Association of Childhood Exposure to School Racial Segregation with Late-Life Cognitive Outcomes among Older Americans

Abstract

Importance: Disparities in cognition, including dementia occurrence, persist between White and Black older adults, and are possibly influenced by early educational differences stemming from structural racism. However, the relationship between school racial segregation and later-life cognition remains underexplored.

Objective: To investigate the association between childhood contextual exposure to school racial segregation and cognitive outcomes in later life.

Design setting and participants: Data from 16,625 non-Hispanic White (hereafter, White) and 3,335 non-Hispanic Black (hereafter, Black) Americans aged 65 or older were analyzed from the Health and Retirement Study.

Exposures: State-level White-Black dissimilarity index for public elementary schools in the late 1960s (range: 0-100) was used to measure school segregation. States were categorized into high segregation (383.6) and low segregation (<83.6) based on the top quintile.

Main outcomes and measures: Cognitive scores, cognitive impairment (with or without dementia), and dementia were assessed using the Telephone Interview for Cognitive Status (TICS) and proxy assessment. Multilevel regression analyses were conducted, adjusting for demographic covariates, socioeconomic status, and health factors. Stratified analyses by race were performed.

Results: The mean (SD) age of participants was 78.5 (5.7) years, and 11,208 (56.2%) were female. Participants exposed to high segregation exhibited lower cognitive scores (12.6 vs. 13.6; P<0.001) and higher prevalence of cognitive impairment (50.8% vs 41.4%; P<0.001) and dementia (26.0% vs. 19.5%; P<0.001), compared to those with low segregation exposure. Multilevel analyses revealed a significant negative association between school segregation and later-life cognitive even after adjusting sequentially for potential confounders, and these associations were stronger among Black than White participants. Notably, in the fully adjusted model, Black participants exposed to high segregation displayed significantly lower cognitive scores (-0.51; 95% CI: -0.94, -0.09) and higher likelihood of cognitive impairment (adjusted Odds Ratio [aOR]: 1.45, 95% CI: 1.22, 1.72) and dementia (aOR: 1.31, 95% CI: 1.06, 1.63).

Conclusions and relevance: Our study underscores that childhood exposure to state-level school segregation is associated with late-life cognition, especially for Black Americans. Given the rising trend of school segregation in the US, educational policies aimed at reducing segregation are crucial to address health inequities. Clinicians can leverage patients' early-life educational circumstances to promote screening, prevention, and management of cognitive disorders.

Keywords: School racial segregation; cognition; cognitive impairment; dementia; early life.

Figure 1.

Flow chart of sample selection process Notes: HRS=Health and Retirement Study.

Figure 2.

Relationship between school segregation and cognitive outcomes by race Notes: The figure presents scatterplots of US states demonstrating the inverse relationships between school segregation (measured by White-Black dissimilarity index) and cognitive function, with dissimilarity index on the x-axis and adjusted cognitive outcomes on the y-axis. The scatterplots are stratified by White (in gray color) and Black (in black color) participants. White refers to non-Hispanic White, and Black refers to non-Hispanic Black. The average cognitive outcomes were estimated respectively for White and Black participants in each state after adjusting for age and sex; and only states with more than 10 observations are plotted. The fitted lines (with 95% CI) denote the linear relationship between school segregation and adjusted average cognitive outcomes for all participants (in blue color), White participants (in gray color), and Black participants (in black color). Chow cross-equation tests were performed to denote if there were statistically significant differences in fitted slopes between White and Black participants for each cognitive outcome, and P-values were displayed at the bottom of the corresponding panel.

Figure 3.

Association between school segregation and cognitive score by race Notes: Multilevel regression models were used to estimate the association between school segregation and cognitive score for the overall sample (in blue color) and for White (in gray color) and Black participants (in black color), respectively. White refers to non-Hispanic White, and Black refers to non-Hispanic Black. Horizontal lines represent the 95% confidence interval, and P-values are displayed alongside each lines to denote statistical significance. Covariates were added sequentially into the regression model. Model A adjusted for demographic covariates, including age, sex, and race. Model B additionally adjusted for life course socioeconomic status (SES), including educational attainment, wealth level, and Medicare enrollment, Medicaid enrollment, VA enrollment, private health insurance coverage, and employer-based health insurance coverage. Model C further adjusted for health factors over the life course including hypertension, diabetes, heart diseases, stroke, obesity, ADL and IADL functional limitations, and smoking behaviors. Random intercepts for each childhood states of residence were included in the models to address unobserved heterogeneity and differences between states. The numerical estimates are presented in Supplementary eTables 1–2. Asterisks denote the statistical significance of the association: *** P < 0.001, ** P < 0.01, * P < 0.05.

Figure 4.

Association between school segregation and cognitive impairment and dementia by race Notes: Multilevel regressions were used to estimate the association between school segregation and cognitive impairment (left panel) and dementia (right panel), for overall sample (in blue color), as well as for White (in gray color) and Black participants (in black color) respectively. White refers to non-Hispanic White, and Black refers to non-Hispanic Black. Horizontal lines represent the 95% confidence interval, and P-values are displayed alongside each lines to denote statistical significance. Covariates were added sequentially into the regression model. Model A adjusted for demographic covariates, including age, sex, and race. Model B additionally adjusted for life course socioeconomic status (SES), including educational attainment, wealth level, and Medicare enrollment, Medicaid enrollment, VA enrollment, private health insurance coverage, and employer-based health insurance coverage. Model C further adjusted for health factors over the life course including hypertension, diabetes, heart diseases, stroke, obesity, ADL and IADL functional limitations, and smoking behaviors. Random intercepts for each childhood states of residence were included in the models to address unobserved heterogeneity and differences between states. The numerical estimates are presented in Supplementary eTables 1–2. Asterisks denote the statistical significance of the association: *** P < 0.001, ** P < 0.01, * P < 0.05.