Abstract
Purpose:
Evidence regarding neighborhood socioeconomic status (nSES) as an upstream determinant of cognitive outcomes has largely lacked a life-course perspective. We examined racial differences in the associations between midlife and late-life nSES and cognitive function in a cohort of 330 Black and White older Americans aged 70 + .
Methods:
General cognitive function was measured using Modified Mini-Mental State Examination up to a 15-year follow-up. Midlife (age 49–58) and late-life (age 70–79) nSES scores were z-standardized based on five census indicators of tract-level socioeconomic characteristics. Mixed-effects linear regression examined the associations between midlife and late-life nSES and cognitive function.
Results:
Higher midlife nSES was associated with higher baseline levels of cognitive function among Black (β: 3.10, 95 % CI: 0.85, 5.33), but not among White participants (β: 0.51, 95 % CI: −0.88, 1.90; p for interaction: 0.037). There were no observed associations between midlife nSES and changes in cognitive function in the overall sample or in either racial group. Late-life nSES was not associated with baseline levels of cognitive function or changes in the overall sample or either racial group.
Conclusions:
Midlife may be a critical period in which neighborhood socioeconomic exposure has a greater impact on late-life cognitive health, particularly for Black individuals.
Keywords: Neighborhood characteristics, Neighborhood environment, Cognitive function, Cognitive decline, Cognitive aging, Census tracts, Racial minorities, Health equity, Health disparities
Introduction
A growing body of literature has linked area-level or neighborhood socioeconomic conditions to the risk of cognitive decline and dementia [1–5]. Neighborhood socioeconomic status (nSES) may operate through intermediate factors, such as cardiometabolic health, physical activity, stress exposures, and exposure to environmental pollutants, in shaping later-life cognitive function [6–8]. Although a wealth of cross-sectional studies suggest that nSES is positively associated with levels of cognitive function above and beyond the effect of individual-level socioeconomic status [9,10], less is known about whether nSES is predictive of future cognitive decline. Among a few studies that have assessed cognitive function longitudinally and incorporated census-based estimates of nSES [1,2,11,12], evidence remains mixed and inconclusive. Notedly, the predominant approach in the existing literature emphasizes the effects of nSES measured at a single time point, largely due to a lack of access to residential history and longitudinal measures of neighborhood characteristics, which has posed a challenge to understanding the association between neighborhood socioeconomic conditions and cognitive health outcomes from a life-course perspective.
The risk of cognitive decline and dementia in late life is shaped by exposures and risk factors across the life course, among which midlife is recognized as a critical period because dementia pathology and many vascular risk factors, including diabetes, obesity, and hypertension, begin to emerge [13]. Neighborhood exposures during midlife may be particularly impactful as they shape opportunities for healthcare access, social connection, and health behavior modification, all of which contribute to vascular disease prevention and subsequent cognitive health. Understanding when during the lifecourse exposure to specific neighborhood conditions is most strongly related to later-life cognitive function may elucidate underlying mechanisms and point to critical windows for early intervention [14,15]. Despite a growing number of studies on this topic [3,16–20], there is little understanding of life-course exposure to neighborhood socioeconomic conditions and how they bear on later-life cognitive function.
Racial differences in the experiences of socioeconomic inequalities have been well-documented. Black individuals are more likely to be exposed to neighborhood socioeconomic disadvantages as a result of historical discriminatory land use and housing policies [21]. Black Americans also bear a disproportionate burden of poor cognitive health outcomes compared to their White counterparts [22,23]. A few studies have noted that neighborhood socioeconomic conditions may have a stronger effect on cognitive health outcomes among Black individuals than that of White individuals [1,17,24]. Consideration of effect modification by race is important to understand the extent to which exposure to neighborhood socioeconomic conditions may operate differently on cognitive health in late life across racial groups.
Using longitudinal data from the Health, Aging, and Body Composition Study and retrospectively collected residential history from LexisNexis, this study examined the associations between midlife (49–58 years) and late-life (70–79 years) neighborhood socioeconomic status and late-life longitudinal cognitive function among adults aged 70 and older, and whether these associations differed between Black and White individuals. We hypothesized that 1) higher nSES in midlife or late-life would be associated with higher baseline cognitive function and a slower rate of cognitive decline, and 2) higher nSES in midlife or late-life would have a stronger protective effect on baseline cognitive function and cognitive change among Black participants compared to White participants.
Methods
Study sample
The Health, Aging, and Body Composition (Health ABC) Study is a longitudinal cohort that consists of 3075 community-dwelling, Black and White adults aged 70–79 who resided in Pittsburgh, Pennsylvania, and Memphis, Tennessee. The cohort was established in 1997–1998 and then followed prospectively with annual clinical visits and semi-annual telephone interviews until 2013–2014. The study recruited a random sample of all age-eligible Black community residents in designated zip codes of Pittsburgh and Memphis. White participants were recruited from a random sample of Medicare beneficiaries in the same zip code areas. Participants were eligible for enrollment if they did not have functional disability or life-threatening cancers at the study baseline. Details of the study design can be found elsewhere [25]. The current study included 330 Health ABC participants from the Pittsburgh site, with data collected from the 1997–1998 visit to the 2011–2012 visit and residential histories dated back to midlife. The analytic sample came from a neighborhood substudy of the Health ABC Pittsburgh site [26], where 745 participants with a valid address in Pittsburgh at the 2006–2007 visit were selected from the original 1527 Pittsburgh participants. Participants from the Memphis site were excluded, as their residential history data were not available. Participants were excluded following the inclusion and exclusion criteria detailed in Fig. 1. The comparisons of the included and excluded samples can be found in Supplementary Tables 1a & 1b. A sample selection weight was constructed (detailed in supplementary methods) to minimize the bias of the analytic sample selection. A demonstration of the current study design and timeline is shown in Fig. 2.
Fig. 1.
The flowchart of analytic sample inclusion, Health, Aging, and Body Composition Study, N = 330.
Fig. 2.
Study design and timeline of the analytic sample, N = 330.
Measures
Outcome: general cognitive function
General cognitive function was measured by Modified Mini-Mental State (3MS) examination (range: 0–100) [27], a brief assessment of several cognitive domains including attention, verbal recall, expressive language, verbal fluency, visual construction, and reasoning. A higher score indicates better cognitive function. The 3MS was administered in 1997–1998, 1999–2000, 2001–2002, 2003–2004, 2005–2006, 2006–2007, 2007–2008, 2008–2009, 2009–2010, and 2011–2012 study visits. Individuals with a missing baseline 3MS assessment were excluded. Each participant had up to ten 3MS assessments; all their available follow-up 3MS assessments were included in the analysis (an average of 7 assessments per person). In the sensitivity analysis, we further excluded individuals with a 3MS score < 80, which is indicative of cognitive impairment [28].
Exposure: midlife and late-life neighborhood socioeconomic status (nSES)
We obtained participants’ residential histories from LexisNexis, a database that collects historical address information from a variety of public and private sources. A previous validation study found that LexisNexis can match at least 87 % of addresses recorded by the study administration and can be a useful tool for reconstructing residential history for participants in epidemiological studies [29–31]. Historical addresses were manually cleaned and prepared for geocoding. If a person had not moved, the same addresses were carried over throughout a census period. We validated historical addresses using the 2006–2007 participant-provided address documented during the clinic visit; participants were excluded if their historical addresses did not have a match with the 2006–2007 address. Historical addresses from 1976 to 1985 were geocoded and used to construct midlife nSES when participants were aged 49–58. Self-reported addresses at study baseline (1997–1998) were geocoded and used to construct late-life nSES when participants were aged 70–79. Addresses were geocoded using 1980 and 2000 Census tract boundaries from IPUMS National Historical Geographic Information System [32].
Five measures were extracted from the 1980 and 2000 Censuses: proportion of residents aged 25 + with a 4-year college degree or higher, median household income, median home values, proportion of civilians aged 16 + in the labor force, and proportion of residents with executive, managerial, or professional occupations. These measures are consistent with the operationalization of neighborhood socioeconomic status in prior literature [1,2,33]. We z-score standardized them corresponding to Allegheny County census tracts (where Pittsburgh is located). An average score for each census measure was calculated for three participants with more than one address collected during the midlife period. nSES is a composite score by z-score standardizing the sum of these five census measures across all census tracts in Allegheny County; then, standardized nSES scores were assigned to each participant. A higher score is indicative of a higher neighborhood socioeconomic status.
Covariates
We adjusted for baseline individual-level covariates, including age (in years), gender (men, women), race (Black, White), education, and individual income. Education and income were considered proxies of individual-level socioeconomic status, which may affect self-selection into the neighborhood and differential risk for adverse cognitive health outcomes [34]. Education was categorized as less than high school, high school degree, or postsecondary education. Individual income was modeled in $1000 per unit in all regression models. We also adjusted for other census-year-specific (i.e.,1980, 2000) neighborhood-level factors, including population density (a proxy of levels of urbanization) and the percent of Black residents at the census tract (a proxy of levels of racial segregation); disparities in cognitive health have been observed across areas of varied levels of urbanization and racial composition [35,36]. We did not adjust for health conditions (e.g., diabetes, hypertension) that may be predictive of cognitive health in later life, as they may lie on the pathway from nSES to cognitive function, particularly for earlier life period neighborhood exposures.
Statistical analysis
Descriptive statistics were conducted to describe characteristics of the overall sample and by race. Covariate-adjusted linear mixed-effects regression was conducted to examine the associations between midlife or late-life nSES and late-life cognitive function over a 15-year follow-up; we then further stratified the analysis by race. A random intercept of person, a random intercept of census tract, and a random slope of linear time were specified in all mixed-effects models. An unstructured covariance matrix was specified, and robust standard errors were reported. Time (coded continuously from 1 to 10) was operationalized as the number of study follow-ups. Further, a two-way interaction (time x nSES) was included to estimate baseline cognitive function and rate of cognitive change by levels of neighborhood socioeconomic status. In the overall model, a three-way interaction (time x nSES x race) was included to estimate whether the effect of nSES varies between Black and White participants. Two interaction terms with race (education x race, income x race) were also included in the overall models to improve the model fit. A p-value of < 0.05 was used to determine the statistical significance of the interaction with race.
All main analyses incorporated the sample selection weight at the person level and an inverse probability of attrition weight at the observation level. An inverse probability of attrition weight was constructed to account for bias related to loss to follow-up over the 15 years. Details of weight construction are documented in the supplementary materials. We conducted a series of sensitivity analyses. First, we re-ran regression analyses without applying weights. Second, we re-ran regression analyses after excluding individuals with a 3MS score < 80, which is indicative of cognitive impairment [28]. All analyses were conducted using Stata (19.5 StataCorp LLC, College Station, TX).
Results
Table 1 presents the baseline individual-level and neighborhood (census tract)-level characteristics of the overall sample and stratified by race. Overall, the mean age at enrollment was 73 (SD: 2.7) years. Approximately 36 % of the sample self-identified as Black. Slightly more than half of the sample (57 %) were women. Black participants had a lower baseline 3MS score (mean: 90.2, SD: 7.6) than White participants (mean: 93.2, SD: 6.1). The averaged midlife nSES score was 0.8 (SD: 0.9) in White participants and was −0.6 (SD: 0.7) in Black participants.
Table 1.
Sample characteristics at the census-tract and individual levels, Health, Aging, and Body Composition Study, N = 330
| Overall N= 330 | White Participants N= 211 | Black Participants N= 119 | P-value | |
|---|---|---|---|---|
|
| ||||
| 1980 census-tract level/midlife neighborhood characteristics | ||||
| Mean (SD) | Mean (SD) | Mean (SD) | ||
| nSES (z-score, range: −2.0–2.5) | 0.3 (1.1) | 0.8 (0.9) | −0.6 (0.7) | < .001 |
| % residents who completed college | 34.8 (19.5) | 42.0 (19.5) | 22.0 (11.0) | < .001 |
| % civilians aged 16 + in the labor force | 91.6 (5.3) | 93.9 (2.8) | 87.4 (6.2) | < .001 |
| % residents with executive, managerial, or professional occupation | 27.7 (15.5) | 33.7 (14.8) | 17.1 (10.3) | < .001 |
| Median home value (in 1979 dollars) | 44,236 (23,646) | 54044 (23152) | 26846 (11314) | < .001 |
| Median household income (in 1979 dollars) | 17,345 (6474) | 19781 (6128) | 13024 (4537) | < .001 |
| Population density | 8105 (5430) | 7097 (5385) | 9894 (5064) | < .001 |
| % Black residents | 28.6 (36.8) | 6.4 (13.4) | 67.8 (32.2) | < .001 |
| 2000 census-tract level/late-life neighborhood characteristics | ||||
| Mean (SD) | Mean (SD) | Mean (SD) | ||
| nSES (z-score, range: −1.6–2.7) | 0.1 (1.0) | 0.6 (0.9) | −0.7 (0.6) | < .001 |
| % residents who completed college | 32.8 (22.6) | 42.0 (22.3) | 16.6 (11.2) | < .001 |
| % civilians aged 16 + in the labor force | 92.0 (6.9) | 93.8 (7.0) | 16.6 (11.2) | < .001 |
| % residents with executive, managerial, or professional occupation | 41.5 (17.7) | 48.4 (17.5) | 29.2 (9.8) | < .001 |
| Median home value (in 1999 dollars) | 87108 (54951) | 106566 (57947) | 52606 (23787) | < .001 |
| Median household income (in 1999 dollars) | 36287 (15912) | 41782 (15831) | 26544 (10496) | < .001 |
| Population density | 6445 (4016) | 6229 (4515) | 6829 (2909) | .213 |
| % Black residents | 32.6 (35.4) | 12.3 (16.3) | 68.6 (31.4) | < .001 |
| Individual characteristics at baseline (1997–1998) | ||||
| Mean (SD) or n (%) | Mean (SD) or n (%) | Mean (SD) or n (%) | ||
| Cognitive (3MS) score | 92.1 (6.8) | 93.2 (6.1) | 90.2 (7.6) | < .001 |
| Age | 73.0 (2.7) | 73.1 (2.7) | 72.8 (2.7) | .314 |
| Women | 187 (57 %) | 108 (51 %) | 79 (66 %) | .007 |
| Education attainment | < .001 | |||
| Less than high school | 50 (15 %) | 21 (10 %) | 29 (24 %) | |
| High school degree | 120 (36 %) | 71 (34 %) | 49 (41 %) | |
| Post-secondary | 160 (40 %) | 119 (56 %) | 41 (35 %) | |
| Individual income | 27634 (10412) | 31852 (9543) | 20156 (7210) | < .001 |
Note. Abbreviations: nSES, neighborhood socioeconomic status; 3MS, Modified Mini-Mental State Exam.
Table 2 presents the association between midlife neighborhood socioeconomic status and late-life cognitive function. In the overall model, the variance of the random intercept was 8.95 at the census tract level and 14.70 at the person level. At baseline, approximately 20 % of the variance was attributable to census tracts and 34 % to between-person differences. Midlife nSES was not associated with baseline levels of cognitive function in the overall sample (β: 0.12, 95 % CI: −1.38, 1.62) or among White participants (β: 0.51, 95 % CI: −0.88, 1.90). Midlife nSES was positively associated with baseline cognitive function in Black participants (β: 3.10, 95 % CI: 0.85, 5.33). In other words, one standard deviation higher in midlife nSES score was associated with 3.10 points higher baseline levels of cognitive function. The association between midlife nSES and baseline cognitive function was significantly different between White and Black participants, with a greater effect for Black individuals (β for interaction: 2.60, 95 % CI: 0.17, 5.05, p = 0.037). Midlife nSES was not associated with the rate of change in cognitive function in the overall sample or in either racial group.
Table 2.
The Association Between Midlife Neighborhood Socioeconomic Status (nSES) and Late-life Cognitive Function, Health, Aging and Body Composition Study, 1997–1998 visit to 2011–2012 visit, Weighted
| Overall N = 330 (2214 observations, 120 census tracts) | White Participants N = 211 (1440 observations, 73 census tracts) | Black Participants N = 119 (774 observations, 59 census tracts) | |
|---|---|---|---|
|
| |||
| Fixed-effects parameters | β (95 % CI) | β (95 % CI) | β (95 % CI) |
| Midlife nSES | 0.12 (−1.38, 1.62) | 0.51 (−0.88, 1.90) | 3.10 (0.85, 5.33)** |
| Time | −0.03 (−0.24, 0.18) | −0.05 (−0.26, 0.16) | 0.00 (−0.49, 0.49) |
| Midlife nSES x Time | 0.01 (−0.13, 0.16) | 0.03 (−0.12, 0.17) | 0.06 (−0.51, 0.63) |
| Race (ref: White) | −3.83 (−10.97, 3.31) | ||
| Midlife nSES x Race | 2.60 (0.17, 5.05)* | ||
| Race x Time | 0.02 (−0.55, 0.59) | ||
| Midlife nSES x Race x Time | 0.05 (−0.56, 0.67) | ||
| Random-effects parameters | Estimate (95 % CI) | Estimate (95 % CI) | Estimate (95 % CI) |
| Variance (census tract) | 8.95 (2.40, 33.37) | 1.43 (0.09, 23.07) | 13.63 (3.57, 51.99) |
| Variance (time) | 0.80 (0.46, 1.40) | 0.40 (0.18, 0.87) | 1.47 (0.70, 3.08) |
| Variance (person) | 14.70 (8.40, 25.75) | 16.98 (9.46, 30.48) | 9.51 (1.30, 69.51) |
| Covariance (time, person) | −1.55 (−2.82, −0.30) | −1.36 (−2.77, 0.06) | −1.16 (−3.99, 1.67) |
| Variance (Residual) | 20.28 (10.12, 40.63) | 10.66 (4.10, 27.75) | 29.44 (15.26, 56.79) |
Note.
All models were adjusted for baseline age, gender, education level, and individual income. The overall model was further adjusted for race, race x education interaction, and race x individual income interaction.
* p < .05, ** p < .01, *** p < .001.
Table 3 presents the association between late-life neighborhood socioeconomic status and late-life cognitive function. In the overall model, the variance of the random intercept was 10.38 at the census tract level and 15.25 at the person level. At baseline, approximately 22 % of the variance was attributable to census tracts and 33 % to between-person differences. Late-life nSES was not associated with baseline levels of cognitive function in the overall sample (β: −1.13, 95 % CI: −2.90, 0.65) or in either racial group (β for White: −0.36, 95 % CI: −1.94, 1.22; β for Black: 1.54, 95 % CI: −1.95, 5.04). The racial differences in the associations were not statistically significant. Late-life nSES was not associated with the rate of change in cognitive function in the overall sample or in either racial group.
Table 3.
The Association Between Late-life Neighborhood Socioeconomic Status (nSES) and Late-life Cognitive Function, Health, Aging and Body Composition Study, 1997–1998 visit to 2011–2012 visit, N = 330, Weighted
| Overall N = 330 (2214 observations, 113 census tracts) | White Participants N = 211 (1440 observations, 76 census tracts) | Black Participants N = 119 (774 observations, 54 census tracts) | |
|---|---|---|---|
|
| |||
| Fixed-effects parameters | β (95 % CI) | β (95 % CI) | β (95 % CI) |
| Late-life nSES | −1.13 (−2.90, 0.65) | −0.36 (−1.94, 1.22) | 1.54 (−1.95, 5.04) |
| Time | −0.04 (−0.23, 0.15) | −0.05 (−0.23, 0.12) | 0.08 (−0.48, 0.63) |
| Late-life nSES x Time | 0.03 (−0.14, 0.21) | 0.05 (−0.13, 0.23) | 0.22 (−0.32, 0.76) |
| Race (ref: White) | −5.50 (−13.72, 2.71) | ||
| Late-life nSES x Race | 1.94 (−1.57, 5.44) | ||
| Race x Time | 0.10 (−0.57, 0.78) | ||
| Late-life nSES x Race x Time | 0.16 (−0.44, 0.76) | ||
| Random-effects parameters | Estimate (95 % CI) | Estimate (95 % CI) | Estimate (95 % CI) |
| Variance (census tract) | 10.38 (2.80, 38.49) | 2.54 (0.31, 21.13) | 8.77 (2.62, 29.36) |
| Variance (time) | 0.81 (0.47, 1.41) | 0.40 (0.15, 1.08) | 1.47 (0.76, 2.85) |
| Variance (person) | 15.26 (8.88, 26.22) | 15.74 (8.23, 30.12) | 21.93 (13.34, 36.05) |
| Covariance (time, person) | −1.74 (−3.06, -0.42) | −1.34 (−2.97, 0.0) | −2.33 (−5.63, 0.98) |
| Variance (Residual) | 20.27 (10.12, 40.62) | 10.66 (4.10, 27.74) | 29.41 (15.23, 56.79) |
Note.
All models were adjusted for baseline age, gender, education level, and individual income. The overall model was further adjusted for race, race x education interaction, and race x individual income interaction.
p < .05, ** p < .01, *** p < .001.
Sensitivity analyses
In the unweighted mode, results for midlife nSES remained unchanged (Supplementary Table 2a). Higher late-life nSES was associated with higher baseline levels of cognitive function (β: 3.08, 95 % CI: 0.24, 5.93) among Black participants (Supplementary Table 2b), while this association was not observed in the weighted model. The observed association may be inflated by unaccounted selection bias related to survival and selection. In the model excluding those with a 3MS score < 80 (Supplementary Tables 3a & 3b), results were not markedly different.
Discussion
Our study is among the first to examine life-course exposure to neighborhood socioeconomic status and late-life cognitive function in a cohort of Black and White adults. We found that higher midlife nSES was associated with higher baseline cognitive function among Black participants, net of individual socioeconomic status. This association was not observed in White participants. Further, midlife nSES was not associated with the rate of cognitive change in the overall sample or in either racial group. Late-life nSES was not associated with baseline levels of cognitive function or cognitive change in the overall sample or in either racial group. Together, these findings suggest that midlife nSES may be more relevant to Black participants compared to White participants in predicting levels of cognitive function in late life, but not the subsequent cognitive change.
Consistent with a few longitudinal studies [1,2,37], we found that late-life nSES was not associated with the rate of cognitive change in either racial group. Conversely, some longitudinal studies observed faster rates of cognitive decline associated with lower late-life nSES [38,39]. Prior inconsistencies in the literature may reflect substantial heterogeneity in the operationalization and timing of neighborhood SES, study populations, follow-up duration, analytic approaches, and cognitive outcome measures. Further, we did not observe an association between midlife nSES and cognitive change. A recent longitudinal study of the Atherosclerosis Risk in Communities cohort found that lower childhood nSES, instead of midlife nSES, was associated with faster rates of cognitive decline in late life [3]. The null findings on the midlife or late life nSES in relation to cognitive change may suggest that cognitive change in later life may be more sensitive to early life or lifetime cumulative neighborhood exposure compared to neighborhood exposure over a single period in later life. The null association could also be attributed to the ceiling effect in our measure of general cognitive function. Alternatively, tests of specific cognitive domains, such as executive function or episodic memory, rather than general cognitive function, may be more sensitive to decline at early stages of mild cognitive impairment or dementia [40].
Contrary to findings from previous studies [1,2,11,37,38], we found that late-life nSES was not associated with baseline cognitive function in either racial group. Midlife nSES was associated with baseline cognitive function for Black individuals, net of individual SES. Our findings suggest that compared to late-life nSES, midlife nSES may be more relevant to levels of cognitive function in late life for Black participants. The midlife period could be a critical window for the emergence of many dementia risk factors, such as hypertension and diabetes, particularly among Black participants who bear a disproportionate risk for cardiometabolic diseases compared to their White counterparts [41,42] These conditions are heavily influenced by lifestyle and the immediate neighborhood environment in which a healthy lifestyle (e.g., physical activity, healthy diet) could either be supported or hindered. The effect of midlife nSES could accumulate and compound over time to affect access to resources that are supportive of a healthy lifestyle, leading to differential risk for the onset of chronic conditions in midlife and subsequent cognitive function in late life.
It is worth noting that midlife nSES has a stronger effect on late-life levels of cognitive function among Black individuals than White individuals. Similar findings from a nationally representative cohort of older adults show that higher census tract-level affluence was associated with higher baseline levels of cognitive function, and this relationship was more pronounced among Black adults compared to their White counterparts [43]. Two other studies have also consistently observed that neighborhood socioeconomic status is more strongly associated with cognitive health outcomes among Black adults than White adults [1,19]. This pattern may suggest that White individuals may be buffered from low neighborhood socioeconomic status given their greater relative advantages across many spectrums of their lives, for example, through the control of a greater share of economic resources or access to education of better quality [44]. In contrast, Black individuals experience systemic disadvantages across societal domains (e.g., housing system, criminal legal system, educational system), which may compound the adverse effects of low neighborhood socioeconomic status on cognitive function due to structural racism [44].
Our study has several limitations. First, we may include a sample of older adults who were healthier than the general older population due to Health ABC recruitment criteria and survival bias. Second, by excluding individuals (n = 353) without historical addresses linked to the 1980 census (1976–1985), we may introduce built-in bias. In our sample, Black participants and women were more likely to have a missing address for the 1980 census linkage, largely because these groups faced systemic barriers to accessing socioeconomic resources at historical times, which may hinder the collection of address data through financial information such as credit card or loan records. Further, we did not know the length of residence at the midlife or late-life addresses; thus, the length of exposure is unknown. We were also unable to account for neighborhood changes occurring concurrently with measures of cognitive function, which may have introduced confounding bias and risk of reversed causation. Although the longitudinal design and adjustment for baseline cognition helped establish temporality, we cannot rule out self-selection into higher–SES neighborhoods, particularly among individuals with higher cognitive function. Such bias may contribute to baseline differences in cognitive function but is less likely to explain the largely null findings for cognitive change over time. Late-life nSES may be the accumulation of nSES from earlier life periods, including midlife. We were not able to estimate the effect of late-life nSES on cognitive function that is independent of midlife nSES due to collinearity concerns (correlation coefficient: 0.88). Lastly, the current study is geographically limited to Allegheny County in Southwest Pennsylvania, which limits the generalizability.
We were able to retrospectively collect participants’ historical addresses in midlife to construct a novel measure of midlife nSES. Prior studies that have validated residential history data from LexisNexis have deemed it a reliable source in epidemiological research [29–31]. By using census tracts as a proxy of neighborhoods and historical census tract boundaries, we were able to build data linkage with historical census-based measures for nSES that is standardized to county level in a population-based cohort. In addition, we utilized a longitudinal design with cognitive assessment data available up to 15 years of follow-up. The sufficient length of follow-up allowed us to observe substantial within-subject and between-subject differences in longitudinal cognitive function among a sample of older adults aged 70 + . Moreover, we had a racially balanced sample to perform race-stratified analysis and test for interactions with race. Overall, our study advanced the understanding of life-course neighborhood socioeconomic conditions in relation to cognitive function in later life and offered implications for community-level and systemic interventions that target promoting equity of access to neighborhood socioeconomic resources
Supplementary Material
Supporting information
Supplementary data associated with this article can be found in the online version at doi:10.1016/j.annepidem.2026.01.017.
Funding
This work was supported in part by the Intramural Research Program at the National Institute on Aging (NIA) of the National Institutes of Health. The Health Aging and Body Composition study was supported by NIA contracts (N01-AG-6–2101, N01-AG-6–2103, N01-AG-6–2106), NIA grant (R01-AG-028050), and National Institute of Nursing Research grant (R01-NR-012459). Addition of neighborhood measures was funded by the NIA (R21 AG054666–01 to A. Rosso).
Footnotes
Declaration of Competing Interest
We have no conflicts of interest to disclose.
CRediT authorship contribution statement
Andrea L. Rosso: Writing – review & editing, Supervision, Methodology, Funding acquisition. Tiffany L. Gary-Webb: Writing – review & editing, Supervision. C. Elizabeth Shaaban: Writing – review & editing, Supervision. Christina F. Mair: Writing – review & editing, Supervision, Methodology. Jeanine M. Buchanich: Writing – review & editing, Supervision. Greta Jianjia Cheng: Writing – review & editing, Writing – original draft, Software, Methodology, Formal analysis, Data curation, Conceptualization.
References
- [1].Rosso AL, Flatt JD, Carlson MC, et al. Neighborhood socioeconomic status and cognitive function in late life. Jun 15 Am J Epidemiol 2016;183(12):1088–97. 10.1093/aje/kwv337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Zeki Al Hazzouri A, Haan MN, Osypuk T, Abdou C, Hinton L, Aiello AE. Neighborhood socioeconomic context and cognitive decline among older Mexican Americans: results from the Sacramento Area Latino Study on Aging. Am J Epidemiol Aug 15 2011;174(4):423–31. 10.1093/aje/kwr095. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Kucharska-Newton AM, Pike JR, Chen J, et al. Association of childhood and midlife neighborhood socioeconomic position with cognitive decline. JAMA Netw Open Aug 1 2023;6(8):e2327421. 10.1001/jamanetworkopen.2023.27421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Coogan PF, Delp L, Hicks JN, et al. Neighborhood disadvantage and the incidence of dementia in US Black women. Alzheimers Dement Apr 2025;21(4):e70125. 10.1002/alz.70125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Meyer OL, Besser L, Tobias M, et al. Neighborhood socioeconomic status and segregation linked to cognitive decline (Jan-Mar) Alzheimers Dement (Amst) 2023;15(1):e12401. 10.1002/dad2.12401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Schiff MD, Mair CF, Barinas-Mitchell E, et al. Longitudinal profiles of neighborhood socioeconomic vulnerability influence blood pressure changes across the female midlife period. Health Place Jul 2023;82:103033. 10.1016/j.healthplace.2023.103033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Jimenez MP, Elliott EG, DeVille NV, et al. Residential green space and cognitive function in a large cohort of middle-aged women. JAMA Netw Open Apr 1 2022;5(4):e229306. 10.1001/jamanetworkopen.2022.9306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Ailshire J, Karraker A, Clarke P. Neighborhood social stressors, fine particulate matter air pollution, and cognitive function among older U.S. adults. Soc Sci Med Jan 2017;172:56–63. 10.1016/j.socscimed.2016.11.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Aneshensel CS, Ko MJ, Chodosh J, Wight RG. The urban neighborhood and cognitive functioning in late middle age. J Health Soc Behav Jun 2011;52(2):163–79. 10.1177/0022146510393974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Sisco SM, Marsiske M. Neighborhood influences on late life cognition in the ACTIVE study. J Aging Res 2012;2012:435826. 10.1155/2012/435826. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Vassilaki M, Aakre JA, Castillo A, et al. Association of neighborhood socioeconomic disadvantage and cognitive impairment. Alzheimers Dement Jun 6 2022. 10.1002/alz.12702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Hunt JFV, Vogt NM, Jonaitis EM, et al. Association of neighborhood context, cognitive decline, and cortical change in an unimpaired cohort. Neurology May 18 2021;96(20):e2500–12. 10.1212/WNL.0000000000011918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet Aug 10 2024;404(10452):572–628. 10.1016/S0140-6736(24)01296-0. [DOI] [PubMed] [Google Scholar]
- [14].Liu S, Jones RN, Glymour MM. Implications of lifecourse epidemiology for research on determinants of adult disease. Public Health Rev Nov 2010;32(2):489–511. 10.1007/BF03391613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Power C, Hertzman C. Social and biological pathways linking early life and adult disease. Br Med Bull Jan 1997;53(1):210–21. 10.1093/oxfordjournals.bmb.a011601. [DOI] [PubMed] [Google Scholar]
- [16].Wilson RS, Scherr PA, Bienias JL, et al. Socioeconomic characteristics of the community in childhood and cognition in old age. Exp Aging Res Oct-Dec 2005;31(4):393–407. 10.1080/03610730500206683. [DOI] [PubMed] [Google Scholar]
- [17].George KM, Lutsey PL, Kucharska-Newton A, et al. Life-course individual and neighborhood socioeconomic status and risk of dementia in the atherosclerosis risk in communities neurocognitive study. Am J Epidemiol Oct 1 2020;189(10):1134–42. 10.1093/aje/kwaa072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Aartsen MJ, Cheval B, Sieber S, et al. Advantaged socioeconomic conditions in childhood are associated with higher cognitive functioning but stronger cognitive decline in older age. Mar 19 Proc Natl Acad Sci USA 2019;116(12):5478–86. 10.1073/pnas.1807679116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Sol K, Morris EP, Lee JH, et al. Histories of neighborhood socioeconomic status contribute to race differences in later-life cognition. Alzheimers Dement May 2024;20(5):3342–51. 10.1002/alz.13786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Cheng GJ, Shaaban CE, Gary-Webb TL, Mair CF, Buchanich JM, Rosso AL. The associations between midlife neighborhood socioeconomic conditions and late-life cognitive trajectories: the mediating roles of hemoglobin A1c and blood pressure. Alzheimers Dement Dec 2025;21(12):e71002. 10.1002/alz.71002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Charles CZ. The Dynamics of Racial Residential Segregation. Annu Rev Sociol 2003;29:167–207. [Google Scholar]
- [22].Caunca MR, Odden MC, Glymour MM, et al. Association of racial residential segregation throughout young adulthood and cognitive performance in middle-aged participants in the CARDIA study. JAMA Neurol Aug 1 2020;77(8):1000–7. 10.1001/jamaneurol.2020.0860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Schwartz BS, Glass TA, Bolla KI, et al. Disparities in cognitive functioning by race/ethnicity in the Baltimore Memory Study. Environ Health Perspect Mar 2004;112(3):314–20. 10.1289/ehp.6727. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Glymour MM, Manly JJ. Lifecourse social conditions and racial and ethnic patterns of cognitive aging. Neuropsychol Rev Sep 2008;18(3):223–54. 10.1007/s11065-008-9064-z. [DOI] [PubMed] [Google Scholar]
- [25].Newman AB, Visser M, Kritchevsky SB, Simonsick E, Cawthon PM, Harris TB. The health, aging, and body composition (Health ABC) study-ground-breaking science for 25 years and counting. J Gerontol A Biol Sci Med Sci Oct 28 2023;78(11):2024–34. 10.1093/gerona/glad167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [26].Rosso AL, Harding AB, Clarke PJ, Studenski SA, Rosano C. Associations of neighborhood walkability and walking behaviors by cognitive trajectory in older adults. Gerontologist Sep 13 2021;61(7):1053–61. 10.1093/geront/gnab005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Teng EL, Chui HC. The Modified Mini-Mental State (3MS) examination. J Clin Psychiatry Aug 1987;48(8):314–8. [PubMed] [Google Scholar]
- [28].Yaffe K, Fiocco AJ, Lindquist K, et al. Predictors of maintaining cognitive function in older adults: the Health ABC study. Neurology Jun 9 2009;72(23):2029–35. 10.1212/WNL.0b013e3181a92c36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Wheeler DC, Wang A. Assessment of residential history generation using a public-record database. Int J Environ Res Public Health Sep 17 2015;12(9):11670–82. 10.3390/ijerph120911670. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Xu W, Agnew M, Kamis C, et al. Constructing residential histories in a general population-based representative sample. Am J Epidemiol Sep 15 2023. 10.1093/aje/kwad188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Jacquez GM, Slotnick MJ, Meliker JR, AvRuskin G, Copeland G, Nriagu J. Accuracy of commercially available residential histories for epidemiologic studies. Am J Epidemiol Jan 15 2011;173(2):236–43. 10.1093/aje/kwq350. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Manson S, Schroeder J, Van Riper D, et al. IPUMS National Historical Geographic Information System: Version 19.0 [dataset] Minneapolis, MN: IPUMS; 2024. [Google Scholar]
- [33].Chen Krieger N, Waterman JT, Soobader PD, Subramanian MJ, Carson SV. R. Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter?: the Public Health Disparities Geocoding Project. Am J Epidemiol Sep 1 2002;156(5):471–82. 10.1093/aje/kwf068. [DOI] [PubMed] [Google Scholar]
- [34].Krueger KR, Desai P, Beck T, et al. Lifetime socioeconomic status, cognitive decline, and brain characteristics. JAMA Netw Open Feb 3 2025;8(2):e2461208. 10.1001/jamanetworkopen.2024.61208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Glauber R Rural depopulation and the rural-urban gap in cognitive functioning among older adults. J Rural Health Sep 2022;38(4):696–704. 10.1111/jrh.12650. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Meyer OL, Besser L, Mitsova D, et al. Neighborhood racial/ethnic segregation and cognitive decline in older adults. Soc Sci Med Sep 2021;284:114226. 10.1016/j.socscimed.2021.114226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Meyer OL, Mungas D, King J, et al. Neighborhood socioeconomic status and cognitive trajectories in a diverse longitudinal cohort. Clin Gerontol Jan-Feb 2018;41(1):82–93. 10.1080/07317115.2017.1282911. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [38].Sheffield KM, Peek MK. Neighborhood context and cognitive decline in older Mexican Americans: results from the Hispanic Established Populations for Epidemiologic Studies of the Elderly. Am J Epidemiol May 1 2009;169(9):1092–101. 10.1093/aje/kwp005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [39].Vassilaki M, Aakre JA, Castillo A, et al. Association of neighborhood socioeconomic disadvantage and cognitive impairment. Alzheimers Dement Mar 2023;19(3):761–70. 10.1002/alz.12702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].Kaye JA, Swihart T, Howieson D, et al. Volume loss of the hippocampus and temporal lobe in healthy elderly persons destined to develop dementia. Neurology May 1997;48(5):1297–304. 10.1212/wnl.48.5.1297. [DOI] [PubMed] [Google Scholar]
- [41].Exalto LG, Quesenberry CP, Barnes D, Kivipelto M, Biessels GJ, Whitmer RA. Midlife risk score for the prediction of dementia four decades later. Alzheimers Dement Sep 2014;10(5):562–70. 10.1016/j.jalz.2013.05.1772. [DOI] [PubMed] [Google Scholar]
- [42].Whitmer RA, Sidney S, Selby J, Johnston SC, Yaffe K. Midlife cardiovascular risk factors and risk of dementia in late life. Neurology Jan 25 2005;64(2):277–81. 10.1212/01.WNL.0000149519.47454.F2. [DOI] [PubMed] [Google Scholar]
- [43].Bell M Neighborhood affluence and cognitive decline: moderation by Race? Soc Sci Med 2025. 10.1016/j.socscimed.2025.118595. [DOI] [PubMed] [Google Scholar]
- [44].Brown TH, Homan P. Structural racism and health stratification: connecting theory to measurement. J Health Soc Behav Mar 2024;65(1):141–60. 10.1177/00221465231222924. [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.