Research has linked neighborhood opportunity to health outcomes in children and adults; however, few studies have examined neighborhood opportunity and mortality risk among children and their caregivers. The objective of this study was to assess associations of neighborhood opportunity and mortality risk in children and their caregivers over 11 years.
Participants included 1 025 000 children drawn from the Mortality Disparities in American Communities study, a cohort developed by linking the 2008 American Community Survey to the National Death Index and followed for 11 years. Neighborhood opportunity was measured using the Child Opportunity Index, a measure designed to capture compounding inequities in access to opportunities for health.
Using hazard models, we observed inverse associations between Child Opportunity Index quintile and deaths among child and caregivers. Children in very low opportunity neighborhoods at baseline had 1.30 times the risk of dying over follow-up relative to those in very high opportunity neighborhoods (95% confidence interval [CI], 1.15–1.45), and this excess risk attenuated after adjustment for household characteristics (hazard ratio, 1.15; 95% CI, 0.98–1.34). Similarly, children in very low opportunity neighborhoods had 1.57 times the risk of experiencing the death of a caregiver relative to those in very high opportunity neighborhoods (95% CI, 1.50–1.64), which remained after adjustment (hazard ratio, 1.30; 95% CI, 1.23–1.38).
Our analyses advance understanding of the adverse consequences of inequitable neighborhood contexts for child well-being and underscore the potential importance of place-based policies for reducing disparities in child and caregiver mortality.
Research has linked neighborhood opportunity to health outcomes in children and adults; however, few studies have examined neighborhood opportunity and mortality risk among children and their caregivers.
Our analyses advance understanding of the adverse consequences of inequitable neighborhood contexts for child well-being and underscore the potential importance of place-based policies for reducing disparities in child and caregiver mortality.
Neighborhoods influence children’s opportunities for healthy development, as shown in both cross-sectional studies and randomized experiments.1–5 Researchers have documented associations between both advantageous and disadvantageous neighborhood-level characteristics and outcomes including birth weight,6 cognitive development,7 BMI,8 asthma,9 and brain structure,10 while adjusting for individual- and household-level characteristics. Most previous studies have focused on single indicators of neighborhoods, including poverty or violence. However, increasing evidence suggests that consideration of neighborhood opportunity, referring to a global measure of the overall quality of neighborhood conditions and resources,11 offers a valuable approach to understanding the geographic patterning of health outcomes for children and their caregivers.11 The Child Opportunity Index (COI) captures multiple sources of neighborhood disadvantage that are known predictors of healthy child development, but, as a composite index, it also captures the larger adversity that multiple sources of disadvantage create, and reflects how structural racism has generated multidimensional, spatial inequities in access to neighborhood opportunity.12
Ecological studies with aggregated Census tract-level data show that neighborhood opportunity correlates with children’s life expectancy across the 100 largest US metropolitan areas.13 Other studies using multilevel data have demonstrated that neighborhood opportunity is associated with improved birth outcomes,14 children’s cardiometabolic risk,15 and pediatric acute care visits.16 To date, most studies on neighborhood opportunity have had a limited geographic scope, and few, if any, have focused on mortality outcomes. Thus, the field currently lacks a comprehensive understanding of how neighborhood opportunity is related to mortality among children and their caregivers.
Using a large, nationally representative sample, we extend previous work by examining a measure of neighborhood opportunity in relation to (1) child mortality, and (2) death of an adult in the child’s household (hereafter referred to as a “caregiver”), an often-overlooked contributor to a broad range of life-course outcomes and the intergenerational transmission of disparities.17–19 We use the COI, a composite measure designed to capture historically rooted, compounding inequities in access to opportunity that has been used to identify underresourced communities for investments.12,13 We hypothesize that children in lower opportunity contexts will have a higher risk of death and death of a caregiver relative to those in very high opportunity contexts. In quantifying these associations, our analyses advance understanding of:
the geographic patterning of mortality risk in the United States across neighborhoods; and
the potential for place-based and neighborhood mobility policies to reduce disparities in child and caregiver deaths.
Methods
Sample
The Mortality Disparities in American Communities (MDAC) project is a collaboration between the US Census Bureau, Centers for Disease Control and Prevention, and the National Institutes of Health, designed to facilitate research on mortality disparities by social and economic characteristics of individuals and communities. The MDAC was constructed by linking the 2008 American Communities Survey (ACS) to death certificate records between 2008 and 2019 through the National Death Index (NDI). The ACS is an annual complex stratified sample of the US population that collects social and demographic information about adults and children. The ACS sampling frame is obtained from the Census Bureau’s master address file, which is continuously updated to track addresses of known living quarters in the United States. In 2008, the ACS included >4.5 million person-records, with a response rate of 97.8% for housing units and 98.0% for group quarters.20 The ACS questionnaire asks respondents to list persons currently residing in the household. To construct the MDAC, ACS participants were linked to the NDI using a probabilistic mortality matching scheme based on social security numbers (91% of MDAC records) or combined information on first and last name and date of birth.21 The MDAC only retained ACS records that had the necessary information to match to NDI records (percentage of invalid records for match to NDI, 0.7%), and ACS weights were reweighted by age, sex, race, ethnicity, and state to match the US annual population estimates (and applied in all analyses) to adjust for exclusion of ineligible records. Our analytic sample was restricted to children from birth to 17 years in ∼73 000 census tracts across all 50 states and the District of Columbia (n = 1 025 000 children). The Center for Economic Studies approved the analysis, and Census Bureau staff reviewed the output to ensure confidentiality.
Exposure
Every census tract in the United States is assigned a COI score on the basis of 29 indicators of resources and community conditions pertaining to (1) education, (2) health and environment, and (3) social and economic factors (see Table 1 for a description of all component indicators, organized by subscale). We linked each person-record to a COI score on the basis of their census tract for 2010, the closest available year to the MDAC baseline. COI uses 2010 census tract boundaries, whereas the 2008 ACS uses 2000 census tract boundaries. We therefore first recalculated subscale and overall COI z scores for 2000 census tracts using Census Bureau relationship files linking 2000 and 2010 census tracts.22 Following previous research,12 we ranked census tracts according to their COI overall z score, divided them into 5 groups containing 20% of the child population in each, and labeled the resulting categories from very low to very high opportunity. We similarly created quintile variables for COI subscales to use in sensitivity analyses.
Child Opportunity Index Component Indicators and Descriptions, Grouped by Domain
Education | |
ECE | |
ECE centers | Number of ECE centers within a 5-mile radius |
High-quality ECE centers | Number of National Association for the Education of Young Children-accredited centers within a 5-mile radius |
ECE enrollment | Percentage of 3- and 4-y-olds enrolled in nursery school, preschool, or kindergarten |
Elementary education | |
Third-grade reading proficiency | Percentage of third graders scoring proficient on standardized reading tests, converted to NAEP scale score points |
Third-grade math proficiency | Percentage of third graders scoring proficient on standardized math tests, converted to NAEP scale score points |
Secondary and postsecondary education | |
High school graduation rate | Percentage of ninth graders graduating from high school on time |
AP course enrollment | Ratio of students enrolled in at least 1 AP course to the number of 11th and 12th graders |
College enrollment in nearby institutions | Percentage of 18–24-y-olds enrolled in college within 25-mile radius |
Educational and social resources | |
School poverty | Percentage of students in elementary schools eligible for free or reduced-price lunches, reversed |
Teacher experience | Percentage of teachers in their first and second y, reversed |
Adult educational attainment | Percentage of adults aged 25 and older with a college degree or higher |
Health and environment | |
Healthy environments | |
Access to healthy food | Percentage of households without a car located farther than a half-mile from the nearest supermarket, reversed |
Access to green space | Percentage of impenetrable surface areas such as rooftops, roads, or parking lots, reversed |
Walkability | US Environmental Protection Agency Walkability Index |
Housing vacancy rate | Percentage of housing units that are vacant, reversed |
Toxic exposures | |
Hazardous waste dump sites | Proximity to hazardous waste dump sites, reversed |
Industrial pollutants | Concentration of industrial pollutants in air, water, or soil, reversed |
Airborne microparticles | Microparticles (PM 2.5) concentration, reversed |
Ozone concentration | Ozone concentration, reversed |
Extreme heat exposure | Extreme heat exposure, reversed |
Health resources | |
Health insurance coverage | Percentage of individuals aged 0–64 with health insurance coverage |
Social and economic | |
Economic opportunities | |
Employment rate | Percentage of adults aged 25–54 who are employed |
Commute duration | Percentage of workers commuting >1 h 1 way, reversed |
Economic and social resources | |
Poverty rate | Percentage of individuals living in households with incomes <100% of the federal poverty threshold, reversed |
Public assistance rate | Percentage of households receiving cash public assistance or food stamps/Supplemental Nutrition Assistance Program, reversed |
Homeownership rate | Percentage of owner-occupied housing units |
High-skill employment | Percentage of individuals aged 16 and over employed in professional, technical, and managerial occupations |
Median household income | Median income of all households |
Single-headed households | Percentage of family households that are single-parent headed, reversed |
Education | |
ECE | |
ECE centers | Number of ECE centers within a 5-mile radius |
High-quality ECE centers | Number of National Association for the Education of Young Children-accredited centers within a 5-mile radius |
ECE enrollment | Percentage of 3- and 4-y-olds enrolled in nursery school, preschool, or kindergarten |
Elementary education | |
Third-grade reading proficiency | Percentage of third graders scoring proficient on standardized reading tests, converted to NAEP scale score points |
Third-grade math proficiency | Percentage of third graders scoring proficient on standardized math tests, converted to NAEP scale score points |
Secondary and postsecondary education | |
High school graduation rate | Percentage of ninth graders graduating from high school on time |
AP course enrollment | Ratio of students enrolled in at least 1 AP course to the number of 11th and 12th graders |
College enrollment in nearby institutions | Percentage of 18–24-y-olds enrolled in college within 25-mile radius |
Educational and social resources | |
School poverty | Percentage of students in elementary schools eligible for free or reduced-price lunches, reversed |
Teacher experience | Percentage of teachers in their first and second y, reversed |
Adult educational attainment | Percentage of adults aged 25 and older with a college degree or higher |
Health and environment | |
Healthy environments | |
Access to healthy food | Percentage of households without a car located farther than a half-mile from the nearest supermarket, reversed |
Access to green space | Percentage of impenetrable surface areas such as rooftops, roads, or parking lots, reversed |
Walkability | US Environmental Protection Agency Walkability Index |
Housing vacancy rate | Percentage of housing units that are vacant, reversed |
Toxic exposures | |
Hazardous waste dump sites | Proximity to hazardous waste dump sites, reversed |
Industrial pollutants | Concentration of industrial pollutants in air, water, or soil, reversed |
Airborne microparticles | Microparticles (PM 2.5) concentration, reversed |
Ozone concentration | Ozone concentration, reversed |
Extreme heat exposure | Extreme heat exposure, reversed |
Health resources | |
Health insurance coverage | Percentage of individuals aged 0–64 with health insurance coverage |
Social and economic | |
Economic opportunities | |
Employment rate | Percentage of adults aged 25–54 who are employed |
Commute duration | Percentage of workers commuting >1 h 1 way, reversed |
Economic and social resources | |
Poverty rate | Percentage of individuals living in households with incomes <100% of the federal poverty threshold, reversed |
Public assistance rate | Percentage of households receiving cash public assistance or food stamps/Supplemental Nutrition Assistance Program, reversed |
Homeownership rate | Percentage of owner-occupied housing units |
High-skill employment | Percentage of individuals aged 16 and over employed in professional, technical, and managerial occupations |
Median household income | Median income of all households |
Single-headed households | Percentage of family households that are single-parent headed, reversed |
Full details on sources available at https://data.diversitydatakids.org/dataset/coi20-child-opportunity-index-2-0-database. AP, advanced placement; ECE, early childhood education; NEAP, National Assessment of Educational Progress.
Deaths Among Children and Adults in Their Household
Mortality information was obtained from death certificates via the NDI, a centralized database of death record information. Our primary analyses focused on all-cause death for children and any adult in their household during the follow-up period. For children with >1 adult in their household, the first adult death was the outcome event. It was impossible to derive parent–child relationships or caregiving responsibilities for the sample child specifically; thus, any adult in the household was used for this outcome. To gain insight on potential mechanisms, for child deaths only, we considered cause-specific death defined by the International Classification of Diseases, 10th Revision, mortality codes for all external causes (ie, causes of death external to body’s system, such as suicides, accidents, poisoning), and suicide deaths and firearm-related deaths separately (ie, the 2 most common causes within external causes); the sample was not sufficient to study congenital or other causes separately (see Supplemental Table 5 for distributions and International Classification of Diseases codes for cause-specific outcomes).
Covariates
We included a selection of covariates to control for potential confounders of the associations of interest. Model covariates included children’s age, sex, race/ethnicity (ie, non-Hispanic white, non-Hispanic Black, non-Hispanic Asian American and Pacific Islander, non-Hispanic American Indian and Alaska Native, Hispanic, and non-Hispanic other), the highest level of education in the household, household income-to-needs ratio (based on total family income in the past 12 months, family size, and age of household members), family structure, presence of subfamilies in the household (ie, >1 family unit within a single household), and grandparent living with child (see Table 2 for categories).
Demographic Characteristics of the Sample Overall, of Children Who Died, and Children Who Experienced Death of a Household Adult Over the 11-Year of Follow-Up
. | Total Sample . | Child Deaths . | Any Death of Adult in Household . | |||
---|---|---|---|---|---|---|
. | n = 1 025 000 . | n = 4600 . | n = 71 000 . | |||
. | n . | Column % . | n . | Column % . | n . | Column % . |
Child’s age, y | ||||||
0–5 | 318 000 | 31.1 | 550 | 12.3 | 17 000 | 24.1 |
6–11 | 337 000 | 32.9 | 1200 | 25.2 | 22 500 | 31.2 |
12–17 | 369 000 | 36.0 | 2900 | 62.5 | 32 000 | 44.7 |
Child’s sex | ||||||
Female | 525 000 | 51.2 | 1400 | 30.9 | 36 500 | 51.2 |
Male | 500 000 | 48.8 | 3200 | 69.1 | 35 000 | 48.8 |
Child’s race/ethnicity | ||||||
NH white | 649 000 | 63.4 | 2800 | 60 | 41 000 | 57.2 |
NH Black | 110 000 | 10.7 | 750 | 15.8 | 11 000 | 15.1 |
NH API | 42 000 | 4.1 | 200 | 4.7 | 5000 | 7.0 |
NH AI/AN | 10 000 | 1.0 | 100 | 2.3 | 1300 | 1.8 |
Hispanic | 180 000 | 17.6 | 650 | 14.1 | 10 500 | 14.9 |
Other | 33 500 | 3.3 | 150 | 3.1 | 2800 | 3.9 |
Highest educational attainment in HH | ||||||
<HS | 70 000 | 6.8 | 350 | 8.0 | 5400 | 7.5 |
GED/HS | 188 000 | 18.4 | 1100 | 22.6 | 17 500 | 24.8 |
Some college/associate degree | 362 000 | 35.3 | 1800 | 39.6 | 28 000 | 39.4 |
Bachelor’s+ | 400 000 | 39 | 1300 | 27.5 | 20 000 | 28.2 |
Missing | 5300 | 0.5 | 100 | 2.4 | 30 | 0.0 |
Poverty, % | ||||||
<100 | 156 000 | 15.2 | 950 | 20.6 | 13 000 | 18.0 |
100–299 | 391 000 | 38.2 | 1800 | 39.9 | 32 000 | 44.7 |
300–499 | 252 000 | 24.6 | 950 | 21 | 15 500 | 21.6 |
500–999 | 211 000 | 20.6 | 700 | 15.1 | 10 000 | 13.9 |
Other (group quarters, <14 years, foster care) | 15 000 | 1.5 | 150 | 3.5 | 1300 | 1.8 |
Household type | ||||||
Family HH married | 737 000 | 72 | 2700 | 59 | 47 500 | 66.8 |
Family HH not married | 277 000 | 27 | 1800 | 37.8 | 23 000 | 32.4 |
Nonfamily HH | 6300 | 0.6 | 60 | 1.3 | 500 | 0.8 |
Missing | 4100 | 0.4 | 100 | 2.1 | 0 | 0.0 |
Grandparent in HH | ||||||
Yes | 42 500 | 4.2 | 250 | 4.9 | 12 000 | 17.0 |
Subfamily in HH | ||||||
Yes | 74 000 | 7.2 | 350 | 7.8 | 18 500 | 26.0 |
. | Total Sample . | Child Deaths . | Any Death of Adult in Household . | |||
---|---|---|---|---|---|---|
. | n = 1 025 000 . | n = 4600 . | n = 71 000 . | |||
. | n . | Column % . | n . | Column % . | n . | Column % . |
Child’s age, y | ||||||
0–5 | 318 000 | 31.1 | 550 | 12.3 | 17 000 | 24.1 |
6–11 | 337 000 | 32.9 | 1200 | 25.2 | 22 500 | 31.2 |
12–17 | 369 000 | 36.0 | 2900 | 62.5 | 32 000 | 44.7 |
Child’s sex | ||||||
Female | 525 000 | 51.2 | 1400 | 30.9 | 36 500 | 51.2 |
Male | 500 000 | 48.8 | 3200 | 69.1 | 35 000 | 48.8 |
Child’s race/ethnicity | ||||||
NH white | 649 000 | 63.4 | 2800 | 60 | 41 000 | 57.2 |
NH Black | 110 000 | 10.7 | 750 | 15.8 | 11 000 | 15.1 |
NH API | 42 000 | 4.1 | 200 | 4.7 | 5000 | 7.0 |
NH AI/AN | 10 000 | 1.0 | 100 | 2.3 | 1300 | 1.8 |
Hispanic | 180 000 | 17.6 | 650 | 14.1 | 10 500 | 14.9 |
Other | 33 500 | 3.3 | 150 | 3.1 | 2800 | 3.9 |
Highest educational attainment in HH | ||||||
<HS | 70 000 | 6.8 | 350 | 8.0 | 5400 | 7.5 |
GED/HS | 188 000 | 18.4 | 1100 | 22.6 | 17 500 | 24.8 |
Some college/associate degree | 362 000 | 35.3 | 1800 | 39.6 | 28 000 | 39.4 |
Bachelor’s+ | 400 000 | 39 | 1300 | 27.5 | 20 000 | 28.2 |
Missing | 5300 | 0.5 | 100 | 2.4 | 30 | 0.0 |
Poverty, % | ||||||
<100 | 156 000 | 15.2 | 950 | 20.6 | 13 000 | 18.0 |
100–299 | 391 000 | 38.2 | 1800 | 39.9 | 32 000 | 44.7 |
300–499 | 252 000 | 24.6 | 950 | 21 | 15 500 | 21.6 |
500–999 | 211 000 | 20.6 | 700 | 15.1 | 10 000 | 13.9 |
Other (group quarters, <14 years, foster care) | 15 000 | 1.5 | 150 | 3.5 | 1300 | 1.8 |
Household type | ||||||
Family HH married | 737 000 | 72 | 2700 | 59 | 47 500 | 66.8 |
Family HH not married | 277 000 | 27 | 1800 | 37.8 | 23 000 | 32.4 |
Nonfamily HH | 6300 | 0.6 | 60 | 1.3 | 500 | 0.8 |
Missing | 4100 | 0.4 | 100 | 2.1 | 0 | 0.0 |
Grandparent in HH | ||||||
Yes | 42 500 | 4.2 | 250 | 4.9 | 12 000 | 17.0 |
Subfamily in HH | ||||||
Yes | 74 000 | 7.2 | 350 | 7.8 | 18 500 | 26.0 |
Disclosure review board approval number: CBDRB-FY22-CES004-045. AI/AN, American Indian/Alaska Native; API, Asian American and Pacific Islander; GED, graduate equivalent degree; HH, household; HS, high school; NH, non-Hispanic.
Statistical Analyses
First, we generated descriptive analyses to display demographic characteristics for the sample overall, for children who died, and for children who lost a caregiver. Second, we displayed the mortality outcomes and demographic characteristics by COI. Third, we conducted Cox proportional hazard models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for child and caregiver mortality. All models used weighted, unrounded cell counts and included a frailty term to account for the clustering of individuals within census tracts (see Supplemental Information for sample code).23 We calculated person-years-contributed and time-to-event for all causes of death, and censored person-years at the end of 2015 for all children or household adults alive at that time (ie, no NDI record linkage). As a first step, we estimated unadjusted HRs for both child and caregiver mortality outcomes. Next, we conducted an identical series of 3 adjusted models to estimate mortality risk for children and their caregivers:
basic adjustment (ie, child’s age and sex);
additional adjustment for household characteristics (ie, education, income-to-needs ratio, family structure, subfamilies in household, grandparent in family); and
full adjustment, which additionally included child’s race/ethnicity.
We note that model 2 adds variables that could be on the pathway connecting COI to mortality risk. Finally, we conducted an identical series of models to examine cause-specific child mortality outcomes.
In sensitivity analyses, we repeated our main models (1) using COI percentile as a continuous score, (2) considering each subdomain of the COI separately, and (3) including state as fixed effects, to account for potential differences in the relationship between the COI and mortality by state. Analyses were conducted using SAS and follow disclosure rules that require all cell counts to be rounded to 4 significant digits.
Results
Sample Characteristics
Over the 11-year follow-up, 4600 children among the 1 025 000 included in the analysis died; 12% of these deaths were among children from birth to 5 years at baseline, 25% were among children ages 6 to 11 years at baseline; the remaining 63% were among children 12 to 17 years at baseline (Table 2). Children who died over the follow-up period were also disproportionately likely to be male, non-Hispanic Black, and live in households with lower educational attainment and income, and less likely to live in a household with married adults. A total of 71 000 children experienced the death of a caregiver. These children were disproportionately older, more likely to be non-Hispanic Black, and to live in households with lower educational attainment and income, and with a grandparent or in a household with multiple families.
Mortality among children disproportionately occurred among those in very low (26%) and low (23%) opportunity neighborhoods, as compared with the high (17%) and very high (14%) neighborhoods (Table 3). The same pattern is present for caregiver deaths, with 27 and 24% of these deaths occurring in very low and low opportunity neighborhoods, as compared with 16 and 13% of these deaths in high and very high opportunity neighborhoods, respectively. All sociodemographic characteristics, other than child’s age and sex, displayed strong patterns by neighborhood. For example, whereas only 9% of white children lived in very low opportunity neighborhoods, this was the case for 38% of Hispanic, 55% of non-Hispanic Black, and 50 of non-Hispanic American Indian/Alaska Native children. In contrast, whereas approximately one-quarter of non-Hispanic white and one-third of non-Hispanic Asian American and Pacific Islander children lived in neighborhoods categorized as very high opportunity, this was the case for <5% of non-Hispanic Black and non-Hispanic American Indian/Alaska Native children and <8% of Hispanic children.
Deaths and Sample Characteristics Stratified by Quintiles of the Child Opportunity Index
. | First (Very Low) . | Second (Low) . | Third (Moderate) . | Fourth (High) . | Fifth (Very High) . | |||||
---|---|---|---|---|---|---|---|---|---|---|
n = 206 000 . | n = 212 000 . | n = 207 000 . | n = 196 000 . | n = 203 000 . | ||||||
N . | Row % . | n . | Row % . | n . | Row % . | n . | Row % . | n . | Row % . | |
Child deaths, all cause | 1200 | 25.6 | 1100 | 23.2 | 950 | 20.3 | 800 | 16.8 | 650 | 14.2 |
Caregiver deaths, all cause | 19 500 | 27 | 17 500 | 24.2 | 14 000 | 19.6 | 11 500 | 15.9 | 9400 | 13.2 |
Child’s age, y | ||||||||||
0–5 | 68 000 | 21.3 | 67 500 | 21.2 | 64 000 | 20.2 | 59 500 | 18.7 | 59 000 | 18.6 |
6–11 | 66 000 | 19.6 | 69 000 | 20.4 | 68 000 | 20.1 | 65 000 | 19.3 | 69 500 | 20.6 |
12–17 | 72 500 | 19.6 | 75 500 | 20.5 | 75 000 | 20.4 | 71 500 | 19.4 | 74 500 | 20.1 |
Child’s sex | ||||||||||
Male | 105 000 | 20 | 109 000 | 20.8 | 107 000 | 20.3 | 101 000 | 19.2 | 104 000 | 19.8 |
Female | 101 000 | 20.2 | 103 000 | 20.7 | 101 000 | 20.2 | 95 500 | 19.1 | 99 000 | 19.8 |
Child’s race/ethnicity | ||||||||||
NH white | 61 500 | 9.4 | 124 000 | 19.2 | 148 000 | 22.9 | 153 000 | 23.5 | 162 000 | 25 |
NH Black | 60 500 | 55.3 | 22 500 | 20.7 | 12 500 | 11.6 | 8600 | 7.8 | 5200 | 4.7 |
NH API | 4900 | 11.7 | 7000 | 16.6 | 7500 | 17.9 | 8400 | 20.1 | 14 000 | 33.7 |
NH AI/AN | 5000 | 50.2 | 2300 | 23.1 | 1500 | 14.2 | 800 | 8 | 400 | 4.2 |
Hispanic | 68 000 | 37.8 | 48 500 | 27 | 30 000 | 16.8 | 19 000 | 10.7 | 14 000 | 7.7 |
Other | 6200 | 18.5 | 7200 | 21.6 | 6900 | 20.5 | 6400 | 19.1 | 6800 | 20.3 |
Highest education in household | ||||||||||
<HS | 35 000 | 50.2 | 19 000 | 27.1 | 10 000 | 14.5 | 4300 | 6.1 | 1500 | 2.2 |
GED/HS | 60 000 | 31.9 | 51 500 | 27.4 | 39 000 | 20.7 | 26 000 | 13.8 | 12 000 | 6.3 |
Some college/associate degree | 78 500 | 21.8 | 87 500 | 24.2 | 83 000 | 23 | 69 000 | 19.1 | 43 000 | 11.9 |
Bachelor’s+ | 30 500 | 7.6 | 53 000 | 13.3 | 74 000 | 18.5 | 96 000 | 24 | 146 000 | 36.6 |
Missing | 1900 | 37 | 1100 | 21.7 | 950 | 17.7 | 700 | 13.3 | 550 | 10.4 |
Household poverty, % | ||||||||||
<100 | 71 000 | 45.5 | 40 000 | 25.4 | 24 000 | 15.6 | 13 500 | 8.7 | 6900 | 4.4 |
100–299 | 96 000 | 24.5 | 100 000 | 25.7 | 88 500 | 22.6 | 67 500 | 17.3 | 39 000 | 10 |
300–499 | 26 500 | 10.4 | 47 000 | 18.6 | 58 500 | 23.1 | 62 500 | 24.8 | 58 000 | 23 |
500–999 | 9000 | 4.3 | 21 000 | 9.9 | 33 500 | 15.8 | 50 000 | 23.8 | 97 500 | 46.3 |
Other (group quarters, <14 years, foster care) | 4300 | 28.3 | 3900 | 25.4 | 3200 | 21 | 2400 | 15.5 | 1500 | 9.8 |
Household type | ||||||||||
Family HH married | 103 000 | 14 | 144 000 | 19.5 | 155 000 | 21 | 159 000 | 21.5 | 177 000 | 24 |
Family HH not married | 100 000 | 36.2 | 66 000 | 23.9 | 50 000 | 18.1 | 36 000 | 12.9 | 24 500 | 8.9 |
Nonfamily HH | 1600 | 25.1 | 1600 | 25.9 | 1500 | 23.2 | 1000 | 16.1 | 600 | 9.8 |
Missing | 1500 | 36.1 | 900 | 21.7 | 750 | 17.8 | 550 | 13.8 | 450 | 10.7 |
Grandparent in HH | ||||||||||
Yes | 194 000 | 19.8 | 201 000 | 20.4 | 198 000 | 20.2 | 190 000 | 19.3 | 199 000 | 20.2 |
No | 12 000 | 27.6 | 11 500 | 27 | 8800 | 20.8 | 6400 | 15 | 4100 | 9.6 |
Subfamily in HH | ||||||||||
Yes | 180 000 | 18.9 | 193 000 | 20.4 | 194 000 | 20.4 | 187 000 | 19.7 | 197 000 | 20.7 |
No | 26 000 | 35.4 | 19 000 | 25.4 | 13 500 | 18.4 | 9200 | 12.4 | 6300 | 8.5 |
. | First (Very Low) . | Second (Low) . | Third (Moderate) . | Fourth (High) . | Fifth (Very High) . | |||||
---|---|---|---|---|---|---|---|---|---|---|
n = 206 000 . | n = 212 000 . | n = 207 000 . | n = 196 000 . | n = 203 000 . | ||||||
N . | Row % . | n . | Row % . | n . | Row % . | n . | Row % . | n . | Row % . | |
Child deaths, all cause | 1200 | 25.6 | 1100 | 23.2 | 950 | 20.3 | 800 | 16.8 | 650 | 14.2 |
Caregiver deaths, all cause | 19 500 | 27 | 17 500 | 24.2 | 14 000 | 19.6 | 11 500 | 15.9 | 9400 | 13.2 |
Child’s age, y | ||||||||||
0–5 | 68 000 | 21.3 | 67 500 | 21.2 | 64 000 | 20.2 | 59 500 | 18.7 | 59 000 | 18.6 |
6–11 | 66 000 | 19.6 | 69 000 | 20.4 | 68 000 | 20.1 | 65 000 | 19.3 | 69 500 | 20.6 |
12–17 | 72 500 | 19.6 | 75 500 | 20.5 | 75 000 | 20.4 | 71 500 | 19.4 | 74 500 | 20.1 |
Child’s sex | ||||||||||
Male | 105 000 | 20 | 109 000 | 20.8 | 107 000 | 20.3 | 101 000 | 19.2 | 104 000 | 19.8 |
Female | 101 000 | 20.2 | 103 000 | 20.7 | 101 000 | 20.2 | 95 500 | 19.1 | 99 000 | 19.8 |
Child’s race/ethnicity | ||||||||||
NH white | 61 500 | 9.4 | 124 000 | 19.2 | 148 000 | 22.9 | 153 000 | 23.5 | 162 000 | 25 |
NH Black | 60 500 | 55.3 | 22 500 | 20.7 | 12 500 | 11.6 | 8600 | 7.8 | 5200 | 4.7 |
NH API | 4900 | 11.7 | 7000 | 16.6 | 7500 | 17.9 | 8400 | 20.1 | 14 000 | 33.7 |
NH AI/AN | 5000 | 50.2 | 2300 | 23.1 | 1500 | 14.2 | 800 | 8 | 400 | 4.2 |
Hispanic | 68 000 | 37.8 | 48 500 | 27 | 30 000 | 16.8 | 19 000 | 10.7 | 14 000 | 7.7 |
Other | 6200 | 18.5 | 7200 | 21.6 | 6900 | 20.5 | 6400 | 19.1 | 6800 | 20.3 |
Highest education in household | ||||||||||
<HS | 35 000 | 50.2 | 19 000 | 27.1 | 10 000 | 14.5 | 4300 | 6.1 | 1500 | 2.2 |
GED/HS | 60 000 | 31.9 | 51 500 | 27.4 | 39 000 | 20.7 | 26 000 | 13.8 | 12 000 | 6.3 |
Some college/associate degree | 78 500 | 21.8 | 87 500 | 24.2 | 83 000 | 23 | 69 000 | 19.1 | 43 000 | 11.9 |
Bachelor’s+ | 30 500 | 7.6 | 53 000 | 13.3 | 74 000 | 18.5 | 96 000 | 24 | 146 000 | 36.6 |
Missing | 1900 | 37 | 1100 | 21.7 | 950 | 17.7 | 700 | 13.3 | 550 | 10.4 |
Household poverty, % | ||||||||||
<100 | 71 000 | 45.5 | 40 000 | 25.4 | 24 000 | 15.6 | 13 500 | 8.7 | 6900 | 4.4 |
100–299 | 96 000 | 24.5 | 100 000 | 25.7 | 88 500 | 22.6 | 67 500 | 17.3 | 39 000 | 10 |
300–499 | 26 500 | 10.4 | 47 000 | 18.6 | 58 500 | 23.1 | 62 500 | 24.8 | 58 000 | 23 |
500–999 | 9000 | 4.3 | 21 000 | 9.9 | 33 500 | 15.8 | 50 000 | 23.8 | 97 500 | 46.3 |
Other (group quarters, <14 years, foster care) | 4300 | 28.3 | 3900 | 25.4 | 3200 | 21 | 2400 | 15.5 | 1500 | 9.8 |
Household type | ||||||||||
Family HH married | 103 000 | 14 | 144 000 | 19.5 | 155 000 | 21 | 159 000 | 21.5 | 177 000 | 24 |
Family HH not married | 100 000 | 36.2 | 66 000 | 23.9 | 50 000 | 18.1 | 36 000 | 12.9 | 24 500 | 8.9 |
Nonfamily HH | 1600 | 25.1 | 1600 | 25.9 | 1500 | 23.2 | 1000 | 16.1 | 600 | 9.8 |
Missing | 1500 | 36.1 | 900 | 21.7 | 750 | 17.8 | 550 | 13.8 | 450 | 10.7 |
Grandparent in HH | ||||||||||
Yes | 194 000 | 19.8 | 201 000 | 20.4 | 198 000 | 20.2 | 190 000 | 19.3 | 199 000 | 20.2 |
No | 12 000 | 27.6 | 11 500 | 27 | 8800 | 20.8 | 6400 | 15 | 4100 | 9.6 |
Subfamily in HH | ||||||||||
Yes | 180 000 | 18.9 | 193 000 | 20.4 | 194 000 | 20.4 | 187 000 | 19.7 | 197 000 | 20.7 |
No | 26 000 | 35.4 | 19 000 | 25.4 | 13 500 | 18.4 | 9200 | 12.4 | 6300 | 8.5 |
Disclosure review board approval number: CBDRB-FY22-CES004-045. AI/AN, American Indian/Alaska Native; API, Asian American and Pacific Islander; GED, graduate equivalent degree; HH, household; HS, high school; NH, non-Hispanic.
Child Opportunity Index and Child Mortality
Figure 1a displays the unadjusted Cox regression HRs comparing the risk of child mortality over the follow-up on the basis of COI quintiles (see values in Supplemental Table 6) without conditioning on other factors, illustrating an inverse stepwise relationship between COI and child mortality. In a model adjusted for child’s age and sex (Table 4, model 1), children in very low opportunity neighborhoods at baseline had 1.4 times the risk of dying over the follow-up period (95% CI, 1.211–1.537) relative to those in very high opportunity neighborhoods. The HRs were attenuated after adjustment for household socioeconomic characteristics (model 2), and the results remained similar after adjustment for child’s race/ethnicity (model 3). In this fully adjusted model, children in very low neighborhoods at baseline had 1.15 (95% CI, 0.98–1.34) times the risk of dying over the follow-up period than those in very high opportunity neighborhoods.
Adjusted Cox Regression HR and 95% CI for All-Cause Child Deaths And Death of a Caregiver Over the 7-Year Follow-Up Period (N = 1 025 000 Children)
. | Model 1 . | Model 2 . | Model 3 . | |||
---|---|---|---|---|---|---|
. | HR . | (95% CI) . | HR . | (95% CI) . | HR . | (95% CI) . |
Child deaths | ||||||
Opportunity quintiles | ||||||
First (lowest) | 1.365 | (1.211–1.537) | 1.115 | (0.965–1.289) | 1.145 | (0.978–1.341) |
Second | 1.244 | (1.101–1.405) | 1.072 | (0.935–1.231) | 1.105 | (0.956–1.276) |
Third | 1.199 | (1.058–1.358) | 1.064 | (0.925–1.223) | 1.090 | (0.943–1.260) |
Fourth | 1.033 | (0.911–1.172) | 0.956 | (0.835–1.094) | 0.968 | (0.841–1.115) |
Fifth (highest) | 1 | — | 1 | — | 1 | — |
Caregiver deaths | ||||||
Opportunity quintiles | ||||||
First (lowest) | 1.596 | (1.527–1.667) | 1.231 | (1.166–1.300) | 1.300 | (1.227–1.376) |
Second | 1.474 | (1.412–1.539) | 1.200 | (1.139–1.264) | 1.247 | (1.183–1.314) |
Third | 1.353 | (1.295–1.414) | 1.186 | (1.129–1.245) | 1.212 | (1.154–1.274) |
Fourth | 1.203 | (1.151–1.258) | 1.106 | (1.052–1.162) | 1.105 | (1.051–1.162) |
Fifth (highest) | 1 | — | 1 | — | 1 | — |
. | Model 1 . | Model 2 . | Model 3 . | |||
---|---|---|---|---|---|---|
. | HR . | (95% CI) . | HR . | (95% CI) . | HR . | (95% CI) . |
Child deaths | ||||||
Opportunity quintiles | ||||||
First (lowest) | 1.365 | (1.211–1.537) | 1.115 | (0.965–1.289) | 1.145 | (0.978–1.341) |
Second | 1.244 | (1.101–1.405) | 1.072 | (0.935–1.231) | 1.105 | (0.956–1.276) |
Third | 1.199 | (1.058–1.358) | 1.064 | (0.925–1.223) | 1.090 | (0.943–1.260) |
Fourth | 1.033 | (0.911–1.172) | 0.956 | (0.835–1.094) | 0.968 | (0.841–1.115) |
Fifth (highest) | 1 | — | 1 | — | 1 | — |
Caregiver deaths | ||||||
Opportunity quintiles | ||||||
First (lowest) | 1.596 | (1.527–1.667) | 1.231 | (1.166–1.300) | 1.300 | (1.227–1.376) |
Second | 1.474 | (1.412–1.539) | 1.200 | (1.139–1.264) | 1.247 | (1.183–1.314) |
Third | 1.353 | (1.295–1.414) | 1.186 | (1.129–1.245) | 1.212 | (1.154–1.274) |
Fourth | 1.203 | (1.151–1.258) | 1.106 | (1.052–1.162) | 1.105 | (1.051–1.162) |
Fifth (highest) | 1 | — | 1 | — | 1 | — |
Model 1 is adjusted for child’s age and sex. Model 2 is additionally adjusted for highest educational attainment in the household, household poverty level, household hold type, grandparent in household, and subfamily in household. Model 3 is additionally adjusted for child’s race/ethnicity. Disclosure review board approval number: CBDRB-FY22-CES004-045. —, not applicable.
HRs and 95% CIs for (a) all-cause child death and (b) caregiver death, by COI Index quintile, over the 11-year follow-up period (n =1 025 000 children). HRs produced using bivariate Cox regression models. Disclosure review board approval number: CBDRB-FY22-CES004-011.
HRs and 95% CIs for (a) all-cause child death and (b) caregiver death, by COI Index quintile, over the 11-year follow-up period (n =1 025 000 children). HRs produced using bivariate Cox regression models. Disclosure review board approval number: CBDRB-FY22-CES004-011.
In models estimating cause-specific child mortality risk (Supplemental Table 7), associations were evident for all external causes of death and firearm-related deaths, but not suicide deaths. For example, for all external causes of death, the HRs for the comparison of very low to very high neighborhood opportunity ranged from 1.30 (95% CI, 1.13–1.50) in the bivariate model (model 1) to 1.18 (95% CI, 0.99–1.42) in model 4. For firearm-related deaths, the HRs for the comparison of very low to very high neighborhood opportunity ranged from 2.79 (95% CI, 2.04–3.83) in the bivariate model to 1.50 (95% CI, 1.00–2.24) in model 4.
Child Opportunity Index and Caregiver Mortality
Figure 1b displays the Cox regression HRs comparing the risk of having a caregiver die over the follow-up period (see values in Supplemental Table 6). Consistent with child mortality, the COI displayed a graded inverse stepwise relationship with death of a caregiver, with children in very low opportunity neighborhoods at baseline having 1.57 times the risk of experiencing the death of a caregiver (95% CI, 1.50–1.64) over the follow-up relative to those in very high opportunity neighborhoods. The HRs were attenuated after adjustment for household characteristics, yet the graded pattern remained, ranging from 1.23 (95% CI, 1.17–1.30) to 1.11 (95% CI, 1.05–1.16) for COI quintile comparisons of very low versus very high and high versus very high, respectively (Table 4, Model 2). The HRs increased slightly after further adjustment for child’s race/ethnicity (Table 4, Model 3).
Sensitivity Analyses
First, we repeated our main models for child and adult mortality using the COI as a continuous measure (Supplemental Table 8). On the basis of this model, we estimate that a 1 percentile increase in COI score is associated with approximately a 0.5% reduction in risk of death over the follow-up period, for deaths among children or the adults in their households. Second, we repeated our main models using each COI domain as separate exposure (Supplemental Table 9); the patterns and magnitudes are generally similar across the 3 subscales. Finally, as shown in Supplemental Table 10, the estimated HRs are largely unchanged with inclusion of state fixed effects in the model.
Discussion
Using a representative, population-based sample with >1 million children followed for 9 years, and the COI as a measure of neighborhood opportunity, we found that very low neighborhood opportunity was associated with elevated risk for mortality among both children and their caregivers. This pattern was evident for all-cause mortality and external causes of death. Importantly, the health consequences of increased mortality are noteworthy for the survivors. For children, the death of a caregiver is a major loss of a critical relationship that typically buffers the effects of adversity on their health.17,24–26 For adults, the death of a child increases the risk for a wide range of stress-related physical and mental health conditions.27–29 Furthermore, the marked discrepancies in neighborhood opportunity by race and ethnicity, as documented in Table 3, call for greater attention to the influence of structural inequities associated with residential segregation driven by racism that perpetuate the disparities in mortality among neighborhoods observed in this study.
The finding that very low neighborhood opportunity was associated with an increased risk of both child and caregiver mortality is consistent with extensive research showing specific neighborhood-level characteristics such as poverty and racial residential segregation are associated with morbidity30,31 and mortality32,33 among adults, and poorer health among children (eg, externalizing disorders, low physical activity34 ). However, given the sample size required for measuring child mortality, there are few studies of neighborhood effects on this outcome, with 1 study finding that children living in poorer neighborhoods had elevated risk of mortality in the emergency department35 and another examining the “causal effect” of being offered a housing voucher on mortality risk, which was imprecise and varied by child gender.36
Our unique data sources, which include a large, population-based sample, prospective follow-up, detailed covariates to control for potential confounders, and a wide-ranging measure of neighborhood opportunity that captures compounding forms of inequality, allow us to extend these previous studies to include outcomes for children and their caregivers. Although further research is needed to elucidate causal mechanisms that explain these findings, several hypotheses are worthy of investigation. For example, longstanding institutional neglect and disinvestment contribute to weakened neighborhood infrastructures that undermine social cohesion and increase the prevalence of violent crime.37 Other research has shown that disadvantaged neighborhood contexts are associated with elevated, stress-related biological changes,38–43 cardiovascular risk,44,45 externalizing behaviors,34 injuries,46,47 and suicidal symptoms,48,49 which could contribute to the increased mortality observed in this study. It will be important to identify structural, behavioral, and biological mechanisms that might explain the findings in a way that could help to prioritize points for intervention. Furthermore, intergenerational research is needed considering the challenges of interpreting our fully adjusted models. Given that a parent’s neighborhood during their childhood shapes their education, income, household formation, and neighborhood during adulthood (ie, as parents), the models that include parental socioeconomic variables inappropriately control for the effect of persistent multigenerational neighborhood disadvantage that likely shapes children’s mortality risks to a strong degree for a subset of children.50 Moreover, in future work, it will be valuable to consider state-level characteristics, including gun laws and health care access, as potential modifiers of the relationship between neighborhood opportunity and mortality.
This study has some limitations. First, the MDAC linkage is not representative of infant mortality outcomes in the United States, thus limiting the generalizability of our results. Second, given the structure of the ACS household roster, we cannot identify parents from other adults in the home. On this basis, we do not know with certainty that the adult in the household who died during the follow-up period had a caregiving or close relationship with the child. Although this is a significant limitation, the vast majority of children in the United States live in households with a biological or adoptive parent (ie, an estimated 4% of US children live with relatives or nonrelatives in foster care or less formal care arrangements outside of the foster care system51 ). Our analyses adjusted for the presence of subfamilies or grandparents, allowing us to control for these characteristics. Furthermore, death of any household adult is a meaningful outcome given its contribution to intergenerational inequalities. Third, we do not have information on the length of residence in a given neighborhood context, which has 2 implications. This limitation (1) introduces variability in the exposure variable, and (2) prohibits questions about duration of exposure2 or movement into or out of disadvantaged contexts.36,52
Conclusions
Since the 1990s, the United States has ranked lowest relative to 19 comparative nations for child mortality,53 and between 2012 and 2019, deaths among individuals ages 5 to 25 have increased.54 Our study finds that neighborhood quality, as quantified by the COI, is associated with excess risk for mortality in children and their caregivers. For caregiver deaths, in particular, a graded association persisted after adjustment for social and economic characteristics of the household. Future studies are needed to explore further the potential value of the COI for identifying geographic patterns in health status indicators and behaviors that increase mortality risks for children and their caregivers within causally informative research designs. This line of research has the potential to inform innovative strategies for place-based and neighborhood mobility interventions to reduce preventable causes of death that contribute to disparities within and across generations. Furthermore, more nuanced research on specific components of the COI in relation to mortality is essential for informing policy reform related to environmental, educational, and health care systems. In addition, to advance both research and public health practice, further research is needed to compare the COI with other tract-level measures, including specific individual indicators of privileged contexts,55,56 historically informed measures of disinvestment (eg, redlining),57,58 and other multicomponent, neighborhood-level measures of racism59–61 and economic security.62,63
Dr Slopen conceptualized and designed the study, and drafted the initial manuscript; Dr Noelke conceptualized and designed the study, critically reviewed the manuscript, and made intellectual contributions related to the interpretation of the results and revisions to the text; Ms Cosgrove conducted the statistical analysis, critically reviewed the manuscript, and made intellectual contributions to the text; Drs Acevedo-Garcia, Hatzenbuehler, and Shonkoff critically reviewed the manuscript and made intellectual contributions related to the interpretation of the results and revisions to the text; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2022-060498.
This article is released to inform interested parties of research and to encourage discussion. Any opinions and conclusions expressed herein are those of the authors and do not reflect the views of the US Census Bureau. These results have been reviewed by the Census Bureau’s disclosure review board to ensure that no confidential information is disclosed. The disclosure review board’s release number is: CBDRB-FY22-CES004-045. This work was conducted with support from Harvard Catalyst | The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health award #UL1 TR002541) and financial contributions from Harvard University and its affiliated academic health care centers. The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University, and its affiliated academic health care centers, or the National Institutes of Health.
FUNDING: Supported by an award from the Foster Family Funds and the Chan Zuckerberg Initiative/Silicon Valley Community Foundation to the Center on the Developing Child at Harvard University, the Robert Wood Johnson Foundation (grant 71192), and the W.K. Kellogg Foundation (grant P3036220).
CONFLICT OF INTEREST DISCLAIMER: The authors have indicated they have no conflicts of interest relevant to this article to disclose.
Comments