Differences in socio-demographic and environmental exposures at home vs. school in the ECHO cohort

Abstract

BACKGROUND:

Social and environmental exposures, typically characterized at residential locations, are linked to adverse respiratory outcomes in children; yet, children spend a significant amount of time in school. Little is known about how sociodemographic and environmental exposures vary between home and school settings.

OBJECTIVES:

The objectives of this study were to compare home and school exposures and to assess whether these differences were consistent across cities and varied by home-school distance.

METHODS:

Home and school addresses were geocoded for 6–13 year olds (n = 923) from the Environmental influences on Child Health Outcomes (ECHO) consortium. Children were drawn from cohorts located in New York City (NYC), Boston, Cincinnati, Baltimore, and St. Louis. Four categories of exposures were assessed: neighborhood socio-demographics (assigned at the censustract level), ambient pollution (fine particulate matter (PM^2.5); 1 × 1 km grids), traffic-exposure (primary/secondary road proximity, traffic/truck density; 400 m-buffer), and green-space (Enhanced Vegetation Index; 500-m, 1500-m, and 2500-m buffer).

RESULTS:

Overall, neighborhood socio-demographic characteristics were less favorable near home vs school including median household income ($58,125 vs $63,315), poverty (7.8% vs 5.9%), percentage of residents with less than high school education attainment (9.8% vs 8.4%), households receiving public assistance (7.6% vs 6.0%; p < 0.002 for all). Home-school differences were greatest for poverty and receiving public assistance in Baltimore (median: 48.9% and 39.3%) and St. Louis (15.8% and 15.7%), whereas the other cities showed no differences (median values of 0). Overall, children who lived farther from school (≥3.4 km), had greater differences in neighborhood socio-demographics (near vs far median percent difference: 0 vs 16.8% less than high school education, 0 vs 43.1% receiving public assistance) with the largest distance-related percent differences observed in NYC and Baltimore. No meaningful home-school differences were observed in particulate pollution, traffic-exposure, green-space (e.g., home vs school PM_2.5: 9.8 vs 9.9 μg/m³) regardless of the distance between home and school.

IMPACT:

• Children within our longitudinal birth cohorts experienced less optimal socio-demographic conditions at home compared to school, particularly those from low-income families and those traveling further to go to school; however, outdoor exposures such as PM_2.5, traffic, and green-space were similar. Our findings suggest that interventions to reduce environmental health inequities may need to address social and structural disparities at home, such as housing quality and neighborhood socioeconomic status (SES). Additionally, schools may offer an opportunity to provide beneficial exposures, potentially buffering some of the adverse SES-related exposures experienced at home. Overall, our findings provide rationale for a holistic assessment of the social and environmental exposure profile in future environmental health research.

INTRODUCTION

Environmental exposures and social conditions are well-established contributors to respiratory impairment and disease in children [1]. Most epidemiologic studies of air pollution estimate exposures based on participants’ home addresses [2]. However, school-aged children spend over 25% of their time at school, which is often located near major traffic sources such as highways and truck routes, especially in urban areas like New York City (NYC) [3, 4]. Failing to account for pollution exposure at school may lead to significant misclassification of children’s total environmental exposure. Additionally, exposure to traffic-related air pollutants (TRAP) such as black carbon (BC) near schools has been associated with airway inflammation among elementary and middle school children, suggesting that the school environment may be an important source of exposure to TRAP, leading to adverse respiratory health outcomes and therefore be an important target for intervention [5].

Across the United States (US), communities with higher percentages of Black and Hispanic residents and those with lower socioeconomic status (SES) tend to experience greater exposure to ambient pollutants like fine particulate matter (PM_2.5) than predominantly White or more affluent neighborhoods [6–9]. Socio-demographic characteristics such as SES, race, and ethnicity are thought to be among the most significant predictors of disparities in asthma prevalence, incidence, and morbidity across the US [10–13]. Race and ethnicity themselves are not causal biological risk factors for adverse respiratory outcomes; rather, structural and social determinants that shape the built environment, and that lead to disparate environmental exposures, are often stratified along lines of race and ethnicity due to systemic racism [9, 11]. For example, historical redlining, the place-based, government sponsored discriminatory housing practice of the 1930s has lead to concentrated sources of pollution (e.g., proximity to highways and primary roads, higher density of industrial emission sources, and diesel exhaust particle (DEP) emissions) in low-income and racially/ethically marginalized communities, contributing to disproportionate exposures and adverse respiratory outcomes [14–18]. Built environment features such as exposure to traffic, point sources and crowded living quarters in densely populated buildings represent an intersection between the physical and social environments that contribute to asthma morbidity [10, 19]. Additionally, greenspace has been linked to lower PM_2.5 and other air pollutants [20], and urban greenness or residential proximity to green-spaces has been inversely associated with respiratory symptoms like wheezing [21]. While these exposures are well-documented, little is known about how sociodemographic and environmental risks differ between school and home settings. Distance between home and school may influence the extent to which home and school environments differ, and SES may modify this relationship, due to structural inequities in neighborhood quality and unequal access to child opportunities.

The objectives of this study were to: 1) compare exposures at home and school environments across four categories: neighborhood socio-demographics, air pollution, traffic-exposure, and green-space; 2) examine whether patterns of these exposure differences between home and school are consistent across different cities, 3) assess whether home-school exposure differences vary as a function of the distance between them; and 4) evaluate how family SES, measured as individual household income, influences differences in environmental exposures between home and school among school-aged children in urban cohorts. We hypothesized that: 1) exposure levels would differ between home and school, especially for children attending school farther from home; 2) the magnitude and direction of these differences would vary by cohort and city, reflecting geographic and socio-environmental context, and 3) children from lower-income families would experience greater disparities in environmental exposures, specifically, worse environmental conditions at home compared to school, than children from higher-income families. This may reflect families’ efforts to seek better opportunities for their children that may be associated with better social and environmental conditions.

METHODS

Study population

From the larger Environmental influences on Child Health Outcomes (ECHO) consortium, we included 1280 children age 6–13 years from four cohorts for whom home and school address data were available. These cohorts were the Cincinnati Childhood Allergy and Air Pollution Study (CCAAPS), Columbia Center for Children’s Environmental Health (CCCEH), Urban Environment and Childhood Asthma (URECA), and Project Viva. Detailed descriptions of each cohort have been published previously [22–27]. Briefly, key eligibility criteria included: CCAAPS-having at least one parent with atopy and residing either near (<400 m) or far (>1500 m) from a major highway in Cincinnati, OH; CCCEH-residence in Northern Manhattan or the South Bronx, NYC; URECA-having at least one parent with an allergic disease or asthma and living in central urban areas of Baltimore, Boston, NYC, or St. Louis where at least 20% of the population lives below the poverty line; and Project Viva-singleton gestation with an English-speaking mother receiving prenatal care in Eastern Massachusetts. Summary characteristics for each cohort are presented in Supplementary Table S1. Eighty-six percent of children in the CCCEH and URECA cohorts were from low-income families, defined as having an annual household income below $40,000. In contrast, over 70% of children in the CCAAPS and Viva cohorts were from higher-income families (Supplementary Table S1). We chose $40,000 as the cutoff for low vs higher- household income based on two considerations: 1) it is slightly above the 2024 Department of Health and Human Services (HHS) poverty guideline of $31,000 for a family of four, and 2) it was the most consistently reported annual household income category across all our cohorts, allowing for comparable classification of families in each cohort. Based on these differences, we further classified the cohorts into two categories for secondary analysis: lower-income (CCCEH and URECA), which represented participant who lived mostly in urban neighborhoods and higher-income (CCAAPS and Viva) cohorts, which represented participants who lived mostly in suburban neighborhoods.

Estimation of home and school social and environmental exposures

Parents or guardians provided home and school addresses via questionnaire. We used the Decentralized and Reproducible Geomarker Assessment for Multi-Site Studies (DeGAUSS) program to geocode participants’ home and school addresses and to derive geospatial data, including estimates of environmental exposures, without sharing protected health information across institutions. DeGAUSS enables multi-site studies to perform standardized and reproducible offline geocoding and geospatial linkages at individual sites before returning de-identified data to a central site for analyses [28].

We assessed four categories of environmental exposures at each participant’s home and school locations. Exposure Category 1: Neighborhood socio-demographic characteristics were derived from the American Community Survey (ACS) 5-year estimates and matched to each address location at the census-tract level. ACS data were selected to correspond to the approximate time of data collection for each cohort between 2010 and 2020. These variables were selected based on their established relevance in the literature as key indicators of social and environmental disadvantage [29]. Neighborhood variables included total population, average household size, median household income, percentage of residents living below the federal poverty line, percentage identifying as Black or African American, percentage identifying as Hispanic or Latino, percentage identifying as White, percentage without high school diploma, percentage of households receiving public assistance (e.g., food stamps/Supplemental Nutrition Assistance Program (SNAP)), and percentage unemployed (e.g., weeks not worked in the past year). Exposure Category 2: Air pollution (i.e., fine particulate matter (PM_2.5)) was estimated using a satellite-based spatiotemporal model that provides daily averages at a 1 × 1 km resolution across the U.S. [30], that we further aggregated into annual averages. Exposure Category 3: Traffic-exposure data were obtained from the 1500 m = 878 due to unavailable data. Environmental Protection Agency’s [31] EJSCREEN. These measures included proximity to primary and secondary roads (km), as well as traffic and truck density, defined as the average daily number of vehicles and trucks multiplied by the lengths of major roads (i.e. primary and secondary roads). Values were calculated within a 400-meter buffer around each address location based on prior research to capture near-roadway exposure [32]. Exposure Category 4: Green-space was assessed using the average Enhanced Vegetation Index (EVI) [33] within 500-meter, 1500-meter, and 2500-meter buffer around each address. EVI was measured from June 10 to June 25, 2018 and ranges from −0.2 to 1 with higher values indicating greater vegetation. Although the 500-meter and 1500-meter buffers had 5–9% missing values, largely due to very little vegetation (e.g., buildings, roads, or water) or unavailable pixels at the finer spatial scale, they were retained in the analysis to provide a more comprehensive assessment of green-space exposure at finer spatial scales.

Data analyses

Analyses were restricted to participants with complete data on home and school addresses information, environmental exposure estimates, distance between home and school, and household income data, resulting in a final sample of 923 children (n = 357 missing one or more of the above). Data analyses included summary statistics, as well as chi-square, Mann-Whitney, Kruskal–Wallis and Wilcoxon rank-sum test, Spearman correlation, and Biserial correlation, as appropriate. Given geographic variability in urban design that may influence differences between home and school environments we explored city level effects by regrouping the data based on recruitment city. For example, the URECA cohort includes participants from four urban cities: New York, Boston, Baltimore, and St. Louis. To assess city-level effects, we combined data from URECA-New York with CCCEH (also based in New York City) and merged URECA-Boston with participants from Project Viva (also based in Boston). We classified children as living far from school if the Euclidean distance (as opposed to traveled distance) between home and school was greater than the upper tertile (3.4 km) for all 4 cohorts combined. We additionally applied city-specific thresholds based on the upper tertile within each city (2.8 km for New York, 2.5 km for Boston, 3.6 km for Cincinnati, 4.4 km for Baltimore, and 6.7 km for St. Louis). Participants were also categorized as lower-income or higher-income based on a parent-reported individual annual household income threshold of $40,000. Individual-level SES (lower vs higher-income) was further compared with neighborhood-level SES (median household income) to determine how closely individual characteristics reflected the socioeconomic conditions of the surrounding neighborhoods.

We calculated percent differences (% Diff) of social/environmental exposures between school and home as:

$$\frac{\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\left(\mathrm{h}\mathrm{o}\mathrm{m}\mathrm{e}\right)-\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\left(\mathrm{s}\mathrm{c}\mathrm{h}\mathrm{o}\mathrm{o}\mathrm{l}\right)}{\mathrm{A}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{g}\mathrm{e}\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\left(\mathrm{h}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{s}\mathrm{c}\mathrm{h}\mathrm{o}\mathrm{o}\mathrm{l}\right)}\mathrm{*}100$$

where a higher % Diff indicates greater exposure at home compared to school.

To address the issue of multiple testing in our analyses of environmental exposures, we applied a Bonferroni-correction across the set of exposures to adjust the significance threshold (Bonferroni corrected p-value = 0.002). All analyses were performed using SPSS version 29 (Chicago, IL, USA).

RESULTS

Overall differences in home-school exposures

Participants included (n = 923) in the study were similar to excluded participants (n = 357) in most demographic characteristics but were slightly older (9.1 vs 8.9 years; p = 0.01), more likely to be Black non-Hispanic and less likely White non-Hispanic (p < 0.001), and more likely to be classified as low-income (43.4% vs 30.1%, p < 0.001). On average, children were exposed to less optimal socio-demographic conditions near home compared to near school (Table 1). Median household income was lower near home (home vs school [median]: $58,125 vs $63,315), while percent below the poverty level (7.8% vs 5.9%), percentage without a high-school diploma (9.8% vs 8.4%), and households receiving public assistance (7.6% vs 6.0%) were higher near home compared to school (p < 0.002). Estimated PM_2.5 was statistically lower near home compared to school (home vs school [median]: 9.8 vs 9.9 μg/m³, 0% difference), though the difference was not meaningful. No significant differences were observed in traffic-exposure (e.g., distance to the nearest secondary road, home vs school [median]: 0.79 km vs 0.73 km) or green-space (e.g., EVI within 500 m: 0.43 vs 0.42) between home and school using multiple measures for each exposure (Table 1. p ≥ 0.002 for all).

Table 1.

Comparison of social/environmental exposure levels (median [IQR] presented) between home and school (N = 923).

Exposure category	Exposure variable	Home	School	P-valuea
Neighborhoods socio-demographicsb	Total population, counts	4946 [2776]	4653 [2575]	<0.001
	Average household size, counts/household	2.6 [0.52]	2.5 [0.57]	<0.001
	Median household income, $	58,125 [42,935]	63,315 [48,238]	<0.001
	Poverty, %	7.8 [18.4]	5.9 [16.4]	<0.001
	Less than high school education attainment, %	9.8 [15.0]	8.4 [13.1]	<0.001
	Receiving public assistance, %	7.6 [23.2]	6.0 [18.1]	<0.001
	Unemployed, %	19.5 [12.3]	18.4 [12.2]	0.08
	Black, non-Hispanic, %	11.4 [45.9]	8.8 [33.6]	<0.001
	Hispanic or Latino, %	2.2 [8.2]	2.2 [6.6]	0.18
	White, non-Hispanic, %	72.6 [69.2]	79.9 [62.0]	<0.001
Air pollution	PM2.5, μg/mc	9.8 [4.0]	9.9 [3.6]	<0.001
Traffic exposurec	Distance to the nearest primary road, km	1.7 [2.5]	1.6 [2.4]	0.21
	Distance to the nearest secondary road, km	0.79 [1.2]	0.73 [1.2]	0.15
	Average daily traffic density, counts*km	7130 [27,452]	8201 [21,102]	0.46
	Average daily truck density, counts*km	158 [948]	153 [687]	0.02
Green spaced	EVI within 500 m	0.43 [0.16]	0.42 [0.14]	0.002
	EVI within 1500 m	0.43 [0.14]	0.43 [0.13]	0.48
	EVI within 2500 m	0.43 [0.16]	0.43 [0.16]	0.09

IQR Interquartile range, PM_2.5 fine particulate matter, EVI Enhanced Vegetation Index.

Bold values represent statistically significant differences after Bonferroni correction.

^aMann–Whitney test performed.

^bNeighborhood demographics assessed at the census tract level between 2010 and 2020.

^cTraffic exposure measured within a 400-meter buffer around address.

^dGreen space calculated within a 500-meter, 1500-meter and 2500-meter buffer around address; n for EVI within 500 m = 839 and n for EVI within

1500 m = 878 due to unavailable data.

City-specific patterns in home-school % differences in socio-demographic and environmental exposures are presented Table 2. Participants from Baltimore and St. Louis (both from the URECA cohort) generally experienced less favorable socio-demographic conditions at home compared to school. For example, in Baltimore there were notable home vs school percent differences in neighborhood poverty (median: 48.9% difference), median household income (−27.3% difference) and those receiving public assistance (39.8% difference), with less favorable conditions near home. Similar results were noted in St. Louis (poverty: 15.8% difference; median household income: −5.9% difference; public assistance: 15.7% difference). In contrast, participants from Cincinnati (CCAAPS cohort) were exposed to higher PM_2.5 and green space at home compared to school ([median] 3.2% difference for PM_2.5, 4.9% difference for EVI within 500 m).

Table 2.

Percent differences (% difference in median [IQR]) in social/environmental exposures between home and school, stratified by city.

Exposure category		New York City (N = 153)	Boston (N = 271)	Cincinnati (N = 378)	Baltimore (N = 45)	St. Louis (N = 76)	p-valuea
Neighborhoods sociodemographicsb	Total population	12.3 [71.0]	0 [8.1]	0 [46.7]	0 [76.0]	0 [69.4]	<0.001
	Average household size	0 [20.0]	0 [3.8]	0 [14.0]	0 [24.5]	6.0 [18.0]	0.008
	Median household income	0 [57.2]	0 [10.6]	0 [39.8]	−27.3 [63.1]	−5.9 [67.1]	<0.001
	Poverty	0 [79.7]	0 [50.8]	0 [115.8]	48.9 [130.6]	15.8 [125.8]	<0.001
	Less than high school education attainment, %	7.6 [69.0]	0 [24.8]	0 [62.4]	14.0 [56.8]	0 [124.5]	0.007
	Receiving public assistance	2.3 [66.6]	0 [44.4]	0 [109.6]	39.3 [102.2]	15.7 [95.3]	<0.001
	Unemployed	0 [44.7]	0 [11.9]	0 [32.3]	14.8 [53.5]	0 [62.9]	0.002
	Black, non-Hispanic	0 [85.3]	0 [57.3]	0 [109.5]	15.6 [116.6]	8.1 [71.3]	<0.001
	Hispanic or Latino	1.1 [50.9]	0 [41.5]	0 [168.1]	−8.9 [169.3]	−0.6 [192.4]	0.005
	White, non-Hispanic	0 [80.2]	0 [7.8]	0 [6.5]	−47.7 [117.2]	−27.7 [126.4]	<0.001
Air pollution	PM2.5	−0.3 [4.7]	0 [4.7]	3.2 [5.7]	0 [0.1]	0 [0.8]	<0.001
Traffic exposurec	Distance to the nearest primary road	−1.6 [124.3]	0 [57.1]	−7.0 [100.9]	−11.4 [58.0]	−12.0 [112.2]	0.24
	Distance to the nearest secondary road	0 [104.4]	2.2 [142.1]	10.9 [138.5]	−19.4 [162.7]	−0.6 [143.5]	0.06
	Average daily traffic density	2.3 [105.9]	0 [145.6]	0 [226.5]	43.2 [131.0]	0 [223.7]	0.10
	Average daily truck density	2.4 [182.8]	0 [153.3]	0 [0]	48.1 [203.8]	0 [281.7]	0.005
Green spaced	EVI within 500 m	0 [16.9]	0.9 [23.8]	4.9 [23.1]	0 [59.1]	−0.9 [48.9]	0.001
	EVI within 1500 m	0 [43.8]	0.4 [13.0]	0.9 [14.4]	−4.4 [25.8]	−1.1 [27.7]	0.33
	EVI within 2500 m	−1.5 [32.4]	−0.1 [10.3]	0.9 [9.5]	1.1 [22.3]	0.3 [19.2]	0.06

New York City (NYC) participants from CCCEH and URECA were combined and Boston participants from Project Viva and URECA were combined.

IQR Interquartile range, PM_2.5 fine particulate matter, EVI Enhanced Vegetation Index.

Bold values represent statistically significant differences after Bonferroni correction.

^aKruskal–Wallis test performed to compare the distribution of %differences on home and school exposure across cities.

^bNeighborhood demographics assessed at the census tract level between 2010 and 2020.

^cTraffic exposure measured within a 400-meter buffer around address.

^dGreen space calculated within a 500-meter, 1500-meter and 2500-meter buffer around address; n for EVI within 500 m = 839 and n for EVI within 1500 m = 878 due to unavailable data.

Overall, home and school values for socio-demographic characteristics and green-space were moderately correlated (Supplementary Table S2 and Supplementary Fig. S1; Spearman correlation coefficients r = 0.41–0.82). PM_2.5 showed strong correlation (r = 0.95), while traffic exposure showed low to moderate correlations (0.34 ≤ r ≤ 0.66). Correlations were weaker among children attending school farther from home (0.19 ≤ r ≤ 0.67), compared to those living closer to school (0.45 ≤r ≤ 0.93). An exception was PM_2.5 that remained strongly correlated regardless of home-school distance (r ≥ 0.93).

Stratification by children living closer vs. farther from school

To determine if distance between home and school influenced differences between home and school exposures we stratified analyses by the upper tertile of distance between home-school for all of the cohorts combined (for greater generalizability) and by city-specific upper tertile distance thresholds. When considering all 4 cohorts together, the median Euclidean distance between home and school was 2.1 km (IQR: 3.6) (Supplementary Table S1) with 3.4 km representing the upper tertile. Participants from the URECA cohort, that included children from 4 different cities, had the longest distances between home and school (median: 3.4 km), largely due to participants in St. Louis (median: 4.8 km). In comparison, those in CCCEH (NYC) and Viva (Boston) lived closer to school on average (median distance <2 km).

Using the cohort-combined distance threshold, children attending school farther from home (≥3.4 km) had a greater difference between home-school socio-demographic exposures (median [IQR]: 16.8% [125] difference in no high school diploma, 43.1% [145] difference in families receiving public assistance, −7.2% [67.9] difference in median household income), with less favorable conditions near home vs school (Table 3). Conversely, children who attended school closer to home had no significant differences in socio-demographic characteristics (i.e. median [IQR]: 0% [31.1] difference in high school diploma, 0% [35.7] difference in public assistance, and 0% [20.9] difference in median household income between home and school, Table 3).

Table 3.

Percent differences in social/environmental exposures between home and school (median [IQR] presented), stratified by distance (near vs. far).

Exposure category	Exposure variable, % Diffa	Near (N = 615)	Fartherb (N = 308)	p-valuec
Neighborhoods socio-demographicsd	Total population, %	0.0 [26.0]	1.7 [75.1]	0.56
	Average household size, %	0.0 [8.1]	1.4 [25.5]	0.16
	Median income, %	0.0 [20.9]	−7.2 [67.9]	0.003
	Poverty, %	0.0 [46.3]	22.7 [163.6]	0.03
	Less than high school education attainment, %	0.0 [31.1]	16.8 [125.0]	<0.001
	Receiving public assistance, %	0.0 [35.7]	43.1 [145.0]	<0.001
	Unemployed, %	0.0 [17.4]	4.4 [54.1]	0.003
	Black, non-Hispanic, %	0.0 [36.4]	11.6 [170.8]	0.01
	Hispanic or Latino, %	0.0 [36.7]	0.0 [204.6]	0.73
	White, non-Hispanic, %	0.0 [8.8]	−2.3 [64.6]	<0.001
Air pollution	PM2.5, %	0.1 [4.7]	0.4 [7.9]	0.48
Traffic exposuree	Distance to nearest primary road, %	−2.7 [60.5]	−11.2 [150.5]	0.32
	Distance to nearest secondary road, %	1.6 [106.2]	8.0 [207.8]	0.48
	Average daily traffic density, %	0.0 [139.3]	0.0 [236.5]	0.07
	Average daily truck density, %	0.0 [85.5]	0.0 [189.8]	0.10
Green spacef	EVI within 500 m	2.5 [22.3]	1.6 [37.8]	0.72
	EVI within 1500 m	0.3 [12.7]	0.8 [28.1]	0.94
	EVI within 2500 m	0.4 [8.6]	1.2 [23.6]	0.30

IQR Interquartile range, PM_2.5 fine particulate matter, EVI Enhanced Vegetation Index.

Bold values represent statistically significant differences after Bonferroni correction.

^aPercent difference between home and school exposures.

^bChildren were classified as attending school “far” from home if the distance between home and school exceeded 3.4 km.

^cMann-Whiteney test performed.

^dNeighborhood demographics assessed at the census tract level between 2010 and 2020.

^eTraffic exposure measured within a 400-meter buffer around address.

^fGreen space calculated within a 500-meter, 1500-meter and 2500-meter buffer around address.

Similar trends of greater differences between home and school socio-demographic characteristics among children living farther from school were observed when using city-specific distance thresholds (Supplementary Table S3). Children from NYC (CCCEH and URECA) who attended school farther from home (≥2.8 km) experienced a greater unemployment rate (14.7% difference), percent with no high school diploma (35.3% difference), and percent receiving public assistance (56% difference) near home compared to school. Similarly, children in Baltimore (URECA) who attended school farther from home (≥4.4 km) were exposed to more people on public assistance (100.6% difference) near home vs school. Fewer differences between home and school socio-demographic exposures were noted for participants in Cincinnati (CCAAPS), Boston (VIVA and URECA) and St. Louis (URECA) when stratified by city-specific distance thresholds.

Impact of household income on home-school differences

To explore the hypothesis that lower income families may send their children to schools farther from home resulting in greater differences between home and school exposures, we further stratified our analysis by household income. There was a non-significant trend whereby children from lower income families tended to live farther from school than children from higher income families (median [IQR]: 2.3 km [4.8] vs. 2.0 [3.2], p = 0.01). Compared to higher-income families, children from lower-income families were exposed to less optimal socio-demographic conditions at home than at school (Supplementary Table S4. lower-income vs higher-income families median [IQR]: −3.0% [51.1] vs 0.0% [24.8] difference in median income, 9.2 [75.2] vs 0.0% [89.1] difference in no high school diploma). The homes of lower income families were located closer to secondary roads compared to their school (Home-school differences for lower-income vs higher-income families: −6.4% [141] vs 11.9% [128.7] difference; p < 0.002).

In exploratory analysis examining how well individual-level SES reflected neighborhood levels SES, we observed that lower-income families lived in neighborhoods with substantially lower median household income, compared with higher-income families (Supplementary Table S5. Median neighborhood income: $41,036 vs $74,205 for lower-income vs higher-income families, respectively; p < 0.001). Additionally, individual-level income was moderately correlated with neighborhood-level income (Supplementary Table S5. r = 0.5; p < 0.001). City-specific analyses further indicated that most cities showed significant correlations between individual- and neighborhood-level SES with modest variations across cities. Associations were strongest among children from Cincinnati, followed by those from Boston and Baltimore (Supplementary Table S5. r = 0.49, 0.34, and 0.33, respectively; p < 0.05 for all), whereas no association was observed among children from St. Louis (r = −0.01; p = 0.40).

DISCUSSION

In a sample of 923 school-aged children from four urban and suburban cohorts in the US, we observed notable differences in neighborhood level socio-demographic characteristics between home and school locations, but this was not true for outdoor environmental exposures (i.e., PM_2.5, traffic-exposure, and green-space). The differences in home-school socio-demographic characteristics were largely driven by children in Baltimore and St. Louis (URECA) and were most evident among children who lived farther from school (i.e., those attending schools ≥3.4 km from home) and those from lower income households. City-specific distance thresholds highlighted greater socio-demographic differences between home and school in NYC and Baltimore, but only among children with longer distances between home and school. Our findings illustrate the importance of considering both home-school distance and socio-demographic context when assessing environmental disparities among children.

Schools are an important source of environmental exposure and may significantly influence children’s respiratory health [5, 15], and academic performance [34–36]. However, the specific contribution of school-based exposures remains underexplored. In our previous study, we found that TRAP exposure in school environments, independent of home exposures, increased airway inflammation in school-aged children, underscoring the need to include school settings in environmental exposure assessments related to respiratory health [5]. Similarly, an analysis of 3,660 Michigan public schools found that those closer to highways and industrial facilities were associated with greater risks of respiratory and neurological diseases and poor student performance [36]. Several studies have also assessed environmental exposures across both home and school settings. For example, a study of over 150,000 public school children in Florida used cumulative distribution functions to estimate proximity-based exposure potential and found that a greater proportion of children were exposed to air pollution at home than at school [37]. In comparison, McConnell et al. (2010) reported that combined TRAP exposures across home and school environments were more strongly associated with new-onset asthma than exposure at either home or school location alone, highlighting the importance of evaluating both settings in studies of environmental health risks [38]. Personal exposure studies of school-aged children, which capture exposures across multiple microenvironments including home and school, further corroborate the importance of evaluating both environments for overall exposure burden and their impact on asthma-related outcomes [39, 40]. Understanding the variability between home and school environments including social and demographic exposures such as built environmental features and neighborhood socioeconomic conditions, which are linked to respiratory health, could offer important insights into how these settings influence childhood respiratory outcomes.

To our knowledge, this is the first study to assess multi-environmental exposures, including neighborhood socio-demographic characteristics and green-space at both home and school locations. A key strength of this study is the use of individually geocoded home and school addresses, allowing for detailed environmental assessment without sharing participants’ protected health information, using extensive data from children within 5 US cities. Our findings suggest that neighborhood socio-demographic characteristics, but not outdoor environmental exposures, differ between home and school, particularly among children who live farther from school. Across cities, the magnitude of home-school differences in socio-demographic conditions varied. For example, children in Baltimore and St. Louis (URECA) had the largest differences, with children often living in more disadvantaged neighborhoods than those surrounding their schools. In contrast, differences were less pronounced in Boston (Viva) and Cincinnati (CCAAPS), where home and school neighborhood conditions were more similar.

Children who lived farther from school were more likely to live in disadvantaged neighborhoods, characterized by lower educational attainment, and higher percentage of residents on public assistance, compared to the neighborhoods surrounding their schools. When using city-specific thresholds to define farther distances, differences in neighborhood socio-demographic conditions were particularly pronounced in NYC and Baltimore. Also, in NYC, children living farther from school had access to less green-space within a 1500 m buffer around their homes compared to near their schools. Our study also provides evidence that socioeconomic status, reflected by self-reported individual household income, may contribute to differences in environmental exposures between home and school, as suggested by lower neighborhood median income and higher percentage on public assistance near home compared to school among lower income households. One potential explanation for the differences in home-school socio-demographic exposures by income is the availability of school choice programs in many urban areas like NYC and Baltimore. These programs allow families, especially those residing in economically disadvantaged neighborhoods, to enroll their children in schools outside their residential zones in pursuit of better academic quality, greater diversity, and safer environments. Over 82% of children in NYC and Baltimore (URECA) were from low-income families where an extensive school choice system is available. Therefore, these families may have had greater opportunities to attend schools in higher income, and thus better resourced neighborhoods. In contrast, cities like Cincinnati and Boston, represented by higher-income cohorts, maintain strong neighborhood-based assignment systems particularly at the elementary school level. This may limit low-income families’ ability to access schools outside their residential areas. The differences in school assignment policies may help explain the city-specific disparities in neighborhood socio-demographic characteristics, particularly among children attending school further from home. Structural factors such as historical redlining, which shape the spatial distribution of SES conditions, can also contribute to the pronounced city-to-city differences in home and school SES. For example, our previous work showed that historically redlined neighborhoods in NYC typically had a higher prevalence of Black residents, higher Deprivation Index (DI), and higher Social Vulnerability Index (SVI), and lower Child Opportunity Index (COI), compared to non-redlined neighborhoods [14]. These patterns may help explain why children in some cities, particularly those from low-income or racially marginalized neighborhoods, experience greater differences in SES conditions between home and school environments. Taken together, these findings highlight the importance of considering local urban and spatial contexts, as well as the underlying structural factors that shape them, when examining social and environmental inequity.

The alignment between individual and neighborhood-levels SES has important implications for environmental health research, especially when neighborhood-level SES is used as a proxy measure for individual SES. Prior studies have examined the alignment of these measures. For example, Moss et al. (2021) reported that census-track -level SES measures more closely approximated individual-level SES than county-level measures (r = 0.23 vs r = 0.16, respectively), although neighborhood-level SES still underestimated SES-mortality associations when used as a proxy for individual-level SES [41]. Another study examining the role of urbanicity found that individual-level and neighborhood-level SES were poorly correlated in less urban areas (e.g., micropolitan and medium metro), compared with large metro or suburban areas [42]. Similarly, a study of pediatric populations conducted in a mixed urban-rural setting reported poor agreement between individual-level and neighborhood-level SES measures [43]. In our multisite analysis of predominantly urban and suburban settings, we observed a moderate overall correlation between individual-level SES and neighborhood-level SES, with substantial variation across cities. Concordance was strongest in suburban or mixed urban-suburban cities, but weaker in dense urban environments like NYC. The absence of an association in St. Louis likely reflects small sample size for higher-income families at that site. Overall, our results suggest that individual- and neighborhood-level SES capture related but distinct aspects of socioeconomic conditions, underscoring the importance of incorporating both measures when interpreting social and environmental disparities, particularly in diverse and dense urban areas.

While the current study did not directly examine respiratory outcomes associated with home-school environmental exposure differences or the modifying role of SES, related research provides relevant insight. For example, Pollack and colleagues examined whether moving from predominantly low-income neighborhoods to higher-income neighborhoods (through Baltimore Regional Housing Program) reduced asthma morbidity among children with persistent asthma [44]. They found significant reductions in asthma exacerbation rate and maximum symptom days after moving, highlighting how changes in social and physical environments can improve asthma outcomes. The study suggests that neighborhood social factors such as cohesion and safety may explain improvements in asthma outcomes by promoting healthier behaviors (e.g., reduced smoking), reducing time spent indoors (and thus exposure to indoor allergens and pollutants), and alleviating psychosocial stress (e.g., reduced fear and stress). Taken together, our findings raise the possibility that children in low-income families who attend schools further from home, particularly in cities like NYC or Baltimore, where school choice programs are available, may navigate between two distinct social and physical environmental contexts daily, potentially shaping their overall exposure burden and health risk. However, because we do not have city-specific data on typical commuting distances for students of similar ages across these cities, these implications should be viewed cautiously and may not generalize to all urban student populations.

In contrast to socio-demographic characteristics, outdoor environmental exposures such as PM_2.5, traffic-exposures, and green-space were similar between home and school locations. These findings are particularly important because school children spend a significant amount of time per day at school. Thus, a similar burden of exposure near homes and schools suggest that targeted interventions to reduce environmental exposures near school could be an effective and efficient way to reach many students at once. School-based environmental interventions may be particularly effective for children that have high exposure burdens, including those who are also highly exposed at home. A recent systematic review reported modest evidence that school-based interventions (e.g., increasing classroom ventilation to improve air quality, greening school environments, implementing Safe-Routes-style infrastructure to incase active travel time) can improve children’s exposure profiles, physical activity and related health outcomes [45]. These findings support the potential of targeted school-based interventions to mitigate environmental risks, especially in settings where opportunities for individualized residential home interventions are limited.

We acknowledge several limitations to our study. First, we chose an arbitrary buffer distance of 400 m to assess traffic exposure that was based on prior research on traffic-related air pollutants, which show significantly elevated TRAP levels within 200–400 m of major roadway [32]. Using multiple buffer distances, particularly smaller buffers, could provide a more nuanced understanding of exposure variability between home and school. Second, we focused solely on PM_2.5 as the air pollution metric in this study and were unable to capture variability in the intensity and duration of exposures at home vs. school, which could play a role in differences in home and school exposure. PM_2.5 is largely a regional pollutant, with relatively uniform background levels across urban areas, that likely explains the spatial consistency in average concentrations. Yet, our previous study demonstrated that children experienced a greater number of peaks per hour (i.e. peak variability) of BC (a component of PM_2.5) in school, compared to home environments, despite similar average BC levels [5, 46]. Moreover, schools located near high truck density and close proximity to highways exhibited greater peak variability, than those located farther from traffic sources [46]. Urban schools in areas like NYC are frequently located near major roadways and high-traffic truck routes [3], making them potential hotspots for episodic exposures. Therefore, relying solely on average exposure metrics may obscure important short-term fluctuations that could be more relevant to health outcomes, especially in dynamic settings like schools. Expanding beyond PM_2.5 to include other traffic-related pollutants such as BC or NO₂, which have higher spatial variability [47] and incorporating short-term exposure metrics like peak exposure could better capture meaningful differences in outdoor environmental exposures even among children living near their schools, and more accurately reflect environmental risks to children’s health. However, the use of outdoor regional pollution data, as opposed to personal monitoring data, limited our ability to examine short term exposure peaks. In addition, future studies incorporating ozone, another pollutant strongly linked to respiratory outcomes [48] and highly reactive with other atmospheric pollutants [49], alongside traffic-related pollutants, could provide a more comprehensive assessment of children’s warm-season exposure risk, although ozone may be less informative for distinguishing home vs school exposures given its predominance during summer months and school breaks. Third, our study did not account for indoor environmental factors. Important exposure to pollutants such as PM_2.5 and NO₂ can occur indoors (e.g., gas stove use, cooking activity, and ETS exposure) and may differ substantially from outdoor exposure. Fourth, our study may be subject to potential exposure misclassification. We used census tract-level SES data as a proxy for neighborhood context. However, the use of ACS data without accounting for uncertainties may lead to misclassification of neighborhood level SES [50]. In addition, Euclidean distance often underestimates true travel distance because it ignores road networks, and physical barriers, and population distribution within geographic units [50]. Fifth, our greenspace measure relied on EVI, which is a widely used and validated measure of vegetation. However, this measure may oversimplify the construct of green-space, as it does not capture important dimensions of greenspace such as temporal variability across years or season, vegetation quality, type (e.g., trees vs grass) or accessibility, all of which may have health relevance. Sixth, we did not account for time-weighted exposures (e.g., 75% of time spent at home vs 25% at school). A time-weighed exposure model might yield different insights into the environmental risks children face across these settings. Also, while the data were derived from longitudinal cohort studies, the current analyses are cross-sectional. Finally, it is possible that enrollment and self-selection bias influenced our findings. Families who voluntarily chose to participate in the ECHO cohorts, compared to non-participants, may have greater health awareness or motivation to seek educational opportunities perceived as beneficial for children, which may limit the generalizability of our findings to all urban/suburban populations.

CONCLUSIONS

Our analyses of multiple social and environmental exposure differences between home and school highlight important implications for health-effect studies and underscore the need to differentiate strategies addressing social and outdoor environmental exposures for interventions targeting health equity. Less favorable home-SES compared to school-SES highlights that interventions may need to address social and structural disparities at home, such as housing quality and neighborhood SES, to reduce environmental health inequities. At the same time, school may offer an opportunity to provide beneficial exposures, potentially buffering some of the adverse SES-related exposures experienced at home. The similarities observed in PM_2.5, traffic-related exposures, and green-space between home and school, offer rationale for support of school-based interventions that can target large numbers of children at once (e.g. installing air filtration systems in classrooms and limiting outdoor activities during times of high pollution) when opportunities for individualized residential interventions may be limited. The substantial differences in neighborhood socio-demographic characteristics between home and school, particularly for children traveling further distances, underscore the need for future studies to evaluate how these environmental differences influence pediatric respiratory outcomes especially among socioeconomically disadvantaged populations. Overall, our findings provide rationale for a holistic assessment of the social and environmental exposure profile in future environmental health.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41370-026-00875-6.

DATA AVAILABILITY

Select de-identified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). Information on study data not available on DASH, such as some Indigenous datasets, can be found on the ECHO study DASH webpage.