The longitudinal association between neighborhood quality and cardiovascular risk factors among youth receiving obesity treatment

Abstract

Background:

Neighborhood factors are associated with cardiovascular health in adults, but these relationships are under-explored in youth.

Objectives:

To characterize the associations between neighborhood factors and child and adolescent health among youth with obesity.

Methods:

Data were drawn from patient health records at a pediatric weight management clinic (n=2838) and the Child Opportunity Index (COI). Exposures were area-level neighborhood factors (commute duration, walkability, greenspace, industrial pollutants). Outcomes included BMI relative to the 95^th percentile (BMIp95) and blood pressure (continuous variables). Longitudinal models examined associations between COI indicators and outcomes.

Results:

Shorter commute duration (β=−4.31, 95% CI: −5.92, −2.71) and greater walkability (β=−4.40, 95% CI: −5.98, −2.82) were negatively associated with BMIp95. Increased greenspace availability was positively associated with BMIp95 (β=1.93, 95% CI: 0.19, 3.67). None of the COI indicators were associated with cardiovascular outcomes in the full sample. Analyses stratified by sex and race/ethnicity showed similar patterns for BMIp95. For commute duration, there was a negative association with blood pressure for female, non-Hispanic White, and other race/ethnicity youth.

Conclusions:

Neighborhood factors should be considered as contextual factors when treating youth with obesity. Additional research is needed to understand the relationship between neighborhood factors and cardiovascular outcomes.

Introduction

The childhood obesity epidemic continues to be a global health crisis, with 124 million school-aged children having obesity in 2016.¹ In the United States (US), 19.3% of children and adolescents aged 2–19 years had obesity in 2019, which has been further exacerbated by the COVID-19 pandemic.² Obesity in childhood and adolescence is associated with increased cardiometabolic risk and associated morbidity and mortality later in life, indicating a need to address factors that contribute to youth obesity.^3–5

Behavioral strategies such as lifestyle modification to improve diet or physical activity are important for children with obesity, but there is a high degree of heterogeneity in response to individual-level treatment.⁶ From a socioecological approach, individuals’ weight status is the result of the complex interaction of individual, social and environmental factors.⁷ Therefore, it is necessary to necessary to consider contextual factors that may factilitate or serve as barriers to a healthy weight and wellness.

Neighborhood factors including the built environment are key contributors to health, driving behaviors and health disparities predictive of cardiovascular risk in children and adolescents. For instance, lower perceived neighborhood safety, higher crime rates, and less access to supermarkets have all been associated with higher body mass index (BMI) in children and adolescents.^8–10 Outdoor play and physical activity are associated with neighborhood greenness, parks, and playgrounds in early childhood, but these spaces are often distributed unequally and lacking in communites that serve low wealth populations and racial and ethnic minorities.^11–13 Additionally, evidence suggests that low neighborhood walkability (i.e., lower street connectivity, increased density of population and dwellings) is associated with lower physical activity and higher weight status among children.¹⁴ Furthermore, increased neighborhood walkability has been associated with improvements in youth cardiovascular outcomes.¹⁵ Commute duration in adults is also associated with decreased levels of health-promoting activities including food preparation and physical activity, which may influence parenting behaviors that could affect child health behaviors.^16,17 Exposure to pollutants, such as fine particular matter and ozone gas, are also linked to adverse health effects and cardiovascular disease in the adult population; however much less is known about exposure to pollutants on youth cardiovascular health.¹⁸

Compared to the adult literature, the relationships between neighborhood factors, obesity and cardiovascular risk remain understudied in child and adolescent populations. Even less is known about these associations in youth with obesity. Therefore, the aim of the current study was to examine the association between area level neighborhood factors (i.e., greenspace, walkability, commute duration, and industrial pollutants), adiposity, and blood pressure among a population of youth receiving obesity treatment. Additionally, findings were examined across sex and racial/ethnic subgroups. Findings from this work can promote better understanding of the relationship between neighborhood context and child health outcomes, to ultimately inform clinical and public health initiatives targeting health disparities reduction.

Methods

Population

This analysis used patient electronic health record data from the Duke Children’s Healthy Lifestyles pediatric weight management clinic in Durham, North Carolina, US. Data were drawn from July 2013 to February 2020. Healthy Lifestyles is a multidisciplinary specialty care clinic that treats children and adolescents with obesity. Patients with obesity and their families see a medical provider as well as a registered dietitian, physical therapist, and licensed counselor, to address diet, activity, and medical needs to improve health and treat co-morbidities. The population for this analysis included Healthy Lifestyles patients between the ages of 6 and 18 years with a BMI at or above the 95^th percentile for age and sex. Patients were included if they had at an initial visit (baseline) and at least one subsequent visit. Children were excluded if they lived outside of North Carolina (n=24) or had multiple missing values for the outcomes of interest (n=46).

Exposures

All exposure variables were drawn from the Child Opportunity Index (COI). The COI is a publicly available dataset created by diversitydatakids.org and managed by the Heller School for Social Policy and Management at Brandeis University.¹⁹ It provides a measure of neighborhood resources and conditions that impact children for all census tracts across the US. The overall COI includes 29 indicator variables. A total COI z-score is available as well as z-scores for each indicator variable. As part of the Duke Health electronic health record, there is an automated geocoding process that provides census tract information for the patients’ home address, which was linked to the COI dataset. We used the COI 2.0 from 2015 based on the time range of patient data (2013–2020).

Four individual COI indicators were included in analyses: commute duration, walkability, greenspace, and industrial pollutants. Commute duration was the percent of workers commuting more than an hour, with data coming from the American Community Survey and averaged across 2012–2017. Commute duration was included as a measure of transportation vulnerability, as prior literature has shown that commute time may be related to vulnerability in child health and development.²⁰ Walkability was measured using the 2010–2012 Environmental Protection Agency (EPA) Walkability Index that takes street intersection density, distance to transit stops, employment types, and housing into account to predict walking trips. Greenspace availability was defined as the inverse of the percent of each tract covered in impervious surfaces (e.g. rooftops, parking lots), from 2011 satellite data from the National Land Cover Database. Exposure to industrial pollutants in air, water, or soil data came from the 2015 EPA Risk-Screening Environmental Indicators.²⁰ For each indicator, scores range from 0–100 and were coded such that higher scores reflect more ideal neighborhood characteristics (i.e., shorter commute, greater walkability, increased greenspace availability, and fewer industrial pollutants).

Outcomes

BMI relative to the 95^th percentile.

At each Healthy Lifestyles clinic visit, patient height and weights are measured by a nurse. Along with age at the time of the visit, this was used to calculate youth BMI relative to the 95^th percentile (BMIp95) based on the Centers for Disease Control and Prevention growth charts.²¹ BMIp95 is a continuous measure that expresses BMI as a percentage of the 95^th percentile. BMIp95 was used instead of other BMI measures (i.e., BMI or BMI z-scores) based on evidence that BMIp95 is a better indicator of adiposity and captures more variability among children with obesity.²²

Blood pressure.

Blood pressure is also measured at each visit by a nurse using a calibrated auscultatory sphygmomanometer with an appropriately sized cuff using standard methods. Due to expected changes in blood pressure based on age and sex, blood pressure values were transformed to age, height, and sex-specific percentiles according to American Academy of Pediatrics guidelines using a SAS macro.²³ The distribution of systolic blood pressure percentiles had a strong negative skew, so diastolic blood pressure percentile (normally distributed )was used as the primary continuous blood pressure-related outcome given that both diastolic and systolic blood pressure contribute to cardiovascular risk.²⁴

Covariates

Race/ethnicity (non-Hispanic Black, non-Hispanic White, Hispanic, and Other), sex (male or female), health insurance status (public, private, other, missing) and age at each encounter were drawn from patient electronic health records. “Other” was a response option in the electronic health record and could not be further defined. Race/ethnicity were included as covariates because of disparities in obesity prevalence among these groups, variation in treatment outcomes by race/ethnicity, and because race/ethnicity is one of the primary predictors of patient retention in pediatric weight management programs.^25–27

Statistical Analysis

Descriptive statistics were computed to describe characteristics of the sample. Means and standard deviations were calculated for continuous variables, and frequencies and percentages were calculated for categorical variables. Three-level generalized linear mixed models were fit to investigate the longitudinal association between COI indicators and youth outcomes across all participants and stratified by sex and race/ethnicity. To account for repeated measures over time, random effects were included for the repeated measures with individual patient observations nested within patients, which were then nested within census tracts to account for geographic similarities. All models were adjusted for youth age, sex, race/ethnicity, health insurance status, and time between visits. Due to the irregularity of electronic health record data, a spatial covariance structure was used to account for varying time (days) between visits to the clinic.²⁸ For all models, beta coefficients represent the average change in BMIp95 or diastolic blood pressure percentile across the time period when the COI indicator score is increased by one unit. A significance level of at least 0.05 was used in all models. All analyses were conducted using SAS 9.3 (SAS Institute, Inc., Cary, NC).

Results

Characteristics of the patient population (n=2838) are shown in Table 1. The mean number of observations per patient was 5.16 (±5.29) for BMIp95 and 4.79 (±4.72) for blood pressure. Mean child age was 11.08 (± 3.22) years. Slightly more than half of patients were female (55.0%) and the majority were either non-Hispanic Black (33.7%) or Hispanic (27.1%) and receiving public insurance (69.1%). Mean child BMIp95 was 127.7 (SD 23.1). There were 429 unique census tracts captured in the population, which is depicted in Figure 1.

Table 1.

Descriptive characteristics of the analytic sample (n=2838)

Age, years (mean (SD))	11.1 (3.2)
Female (%)	55.0
Race/ethnicity (%)
Non-Hispanic White	14.5
Non-Hispanic Black	33.7
Hispanic	27.1
Other a	24.8
Health insurance status (%)
Public	69.1
Private	27.8
Other	0.6
Missing	2.6
BMIp95 (mean (SD))	127.7 (23.1)
Neighborhood Indicators z scores (mean (SD))
Commute duration	0.38 (0.52)
Greenspace	0.50 (0.50)
Walkability	0.22 (1.03)
Industrial Pollutants	0.21 (0.34)

Abbreviations: standard deviation (SD), body mass index relative to the 95^th percentile for age and sex (BMIp95)

^(a)Other is not specified

Figure 1.

Distribution of census tracts (n=429) and number of patients per each census tract

Association between BMIp95 and COI indicators

Adjusted estimates for the associations between BMIp95 and COI indicators are shown in Table 2. Shorter commute duration (β = −4.31, 95% CI: −5.92, −2.71) and greaterwalkability (β = −4.40, 95% CI: −5.98, −2.82) were negatively associated with BMIp95, while greenspace availability (β = 1.93, 95% CI: 0.19, 3.67) was positively associated with BMIp95. . There were no significant associations between BMIp95 and industrial pollutants.

Table 2.

Adjusted estimates for the association between BMIp95 and COI indicator z-scores for the full cohort and stratified by sex and race/ethnicity ^a,b

	Commute Duration		Walkability		Greenspace		Industrial Pollutants
	Estimate	95 % CI	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI
Full Cohort	−4.31 ***	−5.92, −2.71	−4.40	−5.98, −2.82	1.93 ***	0.19, 3.67	−0.19	−2.65, 2.28
Sex
Female	−4.54***	−6.72, −2.36	−2.92 **	−4.66, −1.18	0.72	−1.66, 3.1	0.51	−2.82, 3.84
Male	−4.15 **	−6.54, −1.76	−4.67 ***	−6.60, −2.73	3.18	0.61, 5.75	−0.98	−4.65, 2.69
Race/ethnicity
Non-Hispanic Black	−5.69 **	−8.76, −2.63	−1.82	−3.68, 0.05	0.84	−2.37, 4.05	−1.71	−6.42, 3.01
Non-Hispanic White	−11.42 ***	−16.02, −6.82	−3.98 **	−6.66, −1.30	6.14	0.14, 12.14	4.41	−2.36, 11.17
Hispanic	−1.14	−3.68, 1.39	−1.92	−3.54, −0.30	0.18	−2.46, 2.81	−4.43	−9.36, 0.49
Other race/ethnicity	−3.04	−6.28, 0.20	−3.72 *	−6.29, −1.14	2.00	−1.57, 5.56	0.46	−3.86, 4.77

Abbreviations: body mass index relative to the 95^th percentile for age and sex (BMIp95), child opportunity index (COI), confidence interval (CI)

^(a)Boldface indicated statistical significance, (p<0.05)

^*indicates p < .01

^**indicates p < .001

^***indicates p <.0001

^(b)Models adjusted for insurance type (private, public, other, missing), age, sex, (full and race/ethnicity subgroups), race/ethnicity (full and sex subgroups)

^(c)Other is not specified; “other” is an option that can be selected within the electronic health record

Sex-stratified adjusted models (Table 2) showed similar significant negative associations for commute duration and BMIp95 for both female (β = −4.54, 95% CI: −6.72, −2.36) and male (β = −4.15, 95% CI: −6.54, −1.76) youth. Walkability remained significantly negatively associated with BMIp95 for both female (β = −2.92, 95% CI: −4.66, −1.18) and male youth (β = −4.67, 95% CI: −6.60, −2.73). The positive association between greenspace availability and BMIp95 remained significant only for male youth (β = 3.18, 95% CI: 0.61, 5.75).

In analyses stratified by race/ethnicity (Table 2), commute duration remained significantly negatively associated with BMIp95 in non-Hispanic Black (β = −5.69, 95% CI: −8.76, −2.63) and non-Hispanic White youth (β = −11.42, 95% CI: −16.02, −6.82).Walkability was significantly negatively associated with BMIp95 for non-Hispanic White (β = −3.98, 95% CI: −6.66, −1.30), Hispanic (β = −1.92, 95% CI: −3.54, −0.30), and other race/ethnicity youth (β = −3.72, 95% CI: −6.29, −1.14). Greenspace availability was positively associated with BMIp95 only among non-Hispanic White youth (β = 6.14, 95% CI: 0.14, 12.14).

Association between diastolic blood pressure and COI indicators

Adjusted estimates for the associations between diastolic blood pressure and COI indicators are shown in Table 3. None of the COI indicators was associated with diastolic blood pressure for the full cohorts. Stratified analyses revealed that commute duration was negatively associated with diastolic blood pressure percentile for female (β = −1.86, 95% CI: −3.31, −0.41), non-Hispanic White (β = −2.89, 95% CI: −5.61, −0.17), and other race/ethnicity youth (β = −3.38, 95% CI: −5.79, −0.96)

Table 3.

Adjusted estimates for the association between cardiovascular outcomes and COI indicator z-scores for the full cohort and stratified by sex and race/ethnicity ^a,b

	Commute Duration		Walkability		Greenspace		Industrial Pollutants
	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI
Diastolic Blood Pressure
Full Cohort	−1.04	−2.08, 0.00	−0.12	−0.67, 0.43	−0.21	−1.34, 0.92	−0.74	−2.32, 0.83
Sex
Female	−1.86	−3.31, −0.41	0.47	−0.3, 1.25	−1.42	−3.01, 0.17	−1.08	−3.31, 1.15
Male	−0.12	−1.62, 1.38	−0.73	−1.51, 0.06	1.03	−0.59, 2.64	−0.16	−2.4, 2.08
Race/ethnicity
Non-Hispanic Black	−0.86	−2.84, 1.13	0.03	−0.96, 1.01	−0.18	−2.25, 1.89	−0.29	−3.32, 2.73
Non-Hispanic White	−2.89	−5.61, −0.17	−0.07	−1.64, 1.51	0.40	−3.04, 3.84	0.17	−3.81, 4.14
Hispanic	0.93	−0.72, 2.57	0.16	−0.75, 1.06	−1.03	−2.76, 0.70	−0.66	−3.74, 2.41
Other race/ethnicity c	−3.38 *	−5.79, −0.96	−0.57	−1.88, 0.73	0.52	−2.21, 3.25	−1.29	−4.31, 1.74

Abbreviations: child opportunity index (COI), confidence interval (CI)

^(a)Boldface indicated statistical significance, (p<0.05)

^*indicates p < 0.01

^(b)Models adjusted for insurance type (private, public, other, missing), age, sex, (full and race/ethnicity subgroups), race/ethnicity (full and sex subgroups)

^(c)Other is not specified; “other” is an option that can be selected within the electronic health record

Discussion

Youth obesity is an ongoing and worsening public health concern, and identifying neighborhood-level factors associated with disease risk is critical in order to inform policy interventions. Our study examined the longitudinal association between neighborhood built environment factors, as measured by the COI, and child BMIp95 and blood pressure among a diverse population of children and adolescents receiving obesity treatment at a multidisciplinary pediatric weight management clinic. Shorter area level commute duration and greater walkability were negatively associated with BMIp95, while higher availability of greenspace was positively associated with BMIp95. Commute duration was associated with diastolic blood pressure percentile, but only for female, non-Hispanic White, and other race youth. Clinically, these environmental factors should be considered when designing treatment plans and interventions for children and adolescents with obesity.

Decreased commuting time at the census tract level was associated with lower BMIp95 and diastolic blood pressure percentile. These findings are consistent with previous research investigating the built environment.¹⁵ Higher commuting distance has been shown to be associated with poorer cardiovascular fitness and increased obesity in adults.^29,30 It is thought that these associations in adults are due to trade-offs in time that may reduce health-promoting behaviors, such as less time for cooking or food preparation and physical activity.³¹ For children and adolescents, parents or adults in the home may have less time to engage in physical activity with their children after work or to prepare healthy meals if they have longer commute time and duration, which may in turn contribute to elevated weight.

Higher neighborhood walkability scores were associated with lower BMIp95 but there was no association with diastolic blood pressure percentile. The association between increased walkability and lower youth BMI has been reported in prior studies.^32–34 Greater walkability promotes physical activity, which in turn may contribute to lower BMI.^35–37 The relationship between blood pressure and walkability in youth is less clear; however, walkability has been associated with lower cardiometabolic risk factors in adult populations.^38,39 Additional research is needed to fully understand how walkability is associated with cardiovascular disease risk in youth. Regardless, walkability remains a clinically important target for public health efforts to reduce burden of obesity in youth, specifically at the policy level where policies can be implanted to improve neighborhood walkability. Data systems that provide this information to clinicians at the point of care may also be helpful, as this could provide clinicians with important contextual information to inform physical activity recommendations.

Increased greenspace in neighborhoods was associated with higher BMIp95, but not diastolic blood pressure percentile. The directionality of this relationship was the opposite of expected. However, past studies have also seen inconsistent relationships between greenspace and cardiovascular risk factors in children and adolescents.^34,40–42 The measure used by the COI for greenspace assesses the percent of impervious surfaces inverted to calculate a score from greenspace, which may inaccurately include surfaces that are not greenspaces and may include some city parks that do include impervious surfaces.⁴³ The COI greenspace score also uses 2011 satellite data, which may not reflect the greenspace conditions in 2013 to 2020 when patient data was collected. Other measures for greenspace, such as distance to nearby parks, may be more relevant to examine the relationship between child health outcomes and neighborhood context in the future. It should also be noted that this measure is assessing availability of greenspace and not actual use. For instance, a study of 10–11 year old children in the United Kingdom found that most of children’s activity outdoors was not spent in greenspace, suggesting that non-green urban environments are just as important for youth physical activity.⁴⁴

Levels of industrial pollutants in air, water, or soil were not associated with BMIp95 or diastolic blood pressure percentile, similar to prior studies that have also demonstrated an inconsistent or lack of relationship.⁴⁵ Research that has shown strong relationships often correlate air pollution with traffic densities in larger metropolitan areas.⁴⁶ Continued research is needed in this area to fully understand the impact of pollutants on child health.

When stratified by sex, the majority of associations remained similar for BMIp95. However, the association between more greenspace and higher BMI was only significant in male compared to female youth. Prior studies have shown that female youth are less likely to use neighborhood parks, so this association was unexpected, but may be in part due to the greenspace measure used.⁴⁷ Commute duration was the only COI indicator associated with blood pressure percentile, but only in female youth. Prior research has shown that preschool-aged girls are less likely to have parental-supervised outdoor play time than boys, partly due to attitudes that they might not need to be as active.^48,49 This may continue beyond preschool into the school-aged years as well. Therefore, it is plausible that increased commute duration and the time constraints that places on parents have more of an impact on female children in our population than male children. Overall, our results are in line with sex differences seen in the literature: stronger associations between neighborhood factors and girls’ health outcomes (lower rates of obesity and lower BMI values).^50–52

When examining differences by race/ethnicity, we observed stronger associations in the non-Hispanic White subgroup between the neighborhood environment and BMIp95. One reason may be that there are cultural contexts or structural influence that contribute to health disparities in these subgroups that are not accounted for in our analysis or reflected in the neighborhood quality measures. Another explanation for differences that are seen could be residential segregation in North Carolina, which is reflected within census tract clustering. North Carolina metropolitan areas show moderate segregation between racial and ethnic groups compared to national standards, so stratification by race/ethnicity may be revealing variability in this relationship across the geographical environments that subgroups inhabit.⁵³

Strengths and limitations

Strengths of this analysis include the diverse sample population across sex and race/ethnicity. Other analyses have occurred in larger, denser metropolitan areas, but the population encompassed in this analysis and in the specialized childhood obesity clinic includes a heterogeneous population in terms of racial/ethnic groups, income, and other factors predictive of obesity. In addition, prior research is often cross-sectional, but this data is longitudinal, drawn from repeated measures patient encounters. However, there are several limitations to the current study. First, the electronic health data used is purely observational, so causality cannot be assumed for any of the demonstrated associations. Second, the COI data largely came from 2010–2015, whereas the child outcome data was collected throughout 2013 to 2020, so the exposure data may not always represent accurate neighborhood conditions as neighborhoods continue to evolve (e.g., building development may impact greenspace). Additionally, there are limitations to some of the exposure variables. For instance, the greenspace measure does not capture incidental greenspace (e.g., empty lot) compared to purposeful greenspace (e.g., parks), while adult commute duration does not directly measure child transportation vulnerability. Third, since patients were drawn from a pediatric weight management clinic in a single geographic area, they represent a unique population of treatment-seeking individuals, so these results cannot be generalized to all children with obesity and those living in different geographic areas. Additionally, the majority of the sample was from urban census tracts, so these findings may not generalize to youth with obesity living in rural areas. Finally, we were unable to account for treatment adherence that may influence results such as changes in BMI.

Conclusion

Prevalence of childhood obesity and cardiovascular disease risk factors continue to increase in children and adolescents in the US. This study of children and adolescents attending a multidisciplinary childhood obesity management clinic demonstrated that neighborhood factors are linked with lower BMIp95 over time for children living in more walkable neighborhoods with shorter commute times. Findings also showed that increased greenspace was associated with increased BMIp95 over time, suggesting that focusing on use of greenspace should be a priority for future obesity treatment. Future research should continue to investigate the association between neighborhood factors and cardiovascular health outcomes to reduce disparities in youth with obesity.