Environmental injustice - Neighborhood characteristics as confounders and effect modifiers for the association between air pollution exposure and cognitive function

Background: Air pollution has been associated with cognitive decline among the elderly. Previous studies have not evaluated the simultaneous effect of neighborhood-level socioeconomic status (N-SES), which can be an essential source of bias. Objectives: We explored N-SES as a confounder and effect modifier in a cross-sectional study of air pollution and cognitive function among the elderly. Methods: We included 12,058 participants age 50+ years from the Emory Healthy Aging Study in Metro Atlanta using the Cognitive Function Instrument (CFI) score as our outcome, with higher scores representing worse cognition. We estimated 9-year average ambient carbon monoxide (CO), nitrogen oxides (NOx), and fine particulate matter (PM2.5) concentrations at residential addresses using a fusion of dispersion and chemical transport models. We collected census-tract level N-SES indicators and created two composite measures using principal component analysis and k-means clustering. Associations between pollutants and CFI and effect modification by N-SES were estimated via linear regression models adjusted for age, education, race and N-SES. Results: N-SES confounded the association between air pollution and CFI, independent of individual characteristics. We found significant interactions between all air pollutants and N-SES for CFI (p-values<0.001) suggesting that effects of air pollution differ depending on N-SES. Participants living in areas with low N-SES were most vulnerable to air pollution. In the lowest N-SES urban areas, interquartile range (IQR) increases in CO, NOx, and PM2.5 were associated with 5.4% (95%-confidence interval, -0.2,11.4), 4.9% (-0.4,10.4), and 9.8% (2.2,18.0) increases in CFI, respectively. In lowest N-SES suburban areas, IQR increases in CO, NOx, and PM2.5 were associated with higher increases in CFI, namely 13.4% (1.3,26.9), 13.4% (0.3,28.2), and 17.6% (2.8,34.5), respectively. Discussion: N-SES is an important confounder and effect modifier in our study. This finding could have implications for studying health effects of air pollution in the context of environmental injustice.


INTRODUCTION 57
During the past decade, multiple investigations explored the effects of long-term exposure to 58 air pollution on cognitive function in the older population after animal models had shown effects 59 of air pollution on the central nervous system via inflammation and oxidative stress (Block &  Individuals living in disadvantaged neighborhoods are often exposed to the highest 66 concentrations of air pollutiona problem known as environmental injustice (Maantay, 2002). 67 Exposure to air pollution is considered to interact with vulnerability, producing a "triple jeopardy" 68 of low socioeconomic position, polluted environment, and impaired health (J. Ailshire,Karraker,69 addresses were mapped to the census tract reference map published by U.S. Census Bureau. 159 As the enrollment year of participants ranged from 2015 to 2020 in our EHAS dataset, and the 160 most recent data release from the ACS is 2018, participants were matched to ACS datasets 2 161 years prior to their enrollment. We employed two dimension reduction approaches to leverage 162 the information captured by the 16 N-SES indicators: principal component analysis (PCA) and k-163 means clustering analysis (KMCA). We used KMCA to generate clusters of census tracts with 164 similar characteristics, and used the Elbow method to determine the optimal number of clusters 165 ( Figure S1B) (Kodinariya & Makwana, 2013). We used PCA to generate census-tract-level PCs 166 (Abdi & Williams, 2010), which captured an additional layer of differences in N-SES within these 167 clusters. We included principal components (PCs) that explained at least 80% of the total 168 variance in the models ( Figure S1A). Briefly, we found that KMCA generated a larger 169 geographic unit of neighborhoods with relatively similar characteristics and the PCs captured an 170 additional layer of differences in N-SES within these clusters ( Figure S2). 171 172

Assessment of Individual-level Characteristics 173
We derived individual-level characteristics of participants from the EHAS online 174 questionnaire, and included age, sex, race, Hispanic ethnicity, household income, education 175 attainment, and disease history of mild cognitive impairment, Alzheimer's disease, or other 176 dementias. We classified racial identification into White, Black, and others that included 177 American Indian or Alaska Native, Asian, and Native Hawaiian or other Pacific Islander. We 178 note that the US Census includes a separate question on Hispanic ethnicity, participants who 179 self-reported as of Hispanic origin also select race from the categories above. We classified 180 continuous annual income into 3 categories: less than $50,000, $50,000-100,000, and more 181 than $100,000. We defined individual education attainment as the highest degree participants 182 reported receiving and categorized values into the following categories: high school or less, 183 some college credit, associate degree, bachelor degree, master degree, and professional or 184

Statistical Analysis 187
We summarized characteristics of the total study population and stratified by quartiles of CFI 188 to describe the potential patterns of covariates in dependence of CFI. We present the 189 distribution of long-term exposure concentrations of ambient air pollutants and the individual N-190 SES characteristics for each N-SES cluster as defined by KMCA. 191 192 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) We tested the associations between air pollutants and CFI using multiple linear regression 193 models taking the natural log of CFI as the dependent variable due to the skewed distribution of 194 CFI values. Due to the potentially complex confounding structure ( Figure S4), we used a 195 stepwise procedure to assess the association of air pollution and CFI, and then the impact of 196 confounding by N-SES on the association between air pollution and CFI. We added two 197 composite measures of N-SES into the models separately or together to assess the 198 confounding of N-SES at different geographic scales. The following single-pollutant models 199 were fitted using adjusted linear regression analyses: 200  Figure S3); and were included as continuous variables; was a 208 categorical variable with six levels (high school or less, some college credit, but no degree, 209 associate degree, bachelor's degree, master's degree, professional or doctorate degree);  was added in Models 2-5 to control for potential individual-SES confounding that may not be 220 captured by education alone ( Figure S4). We further included N-SES PCs and clusters in 221 Models 3-5, to evaluate whether N-SES was a confounder for the association between air 222 pollutants and CFI in addition to individual-level SES. N-SES was represented by two composite 223 measures at different geographic levels in our study, and we assessed the existence of residual 224 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. population had an annual household income more than $100,000, given the median household 257 income in Atlanta was $59,948. Over 70% of the study population had a bachelor's degree or 258 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 26, 2021. ; https://doi.org/10.1101/2021.10.19.21265212 doi: medRxiv preprint higher, while about 50% for the general population. The majority of participants were free of mild 259 cognitive impairment (MCI), Alzheimer's disease (AD), or other dementias. The mean age was 260 marginally higher in the third (Q2 -Q3 CFI) and fourth quartiles of the CFI distribution (Q3 -Max 261 CFI). The first quartile had the highest proportion of White, while the fourth quartile had the 262 highest proportion of racial minorities. In addition, participants with a higher CFI were more likely 263 to have a lower income and education ( Figure S5). Cluster 4 represented the highest N-SES. Cluster 3 consisted of census tracts along major 279 highways in northern Atlanta. As for ambient air pollution, participants living in clusters with a 280 lower level of N-SES (C6 and C7) were generally exposed to higher concentrations of air 281 pollution at their residences; Clusters 6 and 7 had the second and third highest median air 282 pollution concentrations. For those living in Cluster 3, residential air pollution exposures were 283 also high due to the proximity to highways. 284 285

Air pollution, N-SES and cognitive functioning 286
When assessing associations of air pollution and CFI, we observed that higher exposure to 287 ambient PM2.5 was associated with better cognitive function when only adjusting for age, 288 individual-level education, and race [e.g., -2.2% (95% CI, -3.9, -0.4%) CFI per 1. In addition, we found effect modification by N-SES clusters for the association between air 296 pollution and cognitive functioning, with statistically significant interaction terms for all three air 297 pollutants (CO, p = 0.0001; NOx, p = 0.0001; PM2.5, p < 0.0001; Figure 2B and Table S2). The 298 strongest association between air pollution and cognitive functioning was found in participants 299 assigned to N-SES Cluster 7, the cluster characterized by a high proportion of Hispanic and 300 immigrant communities and high air pollution concentrations. Among participants living in 301 Cluster 7, IQR increases in CO, NOx, and PM2.5 were associated with 13.4% (95% CI, 1. increases of CFI, respectively. Among participants living in Cluster 4, the cluster with the highest 307 N-SES, IQR increases in CO, NOx, and PM2.5 were associated with 6.3% (95% CI, 2.1, 10.8%), 308 7.2% (95% CI, 2.9, 11.7%), and 7.3% (95% CI, 2.2, 12.6%) increases of CFI, respectively. We 309 did not observe significant associations of air pollution and CFI for Clusters 3 or 5. Cluster 2 was 310 the only cluster in which we found an inverse association between air pollution and cognitive  Figure S6. All associations and effect modifications were robust to additional 319 adjustment for residential stability. 320

DISCUSSION 321
In the present study of 14,404 individuals 50 years and older from Metro Atlanta, we 322 observed a significant effect modification by N-SES for the association between air pollution and 323 cognitive function, and significant associations between air pollution exposure and cognitive 324 decline among participants living in the most socioeconomically disadvantaged neighborhoods. 325 In addition, our study demonstrates the importance of including N-SES as confounding variable 326 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 26, 2021. ; https://doi.org/10.1101/2021.10.19.21265212 doi: medRxiv preprint when analyzing associations between air pollution and cognitive function. Not adjusting for N-327 SES resulted in a significant negative association between the long-term exposure to air 328 pollution and CFI scores, which indicated a likely "erroneous" protective effect of air pollution, 329 despite adjusting for individual age, education attainment, and race. These findings show the 330 importance of including neighborhood characteristics as confounders and effect modifiers when 331 analyzing associations between air pollution and cognitive function. proposed that late-life cognition was a result of the interaction between individual characteristics 394 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 26, 2021. ;

CONCLUSIONS 430
This study emphasizes the necessity of considering N-SES as confounder and effect 431 modifier when investigating the effect of ambient air pollution on cognitive function among the 432 elderly. We demonstrated that N-SES and the long-term exposure to ambient air pollution could 433 have a joint impact on cognitive function, and participants living in disadvantaged 434 neighborhoods might be more susceptible to ambient air pollution than those living in other 435 areas. This study identifies potentially vulnerable populations, those exposed to both relative 436 higher concentrations of air pollution and disadvantaged social conditions, who may be at an 437 CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) islander, and multi-race. 583 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Cluster 5 Cluster 6 Cluster 7

C.
Percentage (%) 0 100 A. B. Figure 1. Statistics of Air pollution concentrations and neighborhood socioeconomic indicators for the clusters of census tracts. A.