Caregiver-Perceived Neighborhood Safety and Pediatric Asthma Severity: 2017–2018 National Survey of Children’s Health

Abstract

Objective:

To examine the association between caregiver-perceived neighborhood safety and pediatric asthma severity using a cross-sectional, nationally representative sample.

Study Design:

Using data from the 2017–2018 National Survey of Children’s Health, children aged 6–17 years with primary caregiver report of a current asthma diagnosis were included (unweighted N = 3209; weighted N = 3,909,178). Perceived neighborhood safety, asthma severity (mild versus moderate/severe), demographic, household, and health/behavioral covariate data were collected from primary caregiver report. Poisson regression with robust error variance was used to estimate the association between perceived neighborhood safety and caregiver-reported pediatric asthma severity.

Results:

Approximately one-third of children studied had moderate/severe asthma. Forty-two percent of children with mild asthma and 52% of children with moderate/severe asthma identified as Hispanic or non-Hispanic Black. Nearly 20% of children with mild asthma and 40% of children with moderate/severe asthma were from families living below the federal poverty level. Children living in neighborhoods perceived by their caregiver to be unsafe had higher prevalence of moderate/severe asthma compared to those in the safest neighborhoods (adjusted prevalence ratio: 1.34; 95% confidence interval: 1.04–1.74). This association was found to be independent of race/ethnicity, household federal poverty level, household smoking, and child’s physical activity level after adjusting for covariates.

Conclusions:

Children living in neighborhoods perceived by their caregiver to be unsafe have higher prevalence of moderate or severe asthma. Further investigation of geographic context and neighborhood characteristics that influence childhood asthma severity may inform public health strategies to reduce asthma burden and improve disease outcomes.

Introduction

Pediatric asthma is one of the most prevalent chronic conditions of childhood, affecting approximately 1 in 12 children in the United States as of 2016 and accounting for over half a million emergency department visits annually among children aged 17 years and younger.^1–3 This condition is characterized by chronic airway inflammation that may lead to reversible airway obstruction.⁴ Children with moderate and severe asthma experience more persistent symptoms and have more frequent exacerbations, increasing the risk of hospitalization in this population.^{5; 6}

Asthma is also known to be more common among certain populations of children, including those from low-income and minority populations living in urban settings.⁷ Physical environment, psychosocial stressors, and limited access to healthcare within these settings contribute not only to barriers to pediatric asthma care but also increased asthma severity and repeated exacerbations.^7–11 The chronic psychosocial stress inherent to these settings has also been associated with dysregulated immune cell activity and glucocorticoid receptor insensitivity, further predisposing the airways of inhabitants to inflammation, even more so in the setting of exposure to allergens or irritants.^{12; 13}

Recent efforts have examined the impacts of these neighborhood-level factors and how they are associated with pediatric asthma prevalence and its associated morbidity.^{11; 14; 15} Living in neighborhoods perceived by caregivers to be less safe is associated with increased lifetime prevalence of pediatric asthma, as previously reported in a nationally representative sample.¹⁴ Prior work has also described that neighborhoods perceived to be unsafe are associated with poorer asthma control regardless of severity.¹⁶ It is important to note that asthma control is not equivalent to asthma severity, as children with severe asthma may be well-controlled or poorly-controlled.¹⁷ Thus, despite growing evidence in the realm of pediatric asthma and neighborhood-level factors, the extent to which caregiver-perceived neighborhood safety is associated with asthma severity remains unclear.

The objective of this study, therefore, was to examine the association between caregiver-perceived neighborhood safety and childhood asthma severity as reported by caregivers in a large, nationally conducted cross-sectional survey of children in the US. We used methods similar to a prior study that examined the association of perceived neighborhood safety and lifetime prevalence of pediatric asthma in a nationally representative sample.¹⁴ For the present study, we hypothesized that living in neighborhoods perceived to be less safe would be associated with higher prevalence of reporting severe asthma.

Materials and Methods

Population & Data

Data were obtained from the combined 2017–2018 National Survey of Children’s Health (NSCH), a publicly available data set collected by the United States Census Bureau during August 2017-February 2018 and June 2018-January 2019.¹⁸ Data were collected via mail- and web-based survey in English and Spanish with funding provided by the Health Resources and Services Administration Maternal and Child Health Bureau.^{18; 19} A total of 52,129 households with at least one child aged 0–17 years responded from all states and the District of Columbia. For this survey, one child was randomly selected from each randomly sampled qualifying household. The weighted overall response rate was 37.4% for the 2017 NSCH and 43.1% for the 2018 NSCH.¹⁹ Further details concerning the NSCH, methods on data collection, and survey instruments can be found at the following website: https://www.childhealthdata.org/learn-about-the-nsch. The University of Massachusetts Institutional Review Board considered this study exempt because publicly available data were used.

Inclusion and exclusion criteria for our study sample are shown in Figure 1. Of the total surveyed 52,129 households with at least one child aged 0–17 years, we first excluded children aged 0–5 years because covariates such as physical activity were not evaluated (n = 14,737). Among households with a child aged 6–17 years, participants were further excluded due to the lack of a current asthma diagnosis (n = 33,878). Among children aged 6–17 years, 3514 children (9.4%) had a caregiver-reported current asthma diagnosis and met age criteria. Among those, 9 (0.3%) were excluded due to missing asthma severity data and 71 (2%) lacked information on perceived neighborhood safety. Of the remaining sample, 225 (7%) were excluded due to missing covariate data. The final analytic sample consisted of 3209 children (weighted N = 3,909,178).

Figure 1.

Flow diagram. This figure depicts the process of arriving at the final unweighted analytic sample size (weighted N = 3,909,178).

Outcome

Asthma severity was assessed using primary caregiver report. In the survey, caregivers were first asked: “Has a doctor or other health care provider EVER told you that this child has asthma?” For caregivers who responded yes, a follow-up question was asked: “If yes, does this child CURRENTLY have the condition?” Caregivers who responded yes were then asked: “If yes, is it: Mild? Moderate? Severe?” In the NSCH data set, primary caregiver-reported severity was dichotomized into mild versus moderate/severe asthma.

Exposure

Perceived neighborhood safety was also measured through primary caregiver report. Caregivers were asked to what extent they agreed with the following statement: “This child is safe in our neighborhood.” Response options included: “definitely agree,” “somewhat agree,” “somewhat disagree,” and “definitely disagree.” In the NSCH data set, the “somewhat disagree” and “definitely disagree” responses were combined. Therefore, the three categories used for the analysis were strongly agree, somewhat agree, and somewhat/strongly disagree.

Covariates

The following covariates were selected a priori from the 2017–2018 surveys: child’s age, sex, race/ethnicity, health insurance type, household poverty level, highest household education, household smoking status, child having witnessed or heard domestic violence, child’s BMI, caregiver-reported diagnosis of allergies, and child’s physical activity level. These covariates were selected primarily based on a previous study that utilized the 2003–2004 NSCH to examine the association between perceived neighborhood safety and lifetime prevalence of pediatric asthma.¹⁴ Health insurance type was added as a covariate based on a prior examination showing an association between adverse childhood experiences and pediatric asthma severity.²⁰ An additional variable describing whether a “child had witnessed or heard domestic violence” was included as a covariate based on a previously established associations between this variable and pediatric asthma severity and morbidity.^{10; 20}

Child’s age was dichotomized into 6–11 years and 12–17 years. These groups were determined based on the subgroup variable for age provided by the 2017–2018 NSCH data set. Sex was dichotomized into male or female. Race/ethnicity was defined as Hispanic, non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, and non-Hispanic Other, also defined by the NSCH data set.

Health insurance type was classified as public (“Medicaid, Medical Assistance, or any kind of government assistance plan for those with low incomes or a disability”), private (“Insurance through a current or former employer or union, Insurance purchased directly from an insurance company, TRICARE or other military health care, coverage through the Affordable Care Act, or other private insurance), both public and private, or currently uninsured (included “children identified as only having Indian Health Services or a religious health share”).

Child’s household federal poverty level (FPL) was determined from questions regarding family income in the Household Information section of the NSCH. Each year approximately 15% of surveys do not provide information required to calculate FPL, which could limit sample size and bias estimates.²¹ Therefore, missing values for this covariate were imputed by the United States Census Bureau, and these values obtained by single imputation were included in our analysis. The imputation process has previously been described in detail.²¹

Highest household education was collected as “less than high school,” “high school or GED,” “some college or technical school,” and “college degree or higher.” The latter two responses were combined into one category: greater than high school.

Household smoking status was collected via the following survey question: “Does anyone living in your household use cigarettes, cigars, or pipe tobacco?”

Child having witnessed or heard domestic violence was a binary yes/no variable assessed with caregivers answering the following question: “To the best of your knowledge, has this child ever experienced the following: saw or heard parents or adults slap, hit, kick, punch one another in the home?”

Child’s BMI was only collected via caregiver report of height and weight for children aged 10 years and older since parents are more likely to overestimate height and underestimate weight for children younger than 10 years old.²² In this study, child’s BMI was classified as underweight (<5^th percentile), healthy weight (5^th-84^th percentile), and overweight or obese (>85^th percentile). A separate category named “Child aged 6–9 years” was created within this variable to allow the group of children aged 6–9 years for whom BMI was not recorded to be included in the final model. To address the potential bias introduced by the lack of BMI data for children aged 6–9 years, a sensitivity analysis was conducted on children aged 11–17 years, comparing those with and without BMI data. This sensitivity analysis and its results are discussed in later sections of this manuscript.

Child’s allergy status was assessed by a primary caregiver report of current presence of allergies “including food, drug and insect or other.”

Child’s physical activity level was measured by caregiver report of the number of days in the past week that the child had exercised, played a sport, or participated in physical activity for at least 60 minutes.

Primary Analytic Approach

Descriptive statistics and Pearson’s chi-square tests were first applied to describe sample characteristics among children with mild asthma and moderate/severe asthma. Given that the prevalence of moderate/severe asthma was approximately 35%, logistic regression would have overestimated effect estimates.²³ Although log-binomial regression would be a more optimal approach to estimate prevalence ratios,²³ the log-binomial models did not converge. Therefore, we used Poisson regression with robust error variance as an alternative approach to estimate crude and adjusted prevalence ratios (aPR) with 95% confidence intervals (CI) to examine the association of perceived neighborhood safety and caregiver-reported asthma severity. Multivariable adjusted models included a priori determined demographic, household, and health/behavioral covariates, incorporated in a progressive manner. Analyses were conducted using STATA Version 16.1 with the sampling plan and weights provided with the data set to account for the complex sampling design of NSCH and to generate nationwide population estimates.^{24; 25}

Sensitivity Analysis

Given the lack of BMI data for younger children, we used a BMI category “Child aged 6–9 years.” Despite that this allowed a wider age range of children to be included in the final model, this BMI category could be collinear with the age variable category “6–11 years.” Therefore, a sensitivity analysis was conducted on children aged 11–17 years, comparing those with and without BMI data, to provide further insight into the direction and amount of bias associated with not having BMI data for children under 10 years old. Poisson regression with robust error variance was used to estimate crude and adjusted prevalence ratios with 95% confidence intervals.

Results

Sample Characteristics

Characteristics of children aged 6–17 years with caregiver-reported mild or moderate/severe asthma are displayed in Table 1. Overall, 35% had moderate or severe asthma. Hispanic and non-Hispanic Black children accounted for approximately 40% of children with mild asthma and 50% of children with moderate/severe asthma. Approximately 20% of children with mild asthma and 40% of children with moderate/severe asthma were from households living below the FPL. Approximately 50% of children with mild asthma and 75% of children with moderate/severe asthma currently had allergies.

Table 1.

Characteristics of children (6–17 years) with current asthma, 2017–2018 NSCH.

Variable	Mild Asthma	Moderate/Severe Asthma	p-value
	%a (Weighted N = 2,548,784)	%a (Weighted N = 1,360,394)
Population size	65.2	34.8
Child lives in a safe neighborhood
Definitely agree	61.8	57.1	< 0.001
Somewhat agree	33.6	28.4
Somewhat or strongly disagree	4.5	14.5
Demographic Characteristics
Age
6–11 years	46.6	47.1	0.900
12–17 years	53.4	52.9
Female	45.4	47.1	0.683
Race/ethnicity
White, non-Hispanic	47.5	41.0	0.037
Hispanic	21.1	22.3
Black, non-Hispanic	20.4	29.5
Asian, non-Hispanic	4.0	1.5
Other, non-Hispanic	7.0	5.7
Health insurance type
Public only	28.1	50.1	< 0.001
Private only	60.2	38.2
Public and private	7.2	7.3
Currently uninsured	4.5	4.5
Household Characteristics
Household FPL
0–99%	20.8	38.0	< 0.001
100–199%	21.4	20.3
200–399%	27.7	22.8
400%	30.2	19.0
Highest household education
Less than high school	8.9	9.9	0.064
High school degree or equivalent	17.2	26.0
Greater than high school	73.9	64.2
Someone in the child’s household smokes
No one	81.4	79.4	0.490
Someone	18.6	20.6
Child saw or heard domestic violence b
Yes	9.2	11.8	0.284
No	90.8	88.2
Health/Behavioral Characteristics
Child’s BMI
Underweight	3.9	5.6	0.597
Healthy weight	41.2	39.3
Overweight or obese	23.1	25.2
Child aged 6–9 yearsc	31.9	29.9
Child’s allergy status
Does not have condition	42.8	19.7	< 0.001
Was ever told, but does not currently have	6.4	3.7
Currently has condition	50.8	76.6
Child’s physical activity (at least 60 min/day)
0 days	8.9	12.6	0.037
1–3 days	37.7	36.6
4–6 days	30.5	21.8
Everyday	22.9	28.9

^aWeighted percent

^bSaw or heard parents or adults slap, hit, kick, punch one another in the home

^cBMI was not collected for children younger than 10 years old

Univariable and Multivariable Results

Living in neighborhoods perceived to be less safe was associated with increased children’s asthma severity. In the unadjusted model (Table 2, Model 1), children whose caregivers perceived an unsafe neighborhood had higher prevalence of having moderate/severe asthma compared to those whose caregivers strongly agreed that their child lived in a safe neighborhood (prevalence ratio: 1.91; 95% CI: 1.43–2.53).

Table 2.

Associations of perceived neighborhood safety and moderate/severe asthma severity among children aged 6–17 years, 2017–2018 NSCH.

Variable	Model 1a	Model 2b	Model 3c	Model 4d
Child lives in a safe neighborhood
Strongly agree (reference)	-	-	-	-
Somewhat agree	0.94 (0.75–1.18)	0.89 (0.72–1.11)	0.87 (0.70–1.09)	0.86 (0.70–1.07)
Somewhat or strongly disagree	1.91 (1.43–2.53)	1.62 (1.23–2.14)	1.52 (1.17–1.96)	1.34 (1.04–1.74)

^aModel 1: unadjusted

^bModel 2: adjusted for age, sex, race/ethnicity, and health insurance type

^cModel 3: adjusted for age, sex, race/ethnicity, health insurance type, household FPL, highest household education, household smoking, child having witnessed or heard domestic violence

^dModel 4: adjusted for age, sex, race/ethnicity, health insurance type, household FPL, highest household education, household smoking, child having witnessed or heard domestic violence, child’s BMI (not collected for children younger than 10 years old), child’s allergy status, and child’s physical activity level

Demographic covariates included child’s age, sex, race/ethnicity, and health insurance type. When adjusting for these covariates (Table 2, Model 2), the positive association between living in a perceived unsafe neighborhood and increased asthma severity was mildly attenuated but still significant (aPR: 1.62; 95% CI: 1.23–2.14; somewhat/strongly disagree vs. strongly agree). Additionally, after adjusting for household characteristics (Table 2, Model 3), including FPL, highest household education, household smoking, and child having witnessed or heard domestic violence, the aPR decreased to 1.52 (95% CI: 1.17–1.96; somewhat/strongly disagree vs. strongly agree). Our final model which further adjusted for health and behavioral covariates (Table 2, Model 4), including child’s BMI, allergy status, and physical activity level, resulted in an aPR of 1.34 (95% CI: 1.04–1.74; somewhat/strongly disagree vs. strongly agree).

Sensitivity Analysis Results

A sensitivity analysis was completed to compare the associations of perceived neighborhood safety and asthma severity among children aged 11–17 years for whom BMI data were and were not present (Table 3). The lack of BMI data was associated with reduced effect estimates in the unadjusted model and the model that incorporated demographic and household covariates.

Table 3.

Sensitivity analysis comparing associations of perceived neighborhood safety and moderate/severe asthma severity among children aged 11–17 years with and without BMI data, 2017–2018 NSCH.

BMI Data Present	Variable	Model 1a	Model 2b	Model 3c	Model 4d
Yes	Child lives in a safe neighborhood
Weighted N = 2,081,260	Strongly agree (reference)	-	-	-	-
	Somewhat agree	0.90 (0.67–1.21)	0.80 (0.61–1.06)	0.80 (0.61–1.05)	0.82 (0.63–1.06)
	Somewhat or strongly disagree	2.11 (1.49–2.99)	1.61 (1.14–2.26)	1.43 (1.07–1.92)	1.29 (0.96–1.74)
No	Child lives in a safe neighborhood
Weighted N = 118,005	Strongly agree (reference)	-	-	-	-
	Somewhat agree	0.60 (0.20–1.82)	0.46 (0.18–1.21)	0.30 (0.12–0.74)	0.01 (0.00–0.11)
	Somewhat or strongly disagree	1.68 (1.00–2.82)	1.67 (0.66–4.19)	0.55 (0.10–3.02)	-e

^aModel 1: unadjusted

^bModel 2: adjusted for age, sex, race/ethnicity, and health insurance type

^cModel 3: adjusted for age, sex, race/ethnicity, health insurance type, household FPL, highest household education, household smoking, child having witnessed or heard domestic violence

^dModel 4: adjusted for age, sex, race/ethnicity, health insurance type, household FPL, highest household education, household smoking, child having witnessed or heard domestic violence, child’s BMI (not collected for children younger than 10 years old), child’s allergy status, and child’s physical activity level

^eThe final model did not achieve converge due to low sample size

Discussion

Using data from the 2017–2018 NSCH, we found that approximately one-third of children aged 6–17 years reported moderate or severe asthma. One in five children who had mild asthma and two in five children who had moderate/severe asthma were from households living below the FPL. We observed that caregiver-perceived low neighborhood safety was associated with increased asthma severity in children aged 6–17 years. Utilizing a large, diverse, recent, and nationally representative data set, we show an important association between caregiver-reported neighborhood safety and childhood asthma severity.

While many studies have examined the role of social determinants of health in asthma prevalence,^{14; 26; 27} only four have examined the relationship between neighborhood safety and pediatric asthma control.^{15; 16; 28; 29} Our findings align with a cross-sectional study of 219 children in one urban center showing that children whose primary caregivers perceived their neighborhood to be unsafe were more likely to have poorly-controlled asthma.¹⁵ Though asthma control is not equivalent to asthma severity, children with more severe asthma are more likely to have poorer asthma control.³⁰

In our study, the association of negatively perceived neighborhood safety and increased asthma severity was found to be independent of several key covariates, including race/ethnicity, household federal poverty level, household smoking, and child’s physical activity level. The proposed mechanisms for increased asthma severity among children are myriad and have included inadequate investment in neighborhoods of low socioeconomic status, leading to: higher pollutant exposure, higher tobacco exposures, higher prevalence of mice and cockroaches which are known to exacerbate asthma, lack of access to recreational facilities such as playgrounds or green spaces, lack of access to healthy food, inadequate access to medical care for management of chronic conditions, and exposure to violence in communities.^{26; 28; 31; 32} Moreover, psychosocial stress, which is associated with poorer asthma outcomes, is often exacerbated by lack of social supports in the community.^{10; 14; 26} One review highlighted the multiple levels at which psychosocial stress can act to influence asthma morbidity, including the individual, family, and community levels.¹⁰ Ultimately, this chronic stress may play a role in immune dysregulation, leading to persistent inflammatory changes in the airway and impaired glucocorticoid function characteristic of asthma.^{12; 13}

A prior study that utilized similar methods to this work to examine the 2003–2004 NSCH data set demonstrated that living in neighborhoods perceived to be unsafe per parent report was associated with increased lifetime prevalence of pediatric asthma.¹⁴ This study described a statistically significant dose-response relationship, with children living in “usually safe” neighborhoods more likely to have asthma than children in “always safe” neighborhoods. A stronger association with asthma was found among children who lived in “sometimes or never safe” neighborhoods. Of note, our analysis does not indicate that a similar dose-response relationship exists for asthma severity since children of caregivers who “somewhat agreed” that they live in a safe neighborhood were not more likely to have moderate/severe asthma than children of caregivers who “strongly agreed.” It is possible this disparity arises from families responding more accurately to questions about the presence of an asthma diagnosis compared to asthma severity, which may be more difficult to assess objectively. Parents are more likely to underestimate their child’s asthma severity and overestimate asthma control.³³

This study describes an important relationship between neighborhood safety and childhood asthma severity, highlighting the multifactorial role of social determinants of health and the need to address them to mitigate inequities in pediatric asthma morbidity. We recommend future investigation includes examinations of the specific features of a neighborhood that cause it to be perceived as unsafe and their relationships to pediatric asthma. Proximity to environmental pollutants, such as highways or landfills, lack of access to green spaces, and proximity to violence may be associated with increased asthma severity. Qualitative studies may provide more evidence regarding neighborhood context, such as how historical segregation has contributed to lower quality of life, decreased access to health care, and lack of neighborhood safety.

Strengths & Limitations

A major strength of this study lies in its use of a recent, nationally representative sample, which enhances the generalizability of the results. To the best of our knowledge, this is the first study to explore the relationship between perceived neighborhood safety and asthma severity using population-level data. Approximately 93% of eligible cases were included in the final analytic sample, indicating a low proportion of missing data.

The following limitations should be considered. The NSCH is cross-sectional and is not an asthma-focused questionnaire. Thus, it has limited data regarding asthma severity and biological indicators. This study is based on a caregiver-reported diagnosis of current asthma with no clinical validation or review of medical records. Previous studies have found high levels of agreement between parent and child reports of asthma; however, it has also been found that parents often underestimate the severity of their child’s asthma.^{33; 34} In the future, it may be worth considering incorporation of the Asthma Control Test or other validated survey tools used to assess pediatric asthma control.³⁵ Other options for measuring asthma severity and control include pulmonary function testing or assessing rescue medication use.³⁶

These data also rely solely upon caregiver perception of neighborhood safety and do not provide an objective neighborhood safety measure. Nonetheless, these data deliver meaningful insight about how residents internalize the spatial and social components of their communities. A measure quantifying physical and economic aspects of a neighborhood without community context may not fully capture the element of perception that is integral to understanding how stress resulting from perceived threat impacts the physiology relevant to asthma. Despite this potential benefit, the NSCH does not include the child’s perception and for this reason, information may be lacking in its inability to provide multi- and cross-generational experiences of neighborhood. Future studies may consider incorporating data regarding other adverse childhood experiences in addition to child having witnessed or heard domestic violence. The questionnaire is included in the NSCH, but it is important to note that the data are provided by caregivers and not the children themselves.

The covariates were also selected a priori, and we may have overlooked relevant variables. For example, this study does not incorporate data regarding access to health care or frequency of utilization of health care, which may impact the progression and severity of asthma. Other examples of possible covariates not included in the NSCH but noted to play a role in how psychosocial stressors affect asthma morbidity include genetic/epigenetic factors and maternal/paternal stress.¹⁰

For reasons mentioned previously, BMI data were only available for children aged 6–9 years; however, by creating a separate category for these children within the ‘Child’s BMI’ variable, we were able to apply all models to the final analytic sample. A sensitivity analysis was performed to demonstrate possible bias due to the lack of BMI data for children aged 6–9 years. However, the sample size was small enough such that the final model did not converge, producing an unreliable effect estimate for the final model incorporating children aged 11–17 years who did not have recorded BMI data.

Finally, our study included only 5.3% (N = 3,909,178) of the total nationally weighted population (N = 73,428,760). To perform the analysis, we used the sampling weights provided by the NSCH data set to account for the complex survey design. It is important to interpret the results in this context, understanding that utilizing the provided sample weights may have impacted the estimated prevalence ratios and therefore the generalizability of the results.

Conclusions

Utilizing nationally representative data, we found that caregiver perception of low neighborhood safety is associated with higher childhood asthma severity. Future investigations should consider integrating an objective measure of neighborhood safety and should attempt to incorporate the child’s perspective. Public health interventions targeting neighborhood characteristics may provide a means to improve health outcomes for children with asthma, particularly those with moderate or severe disease.

Abbreviations and Acronyms

NSCH

National Survey of Children’s Health

BMI

body mass index

FPL

federal poverty level

aPR

adjusted prevalence ratio

CI

confidence interval

Data Availability Statement:

The data that support the findings of this study are openly available in the Data Resource Center for Child and Adolescent Health at https://www.childhealthdata.org/dataset, reference number 18.