Child Opportunity Index at Birth and Asthma with Recurrent Exacerbations in the U.S. ECHO Program

Abstract

Background:

Environmental exposures and social determinants likely influence specific childhood asthma phenotypes.

Objective:

We hypothesized that the Child Opportunity Index (COI) at birth, measuring multiple neighborhood opportunities, influences incidence rates (IRs) for asthma with recurrent exacerbations (ARE).

Methods:

We tested for COI associations with ARE incidence rates in 15,877 children born between 1990–2018 in the Environmental Influences on Child Health Outcomes (ECHO) program. Parent-reported race and ethnicity and other demographics were assessed as effect modifiers.

Results:

The IR of ARE for children born in very low COI neighborhoods was higher (IR=10.98; 95% Confidence Intervals (CI) 9.71, 12.25) than for other COI categories. Rates for Non-Hispanic Black children (NHB) were significantly higher than Non-Hispanic White (NHW) children in every COI category. The ARE IRs for children born in very low COI neighborhoods were several-fold higher for NHB and Hispanic Black (HB) children (IR=15.30; 95% CI 13.10, 17.49; IR=18.48; 95% CI 8.80, 28.15 respectively) when compared to White children. Adjusting for individual-level characteristics, children born in very low COI neighborhoods demonstrated an ARE incidence rate ratio (IRR) of 1.26 (95% CI 0.99,1.59) with a higher incidence of cases among children ages 2–4 years and with a parental history of asthma.

Conclusions:

Rates of ARE were higher among children born in under-resourced communities and this relationship is strongest for young minoritized children with a parental history of asthma. Higher rates for NHB even in the highest COI categories suggest that risk associated with race persists regardless of social disadvantage.

Capsule Summary:

Rates of childhood asthma with recurrent exacerbations were higher among children born in under-resourced communities. This relationship is strongest for young minoritized children. Higher rates for NHB children were observed across all categories of childhood opportunity.

Graphical Abstract

Introduction:

Childhood asthma with recurrent exacerbations (ARE) is a distinct asthma phenotype that is associated with substantial child morbidity, health care costs, and burdens for the patient and family, including in comparison to more controlled asthma.^1–7 The U.S. Environmental influences on Child Health Outcomes (ECHO) program, a nationwide consortium of pediatric cohorts whose participants are geographically and demographically diverse, was designed to help address causes of childhood asthma and other important health outcomes. In a recent analysis of ECHO data from over three decades of surveillance we reported that the incidence of childhood ARE varied over time, and was higher in one decade (2000–2009) compared to other decades (1990–1999, 2010–2017).⁸ These findings suggest that dynamic environmental exposures, social determinants, or both may contribute to the etiology of ARE.

Research from ECHO and the Children’s Respiratory and Environmental Workgroup (CREW), a group of asthma-focused cohorts funded under ECHO, have identified environmental exposures and neighborhood-level social determinants of health that were associated with the development of childhood wheeze and asthma.⁹ These associations were observed for exposures and social determinants linking census data to birth address using the census year nearest to the birth year, suggesting that prenatal exposures may influence subsequent asthma-related outcomes.⁹ One neighborhood measure, the Child Opportunity Index (COI), combines weighted information from 29 indicators of neighborhood conditions that correlate with greater upward economic mobility and better health and development of children in long-term studies.^10–12 Sub-scores in three domains (education, health and environment, social and economic) are grouped into a final overall opportunity score at the census tract level.^11,12 The higher the COI score, the greater the number of positive features in a neighborhood that are believed to contribute to better economic mobility and healthier child development, including access to quality educational resources, healthy food and green space, lower air pollution, and better employment opportunities. A lower overall COI score has been associated cross-sectionally with higher rates of asthma-related emergency department (ED) visits and hospitalizations for young children.¹³ Further, pooled data from 46 ECHO cohorts around the U.S. has demonstrated that high or very high compared to low COI levels measured at birth were associated with reduced incidence of childhood asthma even after adjusting for individual sociodemographic characteristics, parental history of asthma and other covariates.¹¹ These findings suggest that prenatal and childhood measures of neighborhood characteristics contribute to childhood asthma risk. However, the impact of prenatal neighborhood conditions as measured by the COI on the development of the more burdensome childhood ARE phenotype remains an important public health concern and a substantial research gap.

Based on these previous findings, we hypothesized that lower COI assessed at the birth census tract influences the incidence rates for childhood ARE. We additionally hypothesized that individually reported race and ethnicity would modify this association between COI and ARE. Further understanding of the prenatal and early childhood determinants of this complex childhood disorder at both the individual and neighborhood level could help refine strategies to prevent childhood ARE.

Methods:

1. Study population and ARE outcome:

Children were enrolled from existing cohorts participating in the U.S. ECHO consortium and meeting the following a priori defined additional eligibility criteria, as described previously^8,14: a) born 1990–2018, b) were followed up from 2 years to at least 5 years and up to 19 years of age, and c) had geocoded residential addresses at birth linked to the COI (described below). Procedures for consent, enrollment, and data harmonization have been described.¹⁵ Data were gathered from children starting at 2 years of age (i.e. time zero) who were followed until at least age 5 years and up to age 20 years; through the date of last visit, date of loss to follow-up, or the end of the study follow up period, whenever was earliest. The analysis data set included data collected through March 1^st, 2023. The institutional review boards for each cohort site and the ECHO Data Analysis Center (Johns Hopkins Bloomberg School of Public Health) approved all activities associated with this study and informed consent and assent was obtained.

ARE was determined based on at least two reports of systemic corticosteroid use at any time during the entire follow-up period with each considered an exacerbation event if separated by a minimum of 30 days. ARE was defined at the time of the second exacerbation event.¹⁶ Diagnosis of asthma for this study required that a child diagnosed before 5 years of age also have a reported diagnosis of asthma during a follow-up visit at age 5 years or older to avoid disease misclassification.¹⁷ The date of asthma diagnosis could be at any time, irrespective of the ascertained date of ARE. Asthma diagnoses between ages of 2 and 5 years were classified at the reported age but only if they were later confirmed at or beyond age 5 years (i.e. repeated diagnosis, or evidence of asthma-related provider visit, medication visit, or hospitalization). All reports of systemic corticosteroid use starting at age 2 years were considered as asthma exacerbations ascertained as described previously.⁸ In 49 cases (2.6%) a reported use of corticosteroid was therefore combined with the previous occurrence. There were 16 children treated with corticosteroids at least twice before age 2 years who did not have evidence of an asthma diagnosis during follow-up and were not considered to be cases.

2. Neighborhood socioeconomic and opportunity measures at birth:

The ECHO Data Analysis Center linked COI to participant birth addresses using ArcGIS geospatial software (Esri, Redlands, CA). The COI was built in a supervised fashion with weights based and derived from long-term studies of children, including those that move over time, and established factors of early life that account for healthier outcomes. The COI includes metrics such as poverty and unemployment rates, access to green space, and number of high-quality early childhood education centers from all public sources such as the Census Bureau, National Center for Education Statistics, U.S. Department of Agriculture, and the Environmental Protection Agency, combined into one overall and three domain sub-scores (education, health and environment, social and economic). For each census tract, a nationally standardized z-score for each neighborhood indicator was calculated and subsequently domain-specific and overall COI scores for a total of 72,195 census tracts across all U.S. states and the District of Columbia were derived. The scores were ranked and normed nationally such that higher scores reflected more favorable neighborhood environments relative to other neighborhoods across the US. Each participant’s primary residential address at birth (1994, earliest birth year recorded–2018) was geocoded as determined from monthly residential timelines constructed based on all available addresses and residential history information in ECHO. Census tract at birth was then linked spatially to the geocoded address using the nearest in time decennial US census tract boundaries. Participants only with high quality geocodes (i.e., either a point or specific street address) were included. The version of COI (2010, 2015) closest to the birth year was then linked to the residential census tract. For ECHO participants who could not share address information, this workflow was conducted in an identical but decentralized manner by cohort investigators, and COI estimates at birth stripped of identifiable information were then integrated into the ECHO Data Analysis Center dataset. Consistent with prior ECHO studies,^10,11,18 categories of COI used in these analyses were grouped census tracts based on nationally normed quintiles of the COI score as follows: very low (≤20^th percentile), low (20^th–<40^th percentile), moderate (40^th–<60^th percentile), high (60^th–<80^th percentile) and very high (≥80^th percentile). For the current analyses the high and very high categories were combined.

In addition, several sociodemographic and economic indicators were linked with the birth address census tract, following a similar approach to the COI (that involves combining centralized and decentralized workflows). This was achieved by utilizing data that were temporally adjusted to account for census boundary changes over time, adjusted for inflation, and standardized nationally using z-scores. These data were sourced from the 1990, 2000, and 2010 census as well as the 2019 American Community Survey (ACS) data and were compiled by ECHO investigators. These variables include the percentages of individuals living below the federal poverty line, unemployed, female-headed households, and the representation of specific racial or ethnic backgrounds (also in percent of total population).^9,19

3. Covariates and statistical analyses:

Individual-level reported (by parent) race and ethnicity were categorized as Non-Hispanic White (NHW), Hispanic White (HW), Non-Hispanic Black (NHB), Hispanic Black (HB) and Other race and ethnicity (including as Non-Hispanic Other; Hispanic Other; and Other or Unknown). Self-reported race and ethnicity were viewed as societal constructs, rather than deterministic biological causes of disease risk.²⁰ Individual-level measures of maternal education (highest attained from categories of bachelor’s degree, some college, high school degree or less than high school degree), decade of birth (1990–1999, 2000–2009, 2010–2018), child age and sex (assigned at birth), parent-reported child race and ethnicity, census region of recruitment site, and parental history of asthma (from either or both biological parents) were included. Geographic region was categorized into four regions (West, Northeast, Midwest, South) as defined by the U.S. Census.²¹

Prior to analysis, data from all participants were pooled and inspected for distributions, outliers and missingness. To address missingness (up to 30%) of individual-level data, we implemented RELATE R package to find clusters of cohorts that were common based on variables included in the analysis.²² Multiple imputation with chained equations (MICE) within the clusters of cohorts was used to create 30 datasets after which Rubin’s combination rules were applied to obtain estimates from the multiply imputed data. Distributions of neighborhood-level variables by ARE status obtained using two-sample t-tests and Wilcoxon rank-sum for normally and non-normally distributed data, respectively, were compared. Poisson regression models with a fixed effect were used to adjust for cohort groups (i.e. study design and implementation characteristics) and a cluster term was used for county (i.e. geography/region-based characteristics). Unadjusted incidence rate ratios (IRRs) by individual-level characteristics, followed by adjusted IRRs where we extended models to include the above-mentioned individual-level covariates of interest, were estimated. Where model convergence was not achieved (small cells), clustered sandwich estimators rather than use of a fixed effect to adjust for clustering on groups of cohorts with common characteristics were employed, and the number of imputed datasets to compute estimates was reduced to the maximum possible for achieving standard error calculation. Differences in the association between COI and ARE incidence by race and ethnicity were explored by conducting stratified analyses. The significance of interaction terms to explore potential effect modification by birth decade, age at follow-up, parental history of asthma, sex, census region at birth, and maternal education at childbirth was assessed. Model goodness-of-fit was evaluated by likelihood ratio tests comparing nested (outcome model with all covariates) and extended models (outcome model with all covariates and interaction term). The fitted model estimated (i.e. predicted) number of ARE cases per 1000 child-years by each potential modifier was calculated. Finally, as a sensitivity analysis to further assess cohort heterogeneity, we conducted a meta-analysis estimating IRRs comparing incidence of ARE among moderate, low, and very low COI versus high/very high (reference) groups for each RELATE subgroup, as described.^23,24

All models were performed using Stata version 17 (StataCorp, College Station, TX).

Results:

Of the 63,799 children in ECHO, 25,699 met study eligibility criteria. Of this number, 15,877 were involved in cohorts that collected data on asthma diagnoses and corticosteroid use and composed the analytic cohort. These children were from 60 geographically diverse ECHO cohorts within the U.S. including Puerto Rico and provided 119,068 person-years of follow-up for these analyses (Figure 1). Supplemental Table 1 compares the study-eligible ECHO children with those included in the analytic cohort and suggests that the latter differs by being recruited during earlier decades and included more Non-Hispanic White children. The maternal educational level remained similar while the analytical cohort had more participants with a parental history of asthma, probably because cohorts that collected asthma diagnosis data were more likely to collect such information. Measures of heterogeneity (I²: 0%, Cochran’s Q: p=0.92) indicated an insignificant amount of heterogeneity across RELATE subgroups of cohorts sharing common characteristics. Of the study population of 15,877, 692 children (4.4%) had ≥2 exacerbations and met the definition of ARE. Overall, the incidence rate of ARE was 6.46 per 1,000 person years (95% CI: 5.98, 6.94; n=15,877; 119,068 person-years of follow-up, 692 cases; Supplemental Table 2). Recent decades of birth, younger child age, NHB race and HB race, Northeast, Midwest and South census region compared to West, male sex and parental history of asthma was associated with a higher IR of ARE. A lower level of maternal education attained also was associated with a higher IR of ARE.

Figure 1:

Distribution of the study population. All study-eligible ECHO children (n=25,699) were defined as baseline eligible if born after 1990, contacted between age 2–19 years of age and had a birth census tract coded.

Table 1 shows significant differences in z-scores measured from census tract-derived neighborhood-level sociodemographic and economic characteristics between cases of ARE and those that did not develop ARE (the comparison group). At the neighborhood level, ARE cases were associated with residence in areas with higher percentages of the population below the poverty level, with a female head of household, with higher unemployment, lacking a high school diploma, and with a higher population density. The overall COI and domain specific scores, median household income, and distributions of White, Asian and Hispanic individuals were lower (all p <0.01), and the distribution of Black individuals was higher (both p<0.01) for ARE cases compared to the comparison group.

Table 1.

Distribution of census tract level neighborhood characteristics by ARE status based on z scores

Census-based neighborhood-level characteristics (z scores)	All children N=15,877	Comparison group N=15,185	ARE case N=692	p-value
% Below poverty level	−0.39 (−0.75, 0.39)	−0.40 (−0.75, 0.36)	0.03 (−0.61, 1.05)	<0.01
Median household income	−0.01 (−0.58, 0.65)	0.01 (−0.56, 0.66)	−0.41 (−0.91, 0.28)	<0.01
% Female head of household	−0.25 (−0.67, 0.48)	−0.26 (−0.68, 0.44)	0.25 (−0.44, 2.18)	<0.01
% Unemployed	−0.33 (−0.72, 0.34)	−0.34 (−0.73, 0.31)	0.04 (−0.61, 1.02)	<0.01
% Without HS diploma	−0.39 (−0.84, 0.41)	−0.40 (−0.84, 0.38)	0.03 (−0.70, 0.81)	<0.01
Population density, people per square km	−0.17 (−0.38, 0.15)	−0.17 (−0.38, 0.14)	−0.12 (−0.33, 0.38)	<0.01
% White participant	0.17 (−0.94, 0.76)	0.18 (−0.88, 0.76)	−0.34 (−2.12, 0.64)	<0.01
% Black participant	−0.44 (−0.56, 0.16)	−0.45 (−0.57, 0.11)	0.01 (−0.51, 2.17)	<0.01
% Asian participant	−0.26 (−0.42, 0.19)	−0.26 (−0.42, 0.21)	−0.34 (−0.45, −0.11)	<0.01
% Other participant	−0.30 (−0.55, 0.33)	−0.30 (−0.55, 0.33)	−0.38 (−0.56, 0.18)	<0.01
% Hispanic participant	−0.43 (−0.59, 0.11)	−0.43 (−0.59, 0.12)	−0.48 (−0.59, −0.09)	<0.01
% White, non-Hispanic participant	−0.40 (−0.51, 0.06)	−0.39 (−0.50, 0.05)	−0.44 (−0.52, −0.22)	<0.01
% Black race, non-Hispanic participant	−0.22 (−0.25, 0.08)	−0.23 (−0.25, 0.08)	−0.18 (−0.24, 0.04)	<0.01
Overall COI z score	0.01 (−0.02, 0.03)	0.01 (−0.02, 0.03)	−0.01 (−0.05, 0.02)	<0.01
Education COI domain, mean (SD)	−0.00 (0.06)	0.00 (0.06)	−0.02 (0.06)	<0.01
Health and environment COI domain, mean (SD)	0.01 (0.05)	0.01 (0.05)	−0.01 (0.06)	<0.01
Social and economic COI domain, mean (SD)	0.05 (−0.12, 0.17)	0.05 (−0.12, 0.17)	−0.06 (−0.33, 0.12)	<0.01

Point estimates represent median Z score (IQR) unless specified otherwise. Negative sign indicates median is below the normalized mean.

Other race and ethnicity was defined as Non-Hispanic Other; Hispanic Other; Other or unknown race or ethnicity.

All census-based neighborhood-level characteristics are based on n=15,427 children (n=450; 2.8%) missing data) except for all COI z-scores (n=0 missing data).

Abbreviations: ARE: Asthma with recurrent exacerbations; COI: Child opportunity index; HS: High school; SD: Standard deviation.

Figure 2 and Supplemental Table 3 display IRs for the COI categories stratified by individually-reported child race and ethnicity. The IR of ARE for children residing at birth in a very low overall COI neighborhood was significantly higher than for children born in neighborhoods from all the other categories at 10.98 per 1000 person years (95% CI: 9.71, 12.25). The IRs of ARE for NHB children were 1.76 (moderate COI) to almost 4 (very low COI score category) times higher when compared to NHW children. Within the very low COI category, the IR of ARE was higher among HB children compared to HW children and among NHB compared to NHW. Otherwise, the IRs were not statistically different between the HB and NHB children or between HW children and NHW White children, although the numbers of HB and HW children with ARE were small (<15 in each COI category).

Figure 2:

Overall COI category-specific measures of incidence rates (IRs) of ARE by individual-level race and ethnicity. 95% CI are shown.

We next examined COI associations with ARE for each of the 3 COI domains and found greatest differences across the four categories for the education domain. The IR of ARE for children in the very low category (10.41 per 1000 person years; 95% CI: 8.97, 11.84) was nominally higher than the other groups (Supplemental Table 4). Otherwise, similar findings were observed for the social and economic domain, and the health and environment domain-specific COI categories (Supplemental Tables 5, 6).

Table 2 presents COI category-specific and individual-level characteristic-specific incidence unadjusted and adjusted rate ratios (IRRs) of ARE. Unadjusted analyses showed higher IRRs of ARE among children from very low and low COI neighborhoods, as well as an effect of several individual level characteristics. However, adjustment for individual characteristics diminished the association of ARE with a very low and low overall COI (Table 2). Even when removing the adjustment for individual level maternal education, the IRR was attenuated but remained significant (1.32 (95% CI: 1.05,1.65, p=0.02)). Birth during more recent decades (2000–2009, 2010–2018) remained significantly associated with a higher IRR (adjusted IRR=2.45; 95% CI:1.61, 3.73; adjusted IRR=2.39; 95% CI: 1.47, 3.86, respectively). Adjusted analyses for both the 5–9 and 2–4 year old age groups increased the IRRs to IRR= 25.08; 95% CI: 19.58, 32.12 in the latter group. Individually reported HB and NHB race and ethnicity also retained their associations with a higher IRR for ARE, although adjustment diminished the effect (adjusted IRR=1.93; 95% CI: 1.13, 3.27 and IR=1.97; 95% CI: 1.58, 2.46) respectively. Northeast, South and Midwest census region at birth retained significance when compared to the West region, as did child sex and parental history of asthma, with minimal impact on the IRR upon adjustment for the other characteristics. When examining the adjusted IRRs of ARE for the education domain of COI and individual level demographic characteristics, the 2000–2009 and 2010–2018 decade of birth, NHB, HB, Other race and ethnicity, Northeast, Midwest, South census region, male sex and parental history of asthma similarly retained statistical significance (Table 3). The very low category of COI was associated with a diminished adjusted IRR for ARE that retained statistical significance (IRR=1.29; 95% CI: 1.02, 1.63). Similar trends for the social and economic domain and the health and environment domain of the COI are shown in Supplemental Tables 7, 8.

Table 2.

Overall COI category-specific and individual-level characteristic-specific incidence unadjusted and adjusted^* rate ratios (IRRs) of ARE by multivariate regression model

	Unadjusted IRR (95% CI)	P value	Adjusted IRR (95% CI)	P value
Overall COI categories
High/very high	Reference		Reference
Moderate	1.13 (0.89,1.45)	0.32	1.05 (0.81,1.35)	0.72
Low	1.35 (1.06, 1.72)	0.01	1.04 (0.81,1.34)	0.76
Very low	2.14 (1.78, 2.58)	<0.01	1.26 (0.99,1.59)	0.06
Individual level characteristics
Maternal education
≥Bachelor’s**	Reference		Reference
Some college	1.76 (1.42, 2.19)	<0.01	1.26 (1.00,1.57)	0.04
HS degree	1.71 (1.25, 2.32)	<0.01	1.13 (0.82,1.55)	0.45
<HS	1.64 (1.10, 2.42)	0.01	1.14 (0.76,1.73)	0.52
Decade born
1990–1999	Reference		Reference
2000–2009	2.23 (1.54, 3.22)	<0.01	2.45 (1.61,3.73)	<0.01
2010–2018	4.15 (2.64, 6.54)	<0.01	2.39 (1.47, 3.86)	<0.01
Child age (years)
10–18	Reference		Reference
5–9	2.85 (2.23, 3.66)	<0.01	5.19 (4.00, 6.74)	<0.01
2–4	12.51 (9.88,15.8)	<0.01	25.08 (19.58, 32.12)	<0.01
Child race, ethnicity
Non-Hispanic White	Reference		Reference
Non-Hispanic Black	2.67 (2.23, 3.21)	<0.01	1.97 (1.58, 2.46)	<0.01
Hispanic White	1.01 (0.72,1.43)	0.94	1.15 (0.78, 1.61)	0.45
Hispanic Black	3.09 (1.86, 5.12)	<0.01	1.93 (1.13, 3.27)	0.02
Other	1.47 (1.19, 1.83)	<0.01	1.37 (1.08,1.73)	<0.01
Census region
West	Reference		Reference
Northeast	4.68 (3.36, 6.52)	<0.01	4.32 (3.07, 6.08)	<0.01
Midwest	3.66 (2.53, 5.29)	<0.01	3.86 (2.64, 5.64)	<0.01
South	5.18 (3.39, 7.91)	<0.01	4.05 (2.59, 6.34)	<0.01
Child sex
Female	Reference		Reference
Male	1.41 (1.21,1.64)	<0.01	1.36 (1.17, 1.59)	<0.01
Parental history of asthma
None	Reference		Reference
Either or both	2.80 (2.37, 3.32)	<0.01	2.52 (2.13, 2.99)	<0.01

^*Adjusted IRR estimates were derived from models adjusting for all variables in the table.

^**≥Bachelor’s refers to bachelors, master’s, some professional, or doctorate degree. Other refers to Non-Hispanic Other; Hispanic Other; Other or unknown.

Abbreviations: ARE: Asthma with recurrent exacerbations; CI: Confidence interval; COI: Child opportunity index; HS: high school; IRRs: incidence rate ratios. Model controls for the clustering effect of cohorts by using a fixed effect representing groups of cohorts found to have common characteristics using the RELATE package in R.

Table 3.

Education COI domain and individual-level characteristic-specific incidence unadjusted and adjusted^* rate ratios (IRRs) of ARE by multivariate regression model

	Unadjusted IRR (95% CI)	P value	Adjusted IRR (95% CI)	P value
Education COI domain categories
High/very high	Reference		Reference
Moderate	1.13 (0.90, 1.42)	0.31	1.05 (0.83, 1.32)	0.72
Low	1.68 (1.36, 2.08)	<0.01	1.25 (1.00, 1.58)	0.05
Very low	2.02 (1.65, 2.47)	<0.01	1.29 (1.02, 1.63)	0.03
Individual level characteristics
Maternal education
≥Bachelor’s**	Reference		Reference
Some college	1.76 (1.42, 2.19)	<0.01	1.25 (1.01, 1.57)	0.05
HS degree	1.71 (1.25, 2.32)	<0.01	1.13 (0.83, 1.55)	0.44
<HS	1.64 (1.10, 2.42)	0.03	1.15 (0.76, 1.72)	0.51
Decade born
1990–1999	Reference		Reference
2000–2009	2.23 (1.54, 3.22)	<0.01	2.45 (1.61, 3.74)	<0.01
2010–2018	4.15 (2.64, 6.54)	<0.01	2.38 (1.47, 3.86)	<0.01
Child age (years)
10–18	Reference		Reference
5–9	2.85 (2.23, 3.66)	<0.01	5.21 (4.01,6.76)	<0.01
2–4	12.51 (9.88, 15.83)	<0.01	25.17 (19.64, 32.25)	<0.01
Child race, ethnicity
Non-Hispanic White	Reference		Reference
Non-Hispanic Black	2.67 (2.22, 3.21)	<0.01	1.99 (1.61,2.46)	<0.01
Hispanic White	1.01 (0.72, 1.43)	0.94	1.15 (0.81, 1.65)	0.43
Hispanic Black	3.09 (1.86, 5.12)	<0.01	1.95 (1.15, 3.30)	0.01
Other	1.47 (1.19, 1.83)	<0.01	1.36 (1.08, 1.72)	<0.01
Census region
West	Reference		Reference
Northeast	4.68 (3.36, 6.52)	<0.01	4.40 (3.13, 6.19)	<0.01
Midwest	3.66 (2.53, 5.29)	<0.01	3.87 (2.65, 5.66)	<0.01
South	5.18 (3.39, 7.91)	<0.01	4.05 (2.59, 6.34)	<0.01
Child sex
Female	Reference		Reference
Male	1.41 (1.21, 1.64)	<0.01	1.36 (1.17, 1.59)	<0.01
Parental history of asthma
None	Reference		Reference
Either or both	2.80 (2.37, 3.32)	<0.01	2.52 (2.12, 2.99)	<0.01

^*Adjusted IRR estimates were derived from models adjusting for all variables in the table.

^**≥Bachelor’s refers to bachelors, master’s, some professional, or doctorate degree. Other refers to Non-Hispanic Other; Hispanic Other; Other or unknown.

Abbreviations: ARE: Asthma with recurrent exacerbations; CI: Confidence interval; COI: Child opportunity index; HS: high school; IRRs: incidence rate ratios. Model controls for the clustering effect of cohorts by using a fixed effect representing groups of cohorts found to have common characteristics using the RELATE package in R.

Figures 3 and 4 and Supplemental Figure 1 display the predicted number (i.e. fitted model estimated) of ARE cases per 1000 child-years by COI level, stratified by potential effect modifiers. The number of ARE cases per 1000 person-years was higher among children born during the 2000–2009 with a very low COI (IR_2000–2009= 53.2 per 1000 person-years, (95% CI: 44.27, 70.29) versus IR_1990–1999= 9.65 (95% CI: 4.40, 21.40). The same trend by birth decade was observed to a lesser extent among those with a low COI (Figure 3). The number of ARE cases per 1000 person-years was highest among very young children ages 2–4 years in each COI category compared to their older counterparts of 5–9 years and 10–19 year of age (Figure 4a). The number of ARE cases per 1000 person-years was higher among children with parental history of asthma in compared to children whose parents have no history of asthma, especially among those in the very low COI category (Figure 4b). The differences after adjusting for male sex and maternal education below a bachelor’s degree across the COI categories were relatively smaller (Supplemental Figure 1).

Figure 3:

Predicted number of ARE cases per 1000 child-years by COI category, stratified by birth decade. Fully adjusted data through 2018 are shown. 95% CI are shown.

Figure 4:

Predicted number of ARE cases per 1000 child-years by COI category, stratified by A. Age groups and B. Parental history of asthma. Fully adjusted data are shown. 95% CI are shown.

Discussion:

ARE captures asthma complications that may or may not present to an ED but still cause substantial healthcare burdens for individuals in the U.S. with an overall rate of 6.46 per 100 person years among children aged 5–19 years participating in the nationwide ECHO program. Recent research from the ECHO program showed time-dependent differences in the IR of ARE during childhood that cannot be explained by genetic factors,⁸ and suggest changing environments including physical exposures and social determinants. Utilizing the large sample size of the ECHO program and detailed residential histories, we evaluated whether the COI overall score, believed to capture systemic structural factors and latent causes at the area or neighborhood level that drive child’s opportunity,^11,12 measured near child birth address was associated independently with the IR for the phenotype of ARE during subsequent childhood. We also adjusted for key individual-level characteristics. We further examined effect modification by individually-reported race and ethnicity, young age and other key demographic characteristics. We found that overall, the IRs of ARE for children born in a neighborhood with a very low COI category was significantly higher than the IRs of ARE for the other (high/very high, moderate, low) COI categories. In addition, the IRs of ARE for the overall COI as well as for the three domains (social and economic, education, health and environment domains) were several-fold higher among children who were parent-reported Black children when compared to White children. Adjustment for individual-level characteristics attenuated the significant association of birth in a very low COI neighborhood and higher incidence rate ratio of AR, more clearly evident when examining the overall and educational COI.

The association of neighborhood-level opportunity based on the overall COI score and each of its three domains with ARE is consistent with previous literature suggesting that neighborhood disadvantage in the prenatal period remains an important determinant of asthma and ED or hospital visits for asthma.^9,11,25,26 Cross-sectional studies have reported associations between lower COI and more frequent pediatric ED and hospital visits for asthma.^13,27 Our data from a diverse U.S. population showed that children born in a neighborhood with a very low COI category, in either the total scores or individuals domains, had the highest incidence of ARE, even after adjustment for individual characteristics.

The stronger associations of individually-reported Black race with ARE across all categories of COI may relate to neighborhood stressors or lack of access to resources that are greater for Black children, even after considering neighborhood factors.^28,29 Independent of the affluence level of the neighborhood, ARE may be more frequent for more marginalized families who have relatively reduced access to optimal health care in ways that are not measured with our research surveys, including those related to social stressors, discrimination or structural racism.^6,30 Previous studies have found that communities of color are more likely to live in areas of historical redlining and higher levels of air pollution.^31–33 Disparities within neighborhoods in ARE for Black children compared to non-Hispanic White children are not likely to be a function of biologic differences in heritable vulnerability.³⁴

ARE cases were more common among younger children, children with a parental history of asthma, children born 2000–2009, boys, and among children of mothers with a lower education level. These factors are similar to those associated with childhood asthma, irrespective of exacerbation history.^17,35 Indeed, previous findings on incidence rates for ARE in the ECHO population also suggested the rates were highest in the youngest children aged 2–4 years, among Black children, and among those with a parental history of asthma.⁸ Poor airway function in infancy has been shown to be associated with lower lung function and exacerbations in adolescence and adulthood,^36,37 further supporting the importance of physiological characteristics detected at young ages to longer term airway health. In the present analyses, the increase in ARE cases for young age was particularly notable in the very low COI category, as was an increase associated with parental asthma history. Higher risk associated with parental history of disease and younger age at onset likely suggest an interaction of early environmental exposures with genetic susceptibility.³⁸

Similar to our previous report,⁸ we observed a higher incidence of ARE among children born 2000–2009 compared to the previous decade. We also observed a trend for higher rates for the 2010–2018 birth decade compared to 1990–1999 with the adjustments made here. Notably we report the novel result that among children with a very low COI, those born during the 2000–2009 birth decade experienced the highest rates of ARE (Figure 3). This new finding varies from our previous observations on asthma incidence that was more notable beginning 1990–1999, especially among African American/Caribbean children with a parental history of asthma.³⁵ Altogether these new results suggest that rates of ARE, unlike the rates of child asthma that may be declining,³⁹ are persisting and related to low levels of neighborhood opportunity (lower opportunity for economic mobility and employment, reduced access to quality educational resources, healthy food and green space, and air pollution or other elements of structural racism). The small influence of individually-reported maternal education (IR=1.32) suggests that this particular indicator of an important family resource explains some but not all of the child’s access and opportunity at the neighborhood level on ARE risk. Thus, neighborhood level interventions are still an important target for specific intervention beyond individual level factors.

While the impact of the prenatal compared to childhood time window of exposure was not directly studied here, the findings analyzing the COI measured at birth suggest that pregnancy could be a critical time for intervention. The impact of the physical environmental and social determinants may begin prior to birth but persist during early childhood and over a more prolonged course to induce oxidative or other toxic injury to the airway or allergic sensitization.⁴⁰ Notably, early childhood adversity including socioeconomic disadvantage has been associated with higher levels of proinflammatory biomarkers in children.⁴¹

We acknowledge several study limitations. First, individual cohorts were harmonized to comprise the overall ECHO population and were not randomly selected from defined populations. Therefore, ECHO is not representative of the general national population or any other population.⁴² Participants included in this analysis also differ slightly compared to the study eligible ECHO population by time/decade of enrollment and race. These differences in composition including missing data could potentially introduce bias and limit generalizability. However, because differences in maternal education were not apparent and because we carefully considered variables (SES) that are thought to have the most impact on research participant invitation and acceptance, we speculate that the impact of differences between our analytic population and the general ECHO and US populations on our study inferences is likely modest. Neighborhood measures of exposure at the census tract level are defined by administrative boundaries and do not necessarily mirror what individuals consider their actual neighborhood in terms of where they spend time and interact with others. As such, they may not fully capture exposures in relevant areas where pregnant women spend most of their time. Second, data used for the neighborhood census-cased social determinants (1990, 2000, 2010) and for the COI (2010, 2015) measures were not exactly the birth year but based on closet year. Given the wide range of variability in birth dates for all the ECHO cohorts integrated into this analysis, and the challenges with harmonizing exposure data across these, some cohorts’ neighborhood measures may be less certain than others (especially in earlier birth years and with the COI). Elements of the actual neighborhood (including access to education, and healthcare and social resources, physical buildings, discrimination, stress, air pollution levels) during pregnancy may have variably changed. Unmeasured confounding from other environmental exposures (e.g. respiratory syncytial virus, and human rhinovirus, allergens, ambient air quality) and social stressors (e.g. discrimination) may not have been captured by these measures and could influence the incidence of ARE. ECHO recruitment sites that differed across decades also could have affected the distribution of urbanicity and other potential unmeasured confounders. However, we examined neighborhood effects while adjusting for individual factors raising our confidence that we distinguished effects at the neighborhood level. We did not adjust for other individual factors beyond parental history of asthma such as gestational age and maternal smoking as our focus was on basic demographic factors. We chose individual indicators that best mirrored the contextual indicators included in the COI. Finally, many of the comparisons between HB and other race and ethnic subgroups were hindered by limited power (i.e. small sample size).

Our results indicate that ARE may be more likely to develop among children living in marginalized under-resourced communities. Potential explanations may include diminished access to health care that can predispose to co-morbidities and inadequate medical prescriptions or counseling. Other environmental instabilities such as substandard housing conditions, greater pollution, higher stress, crowding and less green space within the neighborhood^43,44 can increase the likelihood of asthma exacerbations. This study adds to the mounting evidence that improving disadvantaged neighborhoods has a myriad of benefits for children, providing evidence for policy makers of the potential for improving asthma control and decreasing asthma morbidity. Future studies may assess the utility of incorporating the CO as a clinical prediction tool for children with asthma.

Key Messages:

• The incidence rate of asthma with recurrent exacerbations for children born in very low Child Opportunity Index (COI) neighborhoods was significantly higher than rates for the other COI categories.

• The incidence rate of asthma with recurrent exacerbations for children born in very low COI neighborhoods were several-fold higher for parent-reported Non-Hispanic Black and Hispanic Black children when compared to White children.

• Rates for Non-Hispanic Black children were significantly higher than Non-Hispanic White children in every COI category.

• Following adjustment for individual-level characteristics, children born in a very low COI neighborhood demonstrated an ARE incidence rate ratio of 1.26 (95% CI 0.99,1.59, p= 0.06) with a higher incidence of cases among children ages 2–4 years and with a parental history of asthma.

Abbreviations

ARE

asthma with recurrent exacerbations

CI

confidence interval

COI

child opportunity index

ED

emergency department

ECHO

Environmental influences on Child Health Outcomes

HB

Hispanic Black

HW

Hispanic White

IR

incidence rate

IRR

incidence rate ratio

NHB

Non-Hispanic Black

NHW

Non-Hispanic White

U.S

United States