Place-Based Child Opportunity at Birth and Child Development from Infancy to Age 4

Abstract

Objective

To evaluate whether the children’s neighborhood quality, as a measure of place-based social determinants of health, is associated with the odds of developmental delay and developmental performance up to age 4 years.

Study Design

Mothers of 5702 children from the Upstate KIDS Study, a longitudinal population-based cohort of children born from 2008 through 2010, provided questionnaire data and a subset of 573 children participated in a clinic visit. The Child Opportunity Index 2.0 (COI) was linked to home census tract at birth. Probable developmental delays were assessed by the Ages and Stages Questionnaire© up to 7 times between 4 and 36 months, and developmental performance was assessed via the Battelle Developmental Inventory at age 4 years.

Results

In unadjusted models, higher neighborhood opportunity was protective against developmental delays and was associated with slightly higher development scores at age 4. After adjusting for family-level confounding variables, ten-point higher COI (on a 100-point scale) remained associated with a lower odds of any developmental delay (OR = .966, 95%CI = .940-.992), and specifically delays in the personal-social domain (OR = .921, 95%CI = .886-.958), as well as better development performance in motor (B = 0.79, 95%CI = 0.11–1.48), personal-social (B = 0.64, 95%CI = 0.003–1.28), and adaptive (B = 0.69, 95%CI = 0.04–1.34) domains at age 4.

Conclusions

Community-level opportunities are associated with some aspects of child development prior to school entry. A greater understanding of neighborhood quality may be helpful to pediatric providers to inform targeted developmental screening.

Place-based social determinants of health are environmental factors in the places people inhabit that have the potential to impact their health. Examples of place-based social determinants include neighborhood resources like access to quality education, healthy food, and safe and healthy play spaces.¹ Understanding the role of neighborhoods in child development is important because early child development is a foundation for academic performance and mental health.^2–4 Some of the most convincing evidence for the impact of neighborhood on child development comes from the Moving to Opportunity (MTO) study. MTO found that being randomized to a housing voucher before age 13 was associated with better adult economic outcomes including higher likelihood of college attendance, higher college quality, higher earnings, and living in a neighborhood with higher mean income and lower poverty rate.⁵ Similar findings have also been reported from other pseudo-random studies about moving to neighborhoods with less poverty.^{6, 7} Although randomized trials like MTO provide the best evidence of causality, it is uncommon for families to make moves from high- to low-poverty neighborhoods. Most areas have little affordable housing in low-poverty neighborhoods, and families must make complex multifactor decisions about where to live that take into account the family’s history of housing instability as well as balance current and future needs – factors which are often incompatible with making big changes in neighborhood.^{8, 9}

Cohort studies generally find detrimental associations between neighborhood poverty and child academic, social, and behavioral performance.¹⁰ Some previous studies have explored associations between multidimensional indices of neighborhood affordances and child development,^{11, 12} but none have employed the Child Opportunity Index. This study uses the Child Opportunity Index 2.0 (COI), composed of 29 education, health & environment, and social & economic indicators. Although intended to be relevant to children, COI indicators were selected and weighted relative to adult mental and physical health and social mobility.¹³ The COI has demonstrated construct validity for predicting child hospitalizations¹⁴ and cardiometabolic health,¹⁵ but the COI remains to be validated with respect to early child development.

Two literature reviews concluded that neighborhood effect sizes on child development are modest and that more research is needed.^{10, 16} Many high poverty neighborhoods contain social programs (e.g., subsidized preschool) and resources (e.g., parks, libraries, etc) that are not accounted for in research. Evidence is scarce about which combinations of neighborhood resources best predict child developmental outcomes.^{10, 16} Hence, more research about neighborhood institutional resources such as access to high quality preschools and primary schools, and healthy environments, such as green space, walkability, and extreme heat exposure, is needed.¹⁷ These indicators are included in the COI education and health & environment subdomains, which are unique among multi-indicator neighborhood indices.¹⁸

Given the potential importance of neighborhood deprivation early in a child’s life,^{19, 20} this study focused on neighborhood COI score at birth to assess associations with child development from infancy to 4 years of age. A range of family-level covariates were included to help determine whether the COI was uniquely associated with child development over and above family-level factors that might also partially determine where a family lives. Finally, residential instability (i.e., moves) and social mobility (i.e., changes in COI) were explored to examine longitudinal neighborhood changes. We hypothesized that higher neighborhood opportunity at birth would be associated with a lower risk of developmental delay and higher developmental performance, even when adjusting for family-level covariates.

Methods

Participants

Upstate KIDS is a population-based birth cohort designed to evaluate the impact of infertility treatment on child health.²¹ Live births in 2008–2010 in 57 counties from New York (excluding New York City) were sampled, stratified by the state’s seven regional perinatal networks. Singletons conceived by infertility treatment were oversampled (1 treatment: 3 no treatment), and all multiple births were invited to participate regardless of treatment. In a previous study, no association between infertility treatment and early development was observed after accounting for plurality.²² Therefore, the total sample is used and treatment is included as a covariate to account for oversampling.

A total of 6,171 children from 5,034 families were recruited into the study when children were 4 months old. Main analyses were restricted to 5,702 children from 4,599 families with COI and developmental delay data for at least one timepoint from 4 to 36 months of age, including 3,589 singletons and 2,113 multiples. A subgroup of children was invited to participate in a developmental assessment at age 4 years including (1) children who screened positive for a developmental delay on the Ages and Stages Questionnaire©²³ (ASQ) at age 30–36 months or for autism on the Modified Checklist for Autism in Toddlers²⁴ (MCHAT) at 18–24 months; (2) children referred to the New York State Early Intervention Program for evaluation; (3) twins and higher order multiples (regardless of their screening status); and (4) a random selection of singleton children who passed both the ASQ and MCHAT. Of 2,154 children invited, 601 (27.9%) participated and COI data at birth were available for 573 children from 426 families. Most children (458; 80%) passed the ASQ and MCHAT, and 114 (20%) failed one or both. Nearly half of the children (n = 277, 48%) were twins or higher-order multiples.

Human subjects research approval was obtained from participating institutions (NYSDOH IRB #07–097; UAlbany #08–179), and informed consent was obtained. Families received compensation for returning questionnaires and for the developmental assessment.

Measures

Neighborhood opportunity.

The Child Opportunity Index 2.0 (COI) is a national, place-based measure of social determinants of healthy child development.^{1, 25} The COI includes 11 education indicators covering educational resources from high quality preschools through postsecondary enrollment, 10 health & environment indicators assessing healthy environments, toxic exposures, and health insurance coverage, and 8 social & economic indicators such as adult employment, poverty and income, public assistance rate, and single-headed households.

Home addresses were extracted from vital records and geocoded into census tracts using the U.S. Census geocoder.²⁶ Geocoded addresses were then linked to the census-level Child Opportunity Index 2.0 data²⁷ from 2010. Although the study sample was limited to births in New York State, we used the nationally normed COI that can be compared more broadly across the US and other COI studies. The COI is scored from 1–100 with a mean of 50 across all available census tracts in the United States, as are the Education, Health & Environment, and Social & Economic subdomains.

In addition to the continuously distributed COI and subdomains, census tracts are categorized into quintiles called very low, low, medium, high, and very high. To account for residential mobility after birth, we computed two additional variables to use in sensitivity analyses: (1) residential instability, a count of the number of different addresses a child lived in between birth and age 3, and (2) social mobility, or the difference in COI quintile category between birth and age 3 categorized as decreased (0), the same (1), or increased (2).

Child Development.

Mothers completed the Ages and States Questionnaire (ASQ), second edition²⁸ at 4, 8, and 12 months, and the third edition²³ at 18, 24, 30, and 36 months. The ASQ is a validated screening instrument for developmental impairments in five domains: fine motor, gross motor, communication, personal-social functioning, and problem-solving skills.^{29, 30} ASQ items were scored as “yes” (10 points), “sometimes” (5 points), and “not yet” (0 points). Probable developmental delay on a given domain of the ASQ is defined by a score that is two or more standard deviations below the United States national average for the age group. A probable delay on any domain of the ASQ was also examined. Cutoff scores were used rather than continuous scores because there are multiple versions of the ASQ for different ages. The questions and cut-off values for probable delay vary across child age, suggesting that the total score does not have the same meaning over time (measurement non-invariance).

At the 4-year clinic visit, a subset of children was administered the Battelle Developmental Inventory, Second Edition³¹ (BDI-2) by a trained staff member who was blinded to the COI. The BDI-2 is a standardized (M=100, SD=15) assessment of developmental skills with higher scores indicating better developmental performance. The BDI-2 provides a total score as well as 5 subdomain scores: motor, communication, cognitive, personal/social, and adaptive. The BDI-2 has demonstrated reliability, internal consistency, and content and criterion validity.³²

Covariates

Maternal age, parity, plurality, employment during pregnancy, and child sex were obtained from vital records and supplemented with maternal reports when children were 4 months of age when missing. Maternal education, race/ethnicity, infertility treatment use, and insurance status were primarily obtained from the questionnaire and supplemented from vital records where possible.

Statistical Analysis

The association of a 10-point increase in the COI and subdomains with the odds of probable developmental delay were explored with general linear mixed models with a logit link, a random intercept for children nested in families, and repeated effect of child age to account for repeated assessments of the ASQ. The association between a 10-point increase in the COI and Battelle scores were explored with general linear mixed models with an identity link and a random intercept for children nested in families. We chose a 10-point increase in the COI to reflect a meaningful change on a 100-point scale. Due to previous research showing varied neighborhood effects by child sex,^{33, 34} we explored potential interactions between the COI and child sex but all were nonsignificant for the ASQ. Although there were a few significant interactions for the Battelle, none of the simple effects was significant (data not shown), so we used the combined sample. Interactions between the COI and child age were modeled for the ASQ to determine whether associations between the COI and developmental delay varied by child age but no crossover effects were found (see Table I; online), so pooled analyses are presented.

Following unadjusted models, minimally and fully adjusted models were run. The minimally adjusted model covariates were selected to account for unique sample characteristics (eg, use of infertility treatment, plurality) and to increase precision of estimates due to associations with child development alone (child sex, parity, insurance status). Additional covariates in the fully adjusted model (maternal age, race/ethnicity, education, and working status) were selected as potential confounders of the association between neighborhood and child development. Specifically, maternal age, race/ethnicity, education, and employment may impact where a family is able to live (or chooses to live) as well as the child’s level of development. Race/ethnicity was included as a social covariate to account for historical and current processes such as structural racism and discrimination.³⁵

Because those who lived in lower COI neighborhoods at birth moved more between birth and age 3, r(5904) = −.25, p < .001, we computed a sensitivity analysis to assess whether the COI at birth contributed to child development at age 3 and age 4 when adjusting for residential instability between 0 and 3 years and social mobility (downward or upward change in COI quintile compared with same quintile from birth to age 3). For the analysis of the ASQ at age 3, inverse probability of attrition weights were computed and used in the model.³⁶

Results

Children lived in census tracts with COI averaging 62.03 (SD=24.41; range=1–100), meaning they lived in neighborhoods that were above average (50) on a national scale, but still ranged from the lowest to the highest level. Socio-demographics of the families in the main sample and the subsample that completed the BDI-2 are presented in Table II.

Table II.

Sample demographic characteristics

	Total Study Sample	Subsample with BDI-2 data
Number of families	4599 (100.00)	426 (9.3)
Child characteristics
Sex – male	2371 (51.55)	221 (51.88)
Gestational age (weeks) M (SD)	38.04 (2.50)	37.80 (2.62)
Birthweight (grams), M (SD)	3173.38 (697.14)	3098.23 (712.71)
Singleton pregnancy/Plurality	3589 (78.04)	297 (69.72)
Maternal characteristics
Age at birth (years), M (SD)	30.51 (6.05)	32.21 (5.61)
Educational attainment
Less than high school	270 (5.87)	8 (1.88)
High school or GED equivalent	576 (12.53)	31 (7.28)
Some college	1394 (30.32)	115 (27.00)
College	1024 (22.28)	118 (27.70)
Advanced degree	1333 (29.00)	154 (36.15)
Employment during pregnancy	3165 (68.82)	316 (74.18)
Non-Hispanic Asian	120 (2.61)	15 (3.52)
Non-Hispanic Black	222 (4.83)	18 (4.23)
Non-Hispanic White	3707 (80.60)	358 (84.04)
Hispanic	420 (9.13)	28 (6.57)
Mixed race/other	130 (2.83)	7 (1.64)
Fertility treatment	1379 (29.98)	172 (40.38)
Nulliparity	2086 (45.70)	193 (45.73)
Private insurance	3463 (75.36)	366 (86.12)
Neighborhood characteristics
Overall Child Opportunity Index, M (SD)	62.02 (26.41)	63.95 (25.15)
Educational COI, M (SD)	63.84 (24.32)	66.53 (23.18)
Health/Environmental COI, M (SD)	64.71 (26.38)	67.23 (24.89)
Social/Economic COI, M (SD)	59.65 (26.76)	60.94 (25.69)
Residential instability (# addresses), M (SD)	1.48 (.82)	1.44 (.79)
Social mobility
Decreased COI quintile	317 (7.04)	23 (5.45)
Same COI quintile	3769 (83.64)	358 (84.83)
Increased COI quintile	420 (9.32)	41 (9.72)

Note. Statistics presented are n (%) unless otherwise noted. For child characteristics, descriptive statistics were derived from all singletons and one randomly selected twin of each pair.

Missing data: private insurance (n=4;1), maternal educational attainment (n=2;0), nulliparity (n=34;4), social mobility (n=93;4).

Probable Developmental Delays

Overall, 21.8% of children were scored as having a probable delay on the ASQ in any domain at any time from 4 to 36 months (i.e., 78.2% of children never experienced a delay in any domain up to age 3). At any single time, 10% or fewer children were scored as having a probable delay in any domain. Unadjusted models suggested that a 10-point higher COI, and its three subdomains at birth, were associated with 2–9% lower odds of developmental delay from 4–36 months (Figure 1A; Table III, Model 0; online). All associations were significant for the total COI and the social & economic subdomain, and associations were slightly smaller for the education and health & environment subdomains. Results remained virtually the same after adjusting for sample characteristics and variables that contribute to developmental delays (Table 3; Model 1; online). Additionally, after adjusting for family-level socio-demographic characteristics (Figure 1B; Table III, Model 2; online), the total COI and the social & economic subdomain remained significantly associated with the ASQ total and personal-social subdomains, and the COI social & economic subdomain also remained associated with the ASQ fine motor domain.

Figure 1.

A. Unadjusted, and B. Fully-adjusted associations (OR, 95%CI) between COI and child probable developmental delays from 4–36 months

Assessed Developmental Performance

At age 4, children scored nearly one standard deviation above the mean on the Battelle Developmental Inventory (M=113.52, SD=18.59) reflecting the advantaged nature of the sample, but the range of scores was wide (46–145). Unadjusted models found that each 10-point increase on the COI or subdomains at birth was associated with .76 to 1.57 higher standardized scores on the Battelle (Figure 2A; Table IV, Model 0; online). Adjusting for sample characteristics and variables that contribute to developmental performance attenuated some associations with the education and health & environment subscales, but associations remained for the total COI and for the social & economic subdomain (Table IV, Model 1; online). Additional adjustment for family-level factors attenuated most associations, but the total COI, health & environment subdomain, and social & economic subdomain remained significant with child motor, personal-social, and adaptive scores (Figure 2B; Table IV, Model 2; online).

Figure 2.

A. Unadjusted, and B. Fully-adjusted associations (B, 95%CI) between COI and child development performance at age 4 years

Sensitivity Analysis: Residential Instability and Social Mobility

One-third of children (33.42%) moved at least once (and up to 8 times) between birth and age 3. Comparing the COI quintile group at birth and age 3, most children remained in the same category, but 7% moved into a lower quintile, and 9% moved into a higher quintile (see Table II). Associations between the COI at birth and ASQ scores at 3 years and Battelle scores at 4 years were computed accounting for residential instability (# of moves) and social mobility (change in COI) to determine whether the COI at birth was still associated with child development.

In ASQ models, residential instability was associated with higher odds of fine motor delay in all models (Table V; online). Adjusting for residential instability and social mobility, 10-point higher COI total and social & economic subdomain remained associated with lower odds of developmental delay overall and gross motor and personal-social delays. However, those effects attenuated to non-significance once covariates were controlled (despite similar effect sizes as the model above).

In Battelle models, residential instability was associated with 2–4 point lower scores on the total score and all subscales, but all estimates attenuated after adjusting for covariates (Table VI; online). Conversely, even when accounting for residential instability and social mobility, and adjusting for covariates, 10-point higher COI, health & environment, and social & economic scores were still associated with higher motor and adaptive scores. Unlike residential instability, social mobility was rarely associated with child development (data not shown).

Discussion

The COI, as a measure of place-based social determinants, was associated with child probable developmental delays from 4–36 months as well as developmental performance at 4 years. After fully adjusting for family and individual level factors, higher COI total and social & economic subscale scores were associated with lower odds of fine motor and personal-social delays; which are congruent with finding higher scores on the Battelle motor subscale and personal-social and/or adaptive subscales. It should be noted, however, that the sample size for Battelle analysis was small and the subsample that participated was even more socioeconomically advantaged than the total study sample (Table II). Associations were largely consistent in sensitivity analyses accounting for residential instability and social mobility. Residential instability was associated with higher odds of fine motor delay. One possible reason for this association could be that moving throws a house into disarray and may prevent parents from allowing their young child to freely explore the environment and develop their fine motor skills. These findings suggest that the environment a child is reared in can contribute to aspects of child development independent of family-level variables like maternal education, age, race/ethnicity, and employment.

These associations are notable because the sample was population-based, including rural, suburban, and urban areas in New York (although it excluded New York City) from a wide range of neighborhood opportunity. Inclusion of this broad range of neighborhoods promotes generalizability of the study findings. This study also modeled the COI as a continuous score, demonstrating that even among the relatively advantaged families in this sample, more resourced neighborhoods were associated with better outcomes. Still, it is possible that the neighborhood factors that promote child development differ depending on the type of neighborhood (e.g., urban vs. rural; racial/ethnic make-up; etc), and future research should explore potential moderators of these effects.¹⁰ Specifically, although capturing a large socioeconomically diverse sample, the study was predominantly non-Hispanic White. Future work should explore whether associations may differ for children from other racial/ethnic groups.^{33, 37}

One advantage of the COI is inclusion of a wide range of neighborhood factors. In this study, the COI education and health & environment subdomains were rarely associated with development after adjusting for child and family factors. The most robust associations were with the social & economic subdomain which includes commonly used single indicators like mean household income, as well as less common ones like high-skill employment and the public assistance rate.¹⁸ The current focus in the literature on socioeconomic neighborhood indicators may be justified.

There are several potential mechanisms that may explain how place affects child development. The direct path would suggest that having more resources available in the neighborhood supports children’s burgeoning skills. For example, high-quality preschools are associated with better school performance later,^{38, 39} and neighborhood green spaces promote outdoor play and motor development.⁴⁰ Neighborhood disinvestment and poverty may also contribute to chronic stress⁴¹ and can be internalized,⁴² perhaps interfering with learning. Neighborhood effects may also be indirect through brain structure and function,^{43, 44} the quality of the home environment,⁴⁵ and maternal stress and depression.^{12, 44–46} Impoverished neighborhoods may impact children indirectly through parental investments. Evidence suggests that parents living in impoverished neighborhoods may be more likely to hold scarcity mindsets⁴⁷ which can lead them to prioritize immediate family needs over longer-term investments in child development.^{48, 49} Future research should quantify each of the major environments in which a child is embedded – home, school or daycare, etc – to untangle their effects.

If findings are replicated, a child’s neighborhood (ie, family address) could be used for targeted developmental screening, as has been demonstrated in lead abatement research.⁵⁰ Despite recommendations for universal developmental screening,⁵¹ an estimated 37% of pediatricians did not screen all children under 3 years in 2016,⁵² and screening rates vary widely by state.⁵³ There may also be disparities in referral rates. In a large U.S. hospital system, over half of children who screened positive for a developmental problem were not referred to early intervention services, and referral rates differed by child sex, age, and race.⁵⁴ Pediatric providers typically do not have access to detailed information about the quality of the child’s home environment, parenting practices, or even the parents’ demographic and social history. Instead, when universal screening is not feasible, pediatric providers could quickly link the family’s home zip code to neighborhood quality (a web-based search tool is available here: https://data.diversitydatakids.org/dataset/coi20_zipcodes-child-opportunity-index-2-0-zip-code-data) as a proxy for family resources, potentially improving decisions regarding who to prioritize for screening, as well as who to refer to early intervention.

Data Sharing Statement:

Deidentified individual participant data (including data dictionaries) will be made available upon reasonable request following approval of a data use agreement.