Disparities in pediatric obesity during COVID-19: The role of neighborhood social vulnerability and collective efficacy

Abstract Introduction: Childhood obesity increased in the first year of Covid-19 with significant disparities across race, ethnicity, and socioeconomic status. Social distancing led to fewer physical activity opportunities but increased screen time and high-calorie food consumption, all co-determined by neighborhood environments. This study aimed to test the moderation effects of neighborhood socioeconomic and built environments on racial/ethnic disparities in obesity change during Covid-19. Methods Using electronic health records from a large pediatric primary care network in 2018–2022, we cross-sectionally examined 163,042 well visits of 2–17 year-olds living in Philadelphia county in order to examine (1) the pandemic’s effect on obesity prevalence and (2) moderation by census-tract-level neighborhood socioeconomic disadvantage, crime, food and physical activity-related environments using interrupted-time-series analysis, Poisson regression, and logistic regression. Results Weekly obesity prevalence increased by 4.9 percent points (pp) during the pandemic (Jan 2021-Aug 2022) compared to pre-pandemic (Mar 2018- Mar 2020) levels. This increase was pronounced across all age groups, racially/ethnically-minoritized groups, and insurance types (ranging from 2.0 to 6.4 pp) except the Non-Hispanic-white group. The increase in obesity among children racially/ethnically-minoritized groups was significantly larger in the neighborhoods with high social vulnerability (3.3 pp difference between high and low groups), and low collective efficacy (2.0 pp difference between high and low groups) after adjusting for age, sex, and insurance type. Conclusions Racially/ethnically-minoritized children experienced larger obesity increases during the pandemic, especially those in socioeconomically disadvantaged neighborhoods. However, the buffering effect of community collective efficacy on the disparities underscores the importance of environments in pediatric health.


Introduction
Obesity and Covid-19 were concurrent pandemics in 2020 and 2021. 1 In the US, adult obesity prevalence rose by 3 percent points (pp) ,2 as well as childhood obesity, 3 from 1.7 pp 4 to 3.1 pp 5 increase during the rst year of the pandemic.
The literature identi ed school closures, virtual learning, disrupted routines, increased stress, depression, and sedentary time in conjunction with decreased opportunities for physical activity and proper nutrition as reasons for the worsening childhood obesity. 6Studies warned that the short-term changes could become permanently entrenched. 3,7Notably, neighborhood socioeconomic and built environments can play an important role in obesity by determining access to safe and convenient places for physical activity, healthy food options, and community networks to get resources and information. 6,8During the Covid-19 crisis, residential neighborhoods acted as sources of stress or buffers against psychological distress during the pandemic through mechanisms including crime rate, parks, and collective e cacy 9 .In the early pandemic, urban adolescents had a more considerable decrease in physical activity than rural adolescents. 10Increased outdoor activity was more likely to occur by access to parks, even in densely populated areas. 11The pandemic worsened inequities in the social determinants of health through food insecurity, unemployment, poverty and lack of access to the internet. 12,13The structural racism resulted in disparate pandemic impacts on racially/ethnically-minoritized communities, including high Covid-19 incidence and mortality.
Philadelphia is a US Northeast megalopolis with high obesity prevalence, with signi cant regional-racial disparities. 14iladelphia's racial/ethnic disparities in obesity further widened by 3 pp in 2020. 4While previous studies have examined neighborhood moderation effects on racial/ethnic disparities in obesity generally, 6,[15][16][17] the extent to which neighborhood environments exacerbated or buffered against these disparities among children during the pandemic is unknown.
To address this knowledge gap, we examined the moderation effects of neighborhood socioeconomic and built environments on racial/ethnic disparities in obesity prevalence changes during the COVID-19 pandemic among children in a large pediatric primary care network in Philadelphia.We hypothesized that: 1) obesity prevalence had increased by the Covid-19 pandemic, 2) racially/ethnically-minoritized groups experienced higher obesity increases than non-Hispanic white (NHW) children, 3) neighborhood socioeconomic disadvantage and higher crime exacerbated racial/ethnic disparities in obesity increases while tree-cover, walkability, parks/playgrounds, healthy food retailers, and collective e cacy buffered it.

Study design and study sample
This cross-sectional study examined 2-17 year-old children's well visits between March 2018 and August 2022 in a large pediatric primary care network across 31 clinics that serves approximately 300,000 patients annually in the Philadelphia metropolitan area.We included 73,225 patients (163,349 encounters) with a measured height and weight who lived in Philadelphia County, as Covid-19 disproportionately impacted children in densely populated areas. 10,11We excluded patients with implausible BMI by the modi ed z-scores in the CDC growth charts published in 2022 (0.2%). 18This study was deemed exempt by the Children's Hospital of Philadelphia Institutional Review Board.

Assessment and Measures
Obesity.We determined the child BMI Z-score by age-speci c and sex-speci c BMI percentiles using the 2000 US CDC Growth Charts.Obesity was de ned as ≥ the 95th BMI percentile. 19VID-19 pre-pandemic vs. pandemic.We de ned pre-pandemic from March 1 2018 to March 22 2020.After Covid-19 hit, primary care visit volume suddenly dropped due to stay-at-home orders, 4,20 and equipment and staff shortage 21 continued until January 20, 2021 in the care network.We designated the intermittent period from March 23 2020 to January 20 2021.We de ned pandemic period from January 21, 2021 to the end of August 2022, during which both the patient and clinic activities returned to normal.
Other patient characteristics.Patient sex, race/ethnicity, age, and insurance type at each visit were collected from electronic health records.Patient race-ethnicity was classi ed as Hispanic, non-Hispanic black (NHB), NHW, and other, and was included as a marker for potential exposure to structural racism.Neighborhood characteristics.We geocoded patient's home address and de ned their residential census tracts as a proxy for neighborhoods.[24][25] A. Sociodemographic and economic characteristics: We used year-matched American Community Survey data, when available, to de ne below.
1) The CDC's social vulnerability index (SVI): A ranking of a community's resilience following public health disasters, based on 15 social factors in four domains (socioeconomic status, household composition-disability, minority statuslanguage, and housing type-transportation; ranged 1 to 100, with higher scores indicating more vulnerability) 26 .
2) Child's opportunity index (COI) within metropolitan areas: A ranking of neighborhood quality for child development via 29 neighborhood resources and conditions across three domains (education, health-environment, and socialeconomic; ve categories, with higher categories re ecting more favorable opportunities). 27 Index of concentration at the extremes (ICE): Comparing NHW households with the top quintile of US income vs. people of color households with the bottom quintile of US income to capture economic and racial/ethnic segregation (ranged − 1 indicating everyone from the deprived group to 1 indicating everyone from the privileged group).28 1.B.Part 1 crime: The Philadelphia Police Department de nes part one crime as homicides, rapes, robberies, aggravated assaults, and thefts.29 We counted the total number of occurrences in each tract per year.

C.Physical activity-related built environments: The EnviroAtlas community map by the US Environmental
1][32] EPA de ned the National Walkability Index (ranging from 1 to 20; higher scores indicating more walkability) based on measures such as street intersection density, proximity to transit stops, and diversity of land uses. 33,34Percent of park/playground area within the tract was calculated using the data from OpenDataPhilly. 35We de ned tree and park/playground as an area percentage of the neighborhood rather than the distance to the closest one, as families would bene t more from multiple choices during a long period of social distancing.

D.Food environments:
The CDC identi es the modi ed Retail Food Environment Index (mRFEI) as the percent of healthy food retailers vs. the total food retailers. 36Fast-food expenditure ratio was calculated as the percentage of all fast-food restaurants expenditures vs. the total food-related expenditures using Esri Consumer Spending Data. 37

E. Collective e cacy: Collective e cacy indicates neighborhood level of trust, cohesion, and the willingness to
intervene for the common good among residents.9][40] We employed the Southeastern Pennsylvania Household Health Survey's social capital scale, using ve relevant community questions (involvement in local groups and organizations, neighbors work together, community improvement, sense of belonging, and feelings of trust 41 The Likert-scale responses were summed and scored between 1-10 points.After considering the survey sampling weight, we averaged the score to the tract level.

Statistical analysis
We described patients' individual and neighborhood characteristics at the encounter level.Using interrupted-timeseries analysis with three events: 1) Covid-19 started (12th week of 2020), 2) primary care patient volume returned to normal (25th week of 2020) and 3) clinic equipment and staff shortage ended (3rd week of 2021), we explored the change of weekly obesity prevalence over time.
We further analyzed the average weekly obesity point prevalence in the above periods and their changes between the pandemic vs. pre-pandemic by age, race-ethnicity, and insurance type using logistic regression with marginal standardization to calculate the absolute changes (percent point differences) and 95% con dence intervals, and Poisson regression to show the changes on a relative scale (prevalence ratios).
To estimate the neighborhood moderation on obesity increases during the pandemic, we rst examined the differences in neighborhood characteristics across racial/ethnic groups using chi-squared tests.Then we tested the interaction between the obesity increase during the pandemic and the level of each neighborhood characteristic using logistic regression, strati ed by NHW and racially/ethnically-minoritized groups.All racial/ethnic minority categories were combined into one group because the distributions of neighborhood categories in the NHB and Hispanic groups were too skewed to examine their moderation on racial/ethnic-speci c obesity increases.
We selected SVI, ICE, COI, part 1 crime, percent of tree-coverage, and collective e cacy as they signi cantly differed across racial/ethnic groups.Fast-food expenditure ratio, percent of park/playground area, mRFEI, and walkability were dropped for their low variations in Philadelphia County.In the multivariable model, we kept all neighborhood characteristics but dropped COI and ICE due to their high collinearity with SVI.COI raised colinearity issues with other characteristics as it includes many domains.SVI covers more socioeconomic characteristics than ICE.As we duplicated these models using two-level mixed effects model with children nested within clinics, models yielded trivial differences in effect estimates.Statistical analyses used STATA version 17 (Stata Corp, College Station, TX).

Characteristics of the study population
We examined 163,042 encounters of 2-17-year-old children living in Philadelphia County between 2018 and 2022 (Supplementary Table 1).One-fth had obesity; the largest racial/ethnic group was NHB (58%), followed by NHW (19%), and Hispanic (10%).More than half received Medicaid, and most lived within 30 minutes of their clinic (97%).All neighborhoods of Philadelphia County have high population density (mean: 0.009, SD: 0.004).Children's neighborhoods were primarily low in COI (83%) and high in ICE (70%) and social vulnerability (mean (SD) = 70 (25)   percentile).On average, children were exposed to 158 part 1 crimes per year (SD = 87), trees covered 19% of neighborhood land, and 16% had neighborhoods with high collective e cacy.The percentages of park/playground areas and health food stores were low, while walkability and the fast-food expenditure ratio to total food expenditure were high, with small variations.
Percentages present crude prevalence of obesity in three periods.
Percent point differences indicate the absolute change in obesity prevalence from prepandemic to pandemic period, with 95% CIs calculated by using logistic regression and marginal standardization.
Prevalence ratios and interaction p values present the relative change in obesity prevalence in the pandemic period vs. the prepandemic period by using Poisson models.
Percentages present crude prevalence of obesity in three periods.
Percent point differences indicate the absolute change in obesity prevalence from prepandemic to pandemic period, with 95% CIs calculated by using logistic regression and marginal standardization.
Prevalence ratios and interaction p values present the relative change in obesity prevalence in the pandemic period vs. the prepandemic period by using Poisson models.

Racial/ethnic differences in neighborhood characteristics
Children from racially/ethnically-minoritized-groups had signi cantly higher exposure to adverse neighborhood conditions compared to the NHW group (Supplementary Table 2): high in social vulnerability, extreme concentrations, and part 1 crime, while lower in COI, tree-covered land, and collective e cacy.The NHB group had the highest proportion in the most adverse categories of SVI (65% in high), ICE (29% in very high), COI (76% in very low), and part 1 crime (30% in 4th quartile), followed by the Hispanic group.The Hispanic group was more likely to be in the most adverse categories of tree-covered land (53% in < 10%) and collective e cacy (51% in low).

Neighborhood moderation on racial/ethnic in obesity increase during Covid-19
We examined the six neighborhood characteristics' association with the change in obesity prevalence between pre-Covid-19 and the pandemic separately.Overall, SVI, ICE, and part 1 crime consistently had dose-response associations with obesity increase; a higher magnitude of obesity increase was shown in children from neighborhoods with high SVI, ICE, and part 1 crime after adjusting for individual age, sex, insurance type, and Covid-19 community transmission level (all p trends < 0.01; Table 2).However, the magnitude of obesity increase was smaller when tree-coverage, COI, and collective e cacy (all p trends < 0.01) were higher.As strati ed by race-ethnicity, the NHW group had no neighborhood moderation effect on obesity increase during Covid-19.In contrast, SVI, ICE, COI and collective e cacy showed moderation effects on obesity increase among the racially/ethnically-minoritized group with a dose-response relationship.In the multivariable model, the dose-response moderation effects of SVI, tree-coverage and collective e cacy on the obesity increase during Covid-19 remained signi cant for the overall group (Supplementary Table 3).For the SVI (3.3pp difference between high and low categories) and collective e cacy (2pp difference between low and high categories) remained signi cant moderators of the obesity increase during Covid-19, after adjusting for all other individual and neighborhood characteristics.Notably, minorities who lived in the low SVI neighborhoods did not have an obesity increase like the NHW group did (Fig. 2).

Discussion
In this study, we found that pediatric obesity prevalence had a 5 pp increase during the pandemic (Jan 2021-Aug 2022).This increase was especially pronounced among racially/ethnically-minoritized group.Children from racially/ethnically-minoritized-groups lived in neighborhoods with greater socioeconomical disadvantages, high crime, lower tree-coverage, and lower collective e cacy on average compared to the NHW group; their higher increases in obesity during the pandemic were exacerbated by the level of neighborhood social vulnerability and buffered by high collective e cacy, even after adjusting for individual and neighborhood characteristics.
Our ndings are consistent with studies from the early pandemic, which reported increased pediatric obesity and further widened racial/ethnic disparities. 4,5.However, previous studies did not examine the impact of neighborhood environments on the racial/ethnic disparities in obesity increase during the pandemic, despite well-documented impacts of structural racism on both neighborhood conditions and the impact of the pandemic on racial/ethnic minority communities. 12,13The pandemic worsened inequities in the social determinants of health.Particularly high Covid-19 incidence and its mortality rate among racially/ethnically-minoritized groups included factors like crowded housing, limited English pro ciency, and a high proportion of single-parent households 12,13 which are associated with childhood obesity via lack of space for physical activity and di culties accessing health care and health-promoting resources around the community.We addressed this research gap by presenting the different levels of obesity increase in racially/ethnically-minoritized group during the pandemic by their neighborhood sociodemographic, economic and built environmental characteristics.
Our study examined Philadelphia, a city with high obesity prevalence and racial/ethnic disparities in health.Like previous studies, 12,13 children from racially/ethnically minoritized-groups were more likely to live in neighborhoods with high social vulnerability, crime, low incomes, and fewer child opportunities, tree-covered land, and collective e cacy.Minorities' higher obesity increase relative to the NHW group during the pandemic differed by the level of neighborhood disadvantage.Notably, the prevalence of obesity did not increase among children from racially/ethnically-minoritized-groups who lived in the low SVI neighborhoods, which suggests the neighborhood moderation effect on childhood obesity during the pandemic.
Although SVI, COI and ICE describe neighborhood social determinants of health with different subdomains and factors, all three showed signi cant associations with childhood obesity increase in this study.In particular, SVI examines crowding and multi-unit structures in housing and single-parent households.COI includes access to green space, healthy food, walkability, single-headed households, and high-skill employment.These factors are directly linked with the availability of physical activity, healthy diets, the possibility of parenting burnout during the pandemic crisis, and critical for obtaining Covid-19-related information.
The frequency of crime was strongly associated with SVI, COI and ICE in this study.Frequent crime and smaller treecoverage/green space challenge family's outdoor activities and are associated with children's obesity. 6More green spaces were associated with smaller increases in childhood obesity risk in early pandemic. 42However, we did not nd signi cant associations of crime and tree-coverage with racial/ethnic minorities' obesity increase during pandemic.We may not have enough power to detect the effect of tree-covered land on obesity increase among racially/ethnically-minoritized children.The fear of Covid-19 transmission in crowded neighborhoods may challenge racially/ethnically-minoritized families to be active outside regardless of crime rates.Although we couldn't examine the effect of neighborhood parks/playgrounds, and walkability due to their low variations within Philadelphia, their buffering effects on childhood obesity could be less impactful than greenspace or tree-coverage because of the temporal closure of parks/playgrounds and decreased park/playground usage in early pandemic due to Covid-19 transmission risk in public spaces.
While the disadvantaged social and physical conditions may create perpetuation of inequities in pandemic obesity, it is worth noting that collective e cacy showed a buffering effect to the obesity increase among racially/ethnicallyminoritized children.Our nding of the favorable in uence of collective e cacy on childhood obesity is consistent with previous studies that reported lower BMI and overweight risk among adults and adolescents, and black adults with low household income, 15,38,43 and a higher fruit/vegetable intake among mothers of toddlers receiving Medicaid 40 who had a high level of collective e cacy.Collective e cacy has been identi ed as the willingness of community members to intervene together when a problem occurs. 44Thus, the residents in high-e cacy neighborhoods are likely to have more social interactions that result in greater conformity (e.g., adhering more to recommended guidelines for obesity-related health behaviors), 45 and get better infrastructure and social services. 46- 48We speculate that the residents with high e cacy may better share information, opportunities, and advocate for a healthy lifestyle during the pandemic.

Limitations
Our data come from a single care network in one city with a majority of non-Hispanic black, which may limit generalizability.There was low variation in neighborhood walkability, park/playground areas, and food environments, which limited our ability to examine moderation by these features.We did not have enough power to detect moderation by neighborhood characteristics for more granular racial/ethnic categories and have limited years of data for COI and collective e cacy.The dataset only included children who attended well visits during the study period and thus might underrepresent children facing the highest levels of disadvantage.Further follow-up studies are needed to examine the pandemic's long-term effect on racial/ethnic disparities in childhood obesity.

Conclusions
During the rst three years of the Covid-19 pandemic, neighborhood social vulnerability was associated with larger increases in obesity prevalence among racially/ethnically-minoritized children.However, results suggest racial/ethnic disparities could be buffered by neighborhood collective e cacy, which underscores the importance of neighborhood social environments in pediatric health.Further research is warranted to identify how to enhance the community-level interventions in improving continuous obesity disparities among children.

Abbreviations
COI=Child's opportunity index; ICE=Index of concentration at the extremes; mRFEI= modi ed Retail Food Environment Index; NHB=non-Hispanic black; NHW= non-Hispanic white; SVI=social vulnerability index.

Figure 1 Time
Figure 1

Table 1
Changes in weekly obesity prevalence during study periods

Table 2
Separate regression models for the association between each neighborhood characteristic and the obesity increase during Covid-19 in NHW and racially/ethnically-minoritized groups Bold indicates signi cant trends (p trend < 0.05) in the obesity increase across environmental categories.Bivariate Logistic regression models after adjusting for patient sex (female, male), age (2-5y, 6-11y, 12-17y), insurance type (Medicaid vs. others), and covid-19 community transmission level (high vs. others).Fastfood expenditure ratio, park/playground, mRFEI, and walkability were dropped for their low variations in Philadelphia county.