB, docx) Acknowledgement: This research was supported by a pilot grant from the University of California, San Francisco Stress Measurement Network (National Institute on Aging, R24AG048024). RC is partially funded by the National Institute on Aging through the K00 award (K00AG068431). References 1. Barnes LL, De Leon CFM, Lewis TT, Bienias JL, Wilson RS, Evans DA. Perceived discrimination and mortality in a population-based study of older adults. American journal of public health. 2008;98(7):1241–1247. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Kelly ME, Duff H, Kelly S, et al. The impact of social activities, social networks, social support and social relationships on the cognitive functioning of healthy older adults: a systematic review. Systematic reviews. 2017;6(1):1–18. doi: 10.1186/s13643-017-0632-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Stebbins RC, Maselko J, Yang YC, Plassman BL, Edwards JK, Aiello AE. Lifecourse Traumatic Events and Cognitive Aging in the Health and Retirement Study. American Journal of Preventive Medicine. 2022;63(5):818–826. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Aneshensel CS, Ko MJ, Chodosh J, Wight RG. The urban neighborhood and cognitive functioning in late middle age. Journal of health and social behavior. 2011;52(2):163–179. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Sutin AR, Stephan Y, Carretta H, Terracciano A. Perceived discrimination and physical, cognitive, and emotional health in older adulthood. The American Journal of Geriatric Psychiatry. 2015;23(2):171–179. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Ferraro KF, Zaborenko CJ. Race, everyday discrimination, and cognitive function in later life. Plos one. 2023;18(10):e0292617. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Ailshire J, Carr D. Cross-National Comparisons of Social and Economic Contexts of Aging. The Journals of Gerontology: Series B. 2021;76(Supplement_1):S1–S4. doi: 10.1093/geronb/gbab049 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Sternthal MJ, Slopen N, Williams DR. Racial disparities in health: how much does stress really matter? Du Bois Review: Social Science Research on Race. 2011;8(1):95–113. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Slopen N, Meyer C, Williams DR. Cumulative stress and health. The Oxford handbook of integrative health science. Published online 2018:75–93. [ Google Scholar ] 10. Boen C Death by a thousand cuts: Stress exposure and black–white disparities in physiological functioning in late life. The Journals of Gerontology: Series B. 2020;75(9):1937–1950. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Gudi-Mindermann H, White M, Roczen J, Riedel N, Dreger S, Bolte G. Integrating the social environment with an equity perspective into the exposome paradigm: a new conceptual framework of the Social Exposome. Environmental Research. 2023;233:116485. [ DOI ] [ PubMed ] [ Google Scholar ] 12. Chen R, Weuve J, Misra S, Cuevas A, Kubzansky LD, Williams DR. Racial Disparities in Cognitive Function among Middle-Aged and Older Adults: The Roles of Cumulative Stress Exposures Across the Life Course. The Journals of Gerontology: Series A. Published online 2021. doi: 10.1093/gerona/glab099 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Cuevas AG, Chen R, Slopen N, et al. Assessing the role of health behaviors, socioeconomic status, and cumulative stress for racial/ethnic disparities in obesity. Obesity. 2020;28(1):161–170. doi: 10.1002/oby.22648. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Cuevas AG, Chen R, Thurber KA, Slopen N, Williams DR. Psychosocial stress and overweight and obesity: findings from the Chicago community adult health study. Annals of Behavioral Medicine. 2019;53(11):NP-NP. doi: 10.1093/abm/kaz008. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Kuh D, Ben-Shlomo Y, Lynch J, Hallqvist J, Power C. Life course epidemiology. Journal of epidemiology and community health. 2003;57(10):778. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Evans GW. A multimethodological analysis of cumulative risk and allostatic load among rural children. Dev Psychol. 2003;39(5):924–933. doi: 10.1037/0012-1649.39.5.924 [ DOI ] [ PubMed ] [ Google Scholar ] 17. Letellier N, Zamora S, Yang JA, Sears DD, Jankowska MM, Benmarhnia T. How do environmental characteristics jointly contribute to cardiometabolic health? A quantile g-computation mixture analysis. Preventive Medicine Reports. 2022;30:102005. doi: 10.1016/j.pmedr.2022.102005 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Goin DE, Izano MA, Eick SM, et al. Maternal experience of multiple hardships and fetal growth: Extending environmental mixtures methodology to social exposures. Epidemiology. 2021;32(1):18–26. doi: 10.1097/EDE.0000000000001272 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Keil AP, Buckley JP, O ‘Brien Katie M., Ferguson KK, Zhao S, White AJ. A Quantile-Based g-Computation Approach to Addressing the Effects of Exposure Mixtures. Environmental Health Perspectives. 128(4):047004. doi: 10.1289/EHP5838 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort Profile: the Health and Retirement Study (HRS). International Journal of Epidemiology. 2014;43(2):576–585. doi: 10.1093/ije/dyu067 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Steptoe A, Breeze E, Banks J, Nazroo J. Cohort profile: the English longitudinal study of ageing. International journal of epidemiology. 2013;42(6):1640–1648. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Langa KM, Ryan LH, McCammon R, et al. The Health and Retirement Study Harmonized Cognitive Assessment Protocol (HCAP) Project: Study Design and Methods. Neuroepidemiology. 2020;54(1):64–74. doi: 10.1159/000503004 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Smith J, Ryan LH, Fisher GG, Sonnega A, Weir DR. HRS psychosocial and lifestyle questionnaire 2006–2016. Survey Research Center, Institute for Social Research, University of Michigan. Published online 2017:202006–202016. [ Google Scholar ] 24. Wilkinson LR. Financial strain and mental health among older adults during the great recession. Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2016;71(4):745–754. [ DOI ] [ PubMed ] [ Google Scholar ] 25. Cagney KA, Glass TA, Skarupski KA, Barnes LL, Schwartz BS, Mendes de Leon CF. Neighborhood-level cohesion and disorder: Measurement and validation in two older adult urban populations. Journals of Gerontology: Series B. 2009;64(3):415–424. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Williams DR, Yu Y, Jackson JS, Anderson NB. Racial differences in physical and mental health: Socio-economic status, stress and discrimination. Journal of health psychology. 1997;2(3):335–351. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Walen HR, Lachman ME. Social support and strain from partner, family, and friends: Costs and benefits for men and women in adulthood. Journal of social and personal relationships. 2000;17(1):5–30. [ Google Scholar ] 28. Chen Y, Feeley TH. Social support, social strain, loneliness, and well-being among older adults: An analysis of the Health and Retirement Study*. Journal of Social and Personal Relationships. 2014;31(2):141–161. doi: 10.1177/0265407513488728 [ DOI ] [ Google Scholar ] 29. Wilkinson R, Nakamura JS, Kim ES, VanderWeele TJ. Social support and social strain from children and subsequent health and well-being among older U.S. adults. Journal of Social and Personal Relationships. 2023;40(9):2980–3003. doi: 10.1177/02654075231164900 [ DOI ] [ Google Scholar ] 30. Slopen N, Dutra LM, Williams DR, et al. Psychosocial stressors and cigarette smoking among African American adults in midlife. Nicotine & Tobacco Research. 2012;14(10):1161–1169. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Krause N, Shaw BA, Cairney J. A descriptive epidemiology of lifetime trauma and the physical health status of older adults. Psychology and aging. 2004;19(4):637. [ DOI ] [ PubMed ] [ Google Scholar ] 32. Brandt J, Spencer M, Folstein M. The telephone interview for cognitive status. Neuropsychiatry Neuropsychol Behav Neurol. 1988;1(2):111–117. [ Google Scholar ] 33. Herzog AR, Wallace RB. Measures of cognitive functioning in the AHEAD Study. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 1997;52(Special_Issue):37–48. [ DOI ] [ PubMed ] [ Google Scholar ] 34. Manly JJ, Jones RN, Langa KM, et al. Estimating the Prevalence of Dementia and Mild Cognitive Impairment in the US: The 2016 Health and Retirement Study Harmonized Cognitive Assessment Protocol Project. JAMA Neurol. 2022;79(12):1242–1249. doi: 10.1001/jamaneurol.2022.3543 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Gross AL, Li C, Briceño EM, et al. Harmonisation of later-life cognitive function across national contexts: results from the Harmonized Cognitive Assessment Protocols. The Lancet Healthy Longevity. 2023;4(10):e573–e583. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Glymour MM, Manly JJ. Lifecourse social conditions and racial and ethnic patterns of cognitive aging. Neuropsychology review. 2008;18(3):223–254. [ DOI ] [ PubMed ] [ Google Scholar ] 37. Whitmer RA, Gunderson EP, Barrett-Connor E, Quesenberry CP, Yaffe K. Obesity in middle age and future risk of dementia: a 27 year longitudinal population based study. Bmj. 2005;330(7504):1360. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Scult MA, Paulli AR, Mazure ES, Moffitt TE, Hariri AR, Strauman TJ. The association between cognitive function and subsequent depression: a systematic review and meta-analysis. Psychological Medicine. 2017;47(1):1–17. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Ott A, Andersen K, Dewey ME, et al. Effect of smoking on global cognitive function in nondemented elderly. Neurology. 2004;62(6):920–924. doi: 10.1212/01.WNL.0000115110.35610.80 [ DOI ] [ PubMed ] [ Google Scholar ] 40. Parimon T, Cusack B, Rea IM, Carvalho A. Physical activity and cognitive function in individuals over 60 years of age: a systematic review. CIA. Published online April 2014:661. doi: 10.2147/CIA.S55520 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Keil A, Keil MA. Package ‘qgcomp.’ Published online 2020. Accessed February 15, 2025. https://mirrors.sjtug.sjtu.edu.cn/cran/web/packages/qgcomp/qgcomp.pdf 42. Cuevas AG, Ong AD, Carvalho K, et al. Discrimination and systemic inflammation: A critical review and synthesis. Brain, behavior, and immunity. Published online 2020. doi: 10.1016/j.bbi.2020.07.017 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Yang YC, Schorpp K, Harris KM. Social support, social strain and inflammation: Evidence from a national longitudinal study of US adults. Social science & medicine. 2014;107:124–135. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Chen R, Williams DR, Nishimi K, Slopen N, Kubzansky LD, Weuve J. A life course approach to understanding stress exposures and cognitive function among middle-aged and older adults. Social Science & Medicine. 2022;314:115448. doi: 10.1016/j.socscimed.2022.115448 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Helgeson V, Reynolds K, Tomich P. Helgeson VS, Reynolds KA, Tomich PLA meta-analytic review of benefit finding and growth. J Consult Clin Psychol 74(5): 797–816. Journal of consulting and clinical psychology. 2006;74:797–816. doi: 10.1037/0022-006X.74.5.797 [ DOI ] [ PubMed ] [ Google Scholar ] 46. Barskova T, Oesterreich R. Post-traumatic growth in people living with a serious medical condition and its relations to physical and mental health: A systematic review. Disability and Rehabilitation. 2009;31(21):1709–1733. doi: 10.1080/09638280902738441 [ DOI ] [ PubMed ] [ Google Scholar ] 47. Mitchell UA, Shaw BA, Torres JM, Brown LL, Barnes LL. The Effects of Midlife Acute and Chronic Stressors on Black–White Differences in Cognitive Decline. The Journals of Gerontology : Series B. 2023;78(12):2147–2155. doi: 10.1093/geronb/gbad143 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Choi J, Han SH, Ng YT, Muñoz E. Neighborhood Cohesion across the life course and effects on cognitive aging. The Journals of Gerontology: Series B. 2023;78(10):1765–1774. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Harrati A, Glymour MM. Lifecourse epidemiology matures: commentary on Zhang et al.“Early-life socioeconomic status, adolescent cognitive ability, and cognition in late midlife.” Social Science & Medicine. 2020;244:112645. [ DOI ] [ PubMed ] [ Google Scholar ] 50. Riley AR. Neighborhood disadvantage, residential segregation, and beyond—lessons for studying structural racism and health. Journal of racial and ethnic health disparities. 2018;5:357–365. [ DOI ] [ PubMed ] [ Google Scholar ] 51. Zaninotto P, Head J, Steptoe A. Behavioural risk factors and healthy life expectancy: evidence from two longitudinal studies of ageing in England and the US. Scientific Reports. 2020;10(1):6955. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplemental materials NIHMS2162625-supplement-Supplemental_materials.docx (1.3MB, docx) ACTIONS View on publisher site PDF (765.6 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Social determinants of cognitive aging trajectories across 39 countries#356382
• Social and Cardiovascular Risk Factors as Predictors of the Progression from Mild Cognitive Impairment to Dementia in a Large EHR Database#290754
• Loneliness and cognition in older adults: A meta-analysis of harmonized studies from the United States, England, India, China, South Africa, Mexico, and Chile.#186093
• Life Experience and Demographic Influences on Cognitive Function in Older Adults#1132931
• No matter the distance? A theoretical framework of age-diverse friendship effects#237995