Associations of Longitudinal BMI-Percentile Classification Patterns in Early Childhood with Neighborhood-Level Social Determinants of Health

Abstract

Background:

Understanding social determinants of health (SDOH) that may be risk factors for childhood obesity is important to developing targeted interventions to prevent obesity. Prior studies have examined these risk factors, mostly examining obesity as a static outcome variable.

Methods:

We extracted electronic health record data from 2012 to 2019 for a children’s health system that includes two hospitals and wide network of outpatient clinics spanning five East Coast states in the United States. Using data-driven and algorithmic clustering, we have identified distinct BMI-percentile classification groups in children from 0 to 7 years of age. We used two separate algorithmic clustering methods to confirm the robustness of the identified clusters. We used multinomial logistic regression to examine the associations between clusters and 27 neighborhood SDOHs and compared positive and negative SDOH characteristics separately.

Results:

From the cohort of 36,910 children, five BMI-percentile classification groups emerged: always having obesity (n = 429; 1.16%), overweight most of the time (n = 15,006; 40.65%), increasing BMI percentile (n = 9,060; 24.54%), decreasing BMI percentile (n = 5,058; 13.70%), and always normal weight (n = 7,357; 19.89%). Compared to children in the decreasing BMI percentile and always normal weight groups, children in the other three groups were more likely to live in neighborhoods with higher poverty, unemployment, crowded households, single-parent households, and lower preschool enrollment.

Conclusions:

Neighborhood-level SDOH factors have significant associations with children’s BMI-percentile classification and changes in classification. This highlights the need to develop tailored obesity interventions for different groups to address the barriers faced by communities that can impact the weight and health of children living within them.

Introduction

Childhood obesity is a complex multifactorial disease.^1–3 The ecological model of childhood obesity⁴ points to the important role of not only individual-level but also neighborhood-level social determinants of health (SDOH) in the development of childhood obesity. Some of the main neighborhood-level SDOH that have been linked to childhood obesity include single-parent households, housing insecurity, lower socioeconomic status (SES), lower parental educational status, and employment.^5–14 Historically under-resourced populations affected by the negative SDOH are not only disproportionately burdened by childhood obesity but have also shown a lower success rate of obesity prevention interventions.¹⁵ Targeting these community-level risk factors can help provide children with a healthy environment to prevent obesity.

Interventions designed to prioritize at-risk communities could be impactful in early childhood as most childhood and adult obesity begins early in life—as early as age 2.¹⁶ Other important time points for weight gain (e.g., adiposity rebound) occur before age 5 and definitely by around age 7.^17,18 Baidal et al.¹⁶ found that modifiable risk factors like family SES and child care attendance were associated with childhood obesity in the first 2 years of life, and Tester et al.¹⁹ found that poverty and single-parent household status were associated with severity of obesity in children aged 2–5.

There is a wealth of literature describing the relationship between selected SDOH factors and childhood obesity using cross-sectional analyses.^20–34 However, these cross-sectional studies assume that all children with obesity are a homogeneous group with similar growth patterns, but the obesity growth pattern of children while developing obesity could have different implications for intervention.³⁵ For example, children who have obesity from birth may have genetic and epigenetic risk factors and intervention may need to occur in the perinatal period, whereas children born with normal weight but then develop obesity during early childhood may benefit from interventions that can impact their environment and lifestyle. Therefore, to better understand the impact of SDOH on a child’s risk of obesity at different time periods, it would be important to identify how SDOHs are associated with different weight trajectories of children throughout early childhood.^35,36

While many existing longitudinal studies evaluate BMI, BMI-z, or BMI-percentile trajectories in children,^37–39 with many also describing how these trajectories are associated with SDOHs.^40–50 However, there have been no studies to date that have applied advanced algorithmic clustering techniques to establish distinct groups of patients with clinically relevant trajectories (e.g., normal weight to obesity as defined by CDC), and few studies that have examined these associations in early childhood or weight trajectories with ungrouped neighborhood-level SDOH variables to better understand the individual contribution of SDOH that can better inform policy changes to target factors.^39,44,51 Therefore, in this work, we aim to study the association between clinically relevant BMI-percentile status throughout early childhood and a series of major neighborhood-level SDOH and individual-level sociodemographics using an electronic health record (EHR) dataset.

Materials and Methods

Data Source

We obtained patient-level EHR data from Nemours Children’s Health—a large, multistate pediatric health system in the United States. Our study was approved by the Nemours Institutional Review Board. The initial dataset was extracted from PEDSnet dataset of children 0–21 years old with at least 5 years of medical data with patient records from 2002 to 2019 for all inpatient and outpatient visits.⁵² Patients with complex medical conditions (requiring intensive treatment, hospitalizations, etc.) of type 1 diabetes, cancer, or sickle cell diseases (full list in Supplementary Data S8), which are normally excluded from our data center’s extractions are also excluded from this cohort.

Study Sample and Anthropometric Measurements

Due to regular pediatric visits in early childhood, our original data had a higher number of weight and height measurements among younger children (who have at least one visit between 0 and 7 years), with fewer measurements documented with increasing age and the largest drop in measurements after age 7. Because of this and because of our focus on early childhood, we included only 67,242 children of age 0–7. Because our original dataset mostly comprised young children, we only excluded 787 children in this step. We excluded 11 patients with no weight or height measurements or demographic data and 19,052 without a minimum of 5 years of recorded weight and height measurements. Since weight and height measurements for children were recorded over irregular intervals, we segmented the time-series data into nonoverlapping 3-month windows to obtain smooth time series with equal intervals. For every window, we took the median weight and height measurements. If weight and height measurements were missing in a certain 3-month window, we mark it as missing. We excluded 7,286 children with more than 80% missing rate for recorded weight and height measurements. Here we note that out of 26,338 (19,052 + 7286) patients excluded in the previous two steps, 15% lived in a census with more than 20% population with annual income below the federal poverty level compared to 23% in our final cohort. This shows that our final cohort had enough representation of the low and high SES populations compared to the excluded population to fulfill our study objective.

We classify patients based on CDC categories that use BMI percentile, as having a normal weight (BMI percentile <85), overweight (85≤ BMI percentile <95), or obesity (BMI percentile ≥95). Weight and height were used to calculate BMI and reference BMI percentile provided by the World Health Organization⁵³ and CDC⁵⁴ for children between 0 and 2 and older than 2 years, respectively. We choose to use BMI percentile to remain consistent with the measurements used for the cohort across 0–7 years of age.^55–58 Additional notes about our choice of BMI percentile are presented in Supplementary Data S1.

Finally, to keep the focus of our study on studying SDOH associations between children with normal, overweight, or obesity BMI percentiles, we also excluded underweight patients (BMI percentile ≤5) and never crossed the threshold of 85 (more details in Supplementary Data S2 and Supplementary Table S1). This resulted in a cohort of 36,910 patients. All cohort selection steps are summarized in Figure 1, and Table 1 shows the general characteristics of our final cohort.

Figure 1.

Cohort selection and processing steps. The initial extracted data of 68,029 children from Nemours Pediatric EHR only included children who had at least 5 years of medical data irrespective of how many observations in the 5 years. Please note that 5 years of medical data do not assure 5 years of weight and height measurements. Therefore, we exclude 19,052 children in step 3. Since young children tend to have more regular visits, our original dataset was mostly comprised of young children. Therefore, our dataset ends up having more children who have at least one visit between 0 and 7 years. Therefore, in step 1, only 787 children are excluded. EHR, electronic health record.

Table 1.

Characteristics of the Full Cohort

Population (N)	36,910
Male/female	19,920/16,990
Mean BMI percentile (±SD)	64 (±7.18)
Ethnicity
Hispanic	5076 (13.7%)
Non-Hispanic	31,635 (85.70%)
Race
Asian	638 (1.72%)
Black	15,195 (41.16%)
White	16,816 (45.55%)
Other Race	4602 (12.46%)
Insurance
Public	21,431 (58.06%)
Private	15,415 (41.76%)

SD, standard deviation.

Figure 2.

Five groups were identified in our study based on BMI-percentile trajectories of children between 0 and 7 years. We used a smoothing function (geom smooth package in R) to represent each cluster by a smooth solid line with 95% CI. The figure on the left shows the trajectories found using data-driven clustering and partitional clustering with DTW. The figure on the right shows the trajectories found using data-driven clustering and partitional clustering with LCGMM. Technical details about LCGMM experiments and results are presented in Supplementary Data S5. CI, confidence interval; DTW, dynamic time warping; LCGMM, latent class growth mixture modeling.

Linking Patient Data to Neighborhood-Level SDOH Data

Data collected in the EHR about patient race, ethnicity, insurance type, and addresses were included in the cohort. The Census-Geocoder⁵⁹ address look-up tool allowed us to map patient visits to 4237 unique census tract geocodes. We mapped data obtained before 2010–2000 census tract boundaries and after 2010–2010 census tract boundaries. In total, 23% of patients had a relocation, out of which 58% had no relocation after 3 years of age showing that their latest census tract is also the longest-stayed census tract. We used the latest census tract for each patient.

To capture the most essential neighborhood-level SDOH that may influence child development, we identified the union of variables from three standardized indexes: Child Opportunity Index,⁶⁰ Social Vulnerability Index,⁶¹ and Social Deprivation Index.⁶² We grouped these variables into five categories: (1) SES, (2) neighborhood household composition, (3) housing and transportation, (4) education and occupation, and (5) RUCA (rural–urban commuting area) code.⁶³ Evaluating the association between individual SDOHs instead of high-level indexes and BMI-percentile trajectories allows us to determine which aspects of neighborhood risk have the most impact. This allows us to better inform practice and policy. All SDOHs are the ratio of the population belonging to that SDOH except for median household income. RUCA is the ratio of the population with RUCA ≥4. Additional details are presented in Supplementary Data S3.

Clustering BMI-Percentile Trajectories

Data-driven clustering

We first used a traditional data-driven clustering approach (i.e., using only raw data without further analysis) to find any homogeneous groups whose BMI-percentile category remained in the same category (normal, overweight, and obesity) over time. We found two groups: (1) BMI percentile always in the normal weight range and (2) BMI percentile always in the obesity range. The remaining group had significant heterogeneity. The data-driven approach is method-independent to cluster the data for simple stratification.

Algorithmic clustering

However, relying only on raw data for clustering generally offers limited power when working with complex data. An algorithmic approach can capture temporal and nonlinear patterns in the data effectively. Therefore, we used a partitional clustering approach with dynamic time warping (DTW) distance to identify additional homogeneous clusters among the heterogeneous clusters. Technical details about our partitional clustering approach are presented in Supplementary Data S4. Our results from this approach yielded three groups: overweight, increasing, and decreasing BMI percentile.

To further confirm the robustness of the three identified clusters in the choice of algorithmic method, we also used a separate clustering algorithm, latent class growth mixture modeling (LCGMM), which has been used in studying BMI trajectories.^64–67 Technical details about LCGMM experiments and results are presented in Supplementary Data S5.

Statistical Analysis

First, we compare the social characteristics between clusters and then compare the differences in positive and negative SDOH characteristics using analysis of variance (ANOVA) tests. We use multinomial logistic regression to analyze the association between the cluster’s membership and SDOH characteristics. We study the influence of each SDOH on the cluster membership individually. To do that, we dummy-coded each categorical variable (non-Hispanic, Hispanic, Asian, White, Black, Some Other Race, Private Insurance, and Public Insurance) to 0/1 and used the default reference category of 0 for each categorical variable.

Results

Number of Clusters

Using the data-driven clustering technique described in the “Data-driven clustering” section, we found two homogeneous clusters “always having obesity” and “always having normal weight,” leaving all remaining children in a heterogeneous pool of those with BMI percentile recorded from at least two of the normal, overweight, or obesity categories. Next, by applying the algorithmic technique described in the “Algorithmic clustering” section on the heterogeneous pool, we identified three new clusters, which we can label as “overweight most of the time,” “increasing BMI-percentile,” and “decreasing BMI-percentile.” For brevity, we refer to these five clusters as “obesity,” “normal,” “overweight,” “increasing,” and “decreasing” (Figure 2.). From our cohort of 36,910 children, 429 (1.16%) fell into the “obesity” group, 15,006 (40.65%) fell into the “overweight” group, 9,060 (24.54%) fell into the “increasing” group, 5,058 (13.70%) fell into the “decreasing” group, and 7,357 (19.89%) fell into the “normal” group. Technical details about LCGMM experiments and results are presented in Supplementary Data S5.

Race, Ethnicity, Insurance, and BMI Percentile of the Five Groups

Table 2 shows the proportion of children in each cluster, based on race, ethnicity, and insurance status. Groups indicating higher concerns for obesity (“obesity,” “overweight,” and “increasing”) had a higher prevalence of Black (37%–41%) and Hispanic (12%–25%) children than the other groups with <40% Black and <12% Hispanic. They also had a larger proportion with public insurance than private insurance. Whereas the “normal” group had a >50% White children and almost the same proportion with the public (51.92%) and private insurance (47.89%).

Table 2.

Characteristics of Five BMI-Percentile Trajectory Clusters

	Obesity (always having obesity)	Overweight (overweight most of the time)	Increasing (increasing BMI-percentile)	Decreasing (decreasing BMI-percentile)	Normal (always normal weight)	p value^a
Population [n(%)]	429 (1.16%)	15,006 (40.65%)	9060 (24.54%)	5058 (13.70%)	7357 (19.89%)	^
Male/female	263/166	8291/6715	4649/4411	2953/2,105	3764/3593	^**
Mean BMI percentile (±SD)	99.10 (±1.06)	93.26 (±21.37)	65.14 (±26.54)	58.45 (±24.52)	41.18 (±20.40)	^**
Ethnicity
Hispanic [n(%)]	106 (24.7%)	2,470 (16.46%)	1,142 (12.60%)	620 (12.25)	738 (10.03%)	^*
Non-Hispanic [n(%)]	321 (74.28%)	12,536 (83.53%)	7792 (86.%)	4412 (87.22%)	6574 (89.35%)	^**
Race
Asian [n(%)]	3 (0.69%)	190 (1.26%)	181 (1.99%)	121 (2.39%)	143 (1.94%)	^**
Black [n(%)]	164 (38.22%)	6128 (40.83%)	3397 (37.49%)	2013 (39.79%)	2893 (39.32%)	^**
White [n(%)]	178 (41.49%)	6680 (44.51%)	3861 (42.61%)	2395 (47.35%)	3702 (50.31%)	^**
Other Race [n(%)]	90 (20.97%)	2008 (13.38%)	1207 (13.32%)	584 (11.54%)	713 (9.69%)	^**
Insurance
Public [n(%)]	276 (64.33%)	9567 (63.75%)	5407 (59.67%)	2777 (54.90%)	3820 (51.92%)	^**
Private [n(%)]	153 (35.66%)	4991 (33.26%)	4061 (44.82%)	2270 (44.87%)	3524 (47.89%)	^**

Determined using t-test (continuous variables).

p < 0.1 for the difference between trajectory clusters.

p < 0.05.

Association Between the Groups and SDOH

We compare the mean values of the SDOH variables using the population mean for each group, comparing it on continuous heatmap gradient, and ANOVA test across all groups. Figure 3a and b shows the comparison in a heatmap format for negative and positive SDOHs, respectively. The red color shows group mean is higher, and the blue color shows the group mean is lower than the full-cohort mean. At a high level, one can observe that the red boxes are more concentrated on the left side of the 3a, indicating the higher population means in high-risk groups under negative SDOH. The opposite pattern can be observed in 3b. We present a more detailed description of these results in Supplementary Data S6.

Figure 3.

Comparison of mean values for each of the negative (a) and positive (b) SDOH values across the groups using population mean for each group, comparing it on continuous heatmap gradient, and ANOVA test across all groups. Each column in the table corresponds to a group as labeled in the header. Each row corresponds to the SDOH variable and shows an ANOVA p-value for that variable across all groups. Each cell shows the population mean value (with 95% CI) for the SDOH variable, and the color shows the visual comparison of the group mean with the total population mean of the SDOH variable. At the top, we have a heatmap to help understand the color gradient. The red color shows group mean is higher and the blue color shows group mean is lower than the full-cohort population mean. One can especially observe that the higher red heat on the right in (a) is indicative of more negative SDOH factors in the three rightmost high-risk groups and the higher red heat on the left in (b) is indicative of more positive SDOH factors in two low-risk groups identified by our clustering method. The statistical tests were performed using the Researchpy package on Python 3.7. ANOVA, analysis of variance; SDOH, social determinants of health.

Odds of Group Membership Based on the SDOH

Figure 4 shows the results of the multinomial regression for the group membership with “normal” as the reference category. We study the influence of each SDOH listed in Figure 4 on the group membership. The obtained odds ratios show a higher chance of belonging to the higher-risk groups for children with negative SDOH (and vice versa for the positive SDOH). A detailed description of these results is presented in Supplementary Data S7.

Figure 4.

Results from the multivariate logistic regression analysis examining the association of 27 SDOH variables (divided into 8 groups) with the 5 clusters found using our clustering methodology. The units presented in the table are odds ratios. All SDOHs are the ratio of the population belonging to that SDOH except for median household income. RUCA is the ratio of the population with RUCA ≥ 4. Additional details about SDOH are shared in Supplementary Data S3. Odds ratios using “normal (always normal weight)” as the reference category are shown. Each categorical variable—non-Hispanic, Hispanic, Asian, White, Black, Some Other Race, Private Insurance, and Public Insurance dummy-coded to 0/1 and used the default reference category of 0 for each categorical variable. All CIs did not cross 1 and therefore were significant. Because the CIs were so small, they are not visible on the below plot. All categorical variables in the Ethnicity, Race, and Insurance SDOH groups were dummy-coded to 0/1 and used the default reference category of 0. The analysis was performed using the nnet package from R. RUCA, rural–urban commuting area.

Discussion

We identified five distinct groups of BMI-percentile classification trajectories among 36,910 children from birth to 7 years of age. Two groups demonstrated stable BMI-percentile classifications over time, one with a BMI percentile always above 95% (“obesity”) and another with a BMI percentile always <85% (“normal”). Using algorithmic clustering, we also found three additional groups—two groups showed an increasing trajectory, with one group (“overweight”) showing a smaller and the other (“increasing”) showing a larger increase in BMI percentile. A third group (“decreasing”) started with a BMI percentile near 85% but then demonstrated a downward trajectory into a normal BMI-percentile range (<85%).

These five groups are consistent with the groups found in other longitudinal studies that find different BMI trajectories in children.^{37,40,51,68–70} However, the majority of these studies look at trajectories, without a relationship to widely established BMI-percentile classifications. Our identification of a similar five groups, but in relationship to well-established and clinically used BMI-percentile classifications will allow clinicians to better interpret the findings of our study. As an example, it would be easier for a clinician to interpret the higher social risks for children if they are categorized as not only having an increasing weight but going from normal weight to overweight or obese.

Some of these trajectory studies have demonstrated associations between increasing trajectories and individual-level environmental factors like food and housing insecurity and maternal social risk factors like depression and substance use.^51,71–73 Some other studies have examined the association between neighborhood-level SDOH and risk for overweight and obesity,^74–77 but only using cross-sectional data.^{13,45,78–81} Our study is distinct in that it includes the analysis of the association between 27 different social and neighborhood-level SDOH risk factors and five distinct BMI-percentile trajectories based on changes in clinically relevant BMI-percentile classifications during early childhood.

We find three high-risk groups (“obesity,” “overweight,” and “increasing”) with distinct BMI-percentile trajectories and SDOH. This is an important finding of our study because these differences among high-risk groups can be used to inform tailored interventions for each group. Those in the “obesity” and “overweight” groups—children who demonstrated very early onset overweight or obesity from infancy—were more likely to be from neighborhoods with more crowded buildings and households, with a higher percentage of the population from Hispanic and immigrant backgrounds. It, therefore, seems that providing targeted intervention to Hispanic and immigrant communities in the historically under-resourced neighborhoods, especially providing support in the perinatal period, supporting pregnant women and their newborns, and addressing both maternal lifestyle and infant feeding practices, may be important to preventing overweight and obesity in infancy. Whereas the “increasing” group, who developed obesity over time, were more likely to be from neighborhoods with a higher Black population, with higher rates of unemployment, high-school dropouts, and dependence on public assistance. Therefore, providing targeted interventions to Black communities in historically under-resourced neighborhoods to reduce inequities in resources such as employment, and education would be critical to preventing rapid weight gain that might occur in early childhood.

Numerous other studies have shown, factors such as SES, and household structures can also contribute significantly to obesogenic lifestyles.^82–88 Beyond that, neighborhoods can also determine access to health care, education, and transportation and one’s social environment (e.g., norms, relationships, stressors) can negatively impact an individual’s self-efficacy for engaging in health promotion behaviors.^89–93 These barriers disproportionately impact communities of color, who are more likely to live in historically under-resourced neighborhoods due to redlining and lack of investment in their communities from decades of systemic racism.⁹⁴ Therefore, for all of these high-risk groups, it would be critically important to develop effective interventions for reducing obesity-related disparities driven by SDOH and help families overcome barriers to care, such as food insecurity and poverty.^15,95–97

Interestingly, our study also found a group with a decreasing BMI percentile. In direct contrast to increasing BMI-percentile groups, we found that children with decreasing trajectories were more likely to be from neighborhoods with higher rates of preschool enrollment and bachelor’s education, and lower rates of poverty, reliance on public assistance, single-parent households, and crowded households. This is consistent with other studies that demonstrate the importance of education and SES to weight loss.^98–100 While we did not intend to look at children who can successfully lose weight with this study, our ability to distinguish this group with clustering and our finding that there were SDOH associations in the opposite direction as our groups with obesity are very telling about the important role of SDOH not only to the development of obesity but how successful families might be if provided with more social and environmental resources.

Limitations

Our study is limited in several ways. For linking patients to one census tract, we have used their last available address during the 0–7 years. While relocations were not frequent, a more advanced linkage by tracking all the locations over time could enhance this process. We excluded children with very few observations to ensure our analysis of trajectories was reliable, there may be selection bias since children from households with lower SES may be more likely to move out of the health care system. However, 42% of patients had public insurance in our final dataset, and we were still able to rigorously evaluate the contribution of socioeconomic factors on weight trajectories with our dataset. Our study cohort is very large, spans several US states, and is collected across all settings within the health care system and specialties. While this increases the generalizability of the results, there may be regional differences that we did not account for. In addition, there may be some limitations to not excluding certain populations like those on certain medications that impact weight, those with monogenic obesity, or those seeking weight management treatment. However, for our primary aim of evaluating SDOH on weight trajectories, we felt it was important to be inclusive of most of the population to best understand this question but in future studies would consider a stratified analysis based on these factors. Finally, this analysis did not consider the potential interactions between the different SDOH variables studied. Future studies should employ approaches like multilevel analysis of individual heterogeneity and discriminatory accuracy to better understand these intersectionalities.

Conclusion

Our findings, leveraging a large EHR dataset linked to neighborhood-level SDOHs and clustering, found five groups that differed in how their BMI-percentile classification changed throughout early childhood. Three groups demonstrated high risk of developing obesity and were more likely to be from historically under-resourced neighborhoods. Importantly, these three groups were different from one another in terms of the timing of the development of overweight and obesity which supports the need to develop tailored interventions for both perinatal and early childhood obesity prevention.

Impact Statement

This study demonstrates the association between longitudinal BMI-percentile patterns and SDOH in early childhood. Five distinct clusters with different BMI-percentile trajectories are found, and a strong association between these clusters and SDOH is observed. Our findings highlight the importance of targeted prevention and treatment interventions based on children’s SDOH.

Footnotes

Data Availability Statement

This material is previously posted on medRxiv: doi: .

Authors’ Contributions

M.G. conceived and carried out experiments. T.P. and R.B. conceived the experiments and helped in the study design. M.G., F.L., D.E., and H.B. worked on preprocessing steps including data extraction, cleaning, and linkage. All authors were involved in writing the article and had final approval of the submitted and published versions.

Funding Information

This study was supported by NIH awards, U54GM104941, P20GM103446, and P20GM113125.

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

References

Puder

, Munsch

. Psychological correlates of childhood obesity. Int J Obes (Lond), 2010; 34 (Suppl 2)(2):S37–43; doi: 10.1038/ijo.2010.238

Campbell

. Biological, environmental, and social influences on childhood obesity. Pediatr Res, 2016; 79(1–2):205–211; doi: 10.1038/pr.2015.208

Iguacel

, Gasch-Gallen

, Ayala-Marin

, et al. Social vulnerabilities as risk factor of childhood obesity development and their role in prevention programs. Int J Obes (Lond), 2021; 45(1):1–11; doi: 10.1038/s41366-020-00697-y

Davison

, Birch

. Childhood overweight: A contextual model and recommendations for future research. Obes Rev, 2001; 2(3):159–171; doi: 10.1046/j.1467-789x.2001.00036.x

Dwyer-Lindgren

, Freedman

, Engell

, et al. Prevalence of physical activity and obesity in US counties, 2001–2011: A road map for action. Popul Health Metr, 2013; 11(1):7; doi: 10.1186/1478-7954-11-7

Myers

, Slack

, Martin

, et al. Regional disparities in obesity prevalence in the United States: A spatial regime analysis. Obesity (Silver Spring), 2015; 23(2):481–487; doi: 10.1002/oby.20963

von Hippel

, Benson

. Obesity and the natural environment across us counties. Am J Public Health, 2014; 104(7):1287–1293; doi: 10.2105/AJPH.2013.301838

Hales

, Fryar

, Carroll

, et al. Differences in obesity prevalence by demographic characteristics and urbanization level among adults in the United States, 2013-2016. JAMA, 2018; 319(23):2419–2429; doi: 10.1001/jama.2018.7270

Maharana

, Nsoesie

. Use of deep learning to examine the association of the built environment with prevalence of neighborhood adult obesity. JAMA Netw Open, 2018; 1(4):e181535; doi: 10.1001/jamanetworkopen.2018.1535

10.

Han

, Lawlor

, Kimm

. Childhood obesity. Lancet, 2010; 375(9727):1737–1748; doi: 0.1016/S0140-6736(10)60171-7

11.

Isong

, Rao

, Bind

M-A

, et al. Racial and ethnic disparities in early childhood obesity. Pediatrics, 2018; 141(1):e20170865; doi: 10.1542/peds.2017-0865

12.

Williams

, Ge

, Petroski

, et al. Socioeconomic status and other factors associated with childhood obesity. J Am Board Fam Med, 2018; 31(4):514–521; doi: 10.3122/jabfm.2018.04.170261

13.

Tylavsky

, Ferrara

, Catellier

, et al. Understanding childhood obesity in the US: The NIH environmental influences on child health outcomes (ECHO) program. Int J Obes (Lond), 2020; 44(3):617–627; doi: 10.1038/s41366-019-0470-5

14.

Freedman

. Local food environments: They’re all stocked differently. Am J Community Psychol, 2009; 44(3–4):382–393; doi: 10.1007/s10464-009-9272-6

15.

Kumanyika

. A framework for increasing equity impact in obesity prevention. Am J Public Health, 2019; 109(10):1350–1357; doi: 10.2105/AJPH.2019.305221

16.

Baidal

JAW

, Locks

, Cheng

, et al. Risk factors for childhood obesity in the first 1,000 days: A systematic review. Am J Prev Med, 2016; 50(6):761–779; doi: 10.1016/j.amepre.2015.11.012

17.

Zhou

, Zhang

, Qin

, et al. Age at adiposity rebound and the relevance for obesity: A systematic review and meta-analysis. Int J Obes (Lond), 2022; 46(8):1413–1424; doi: 10.1038/s41366-022-01120-4

18.

Freedman

, Goodwin-Davies

, Kompaniyets

, et al. Interrelationships among age at adiposity rebound, BMI during childhood, and BMI after age 14 years in an electronic health record database. Obesity, 2022; 30(1):201–208; doi: 10.1002/oby.23315

19.

Tester

, Phan

T-LT

, Tucker

, et al. Characteristics of children 2 to 5 years of age with severe obesity. Pediatrics, 2018; 141(3):e20173228; doi: 10.1542/peds.2017-3228

20.

Miech

, Kumanyika

, Stettler

, et al. Trends in the association of poverty with overweight among us adolescents, 1971–2004. JAMA, 2006; 295(20):2385–2393; doi: 10.1001/jama.295.20.2385

21.

Goodman

, Slap

, Huang

. The public health impact of socioeconomic status on adolescent depression and obesity. Am J Public Health, 2003; 93(11):1844–1850; doi: 10.2105/ajph.93.11.1844

22.

Powell

, Wada

, Krauss

, et al. Ethnic disparities in adolescent body mass index in the United States: The role of parental socioeconomic status and economic contextual factors. Soc Sci Med, 2012; 75(3):469–476; doi: 10.1016/j.socscimed.2012.03.019

23.

Ogden

, Fryar

, Hales

, et al. Differences in obesity prevalence by demographics and urbanization in US children and adolescents, 2013-2016. JAMA, 2018; 319(23):2410–2418; doi: 10.1001/jama.2018.5158

24.

Gordon-Larsen

, Adair

, Popkin

. The relationship of ethnicity, socioeconomic factors, and overweight in US adolescents. Obes Res, 2003; 11(1):121–129; doi: 10.1038/oby.2003.20

25.

Goodman

. The role of socioeconomic status gradients in explaining differences in US adolescents’ health. Am J Public Health, 1999; 89(10):1522–1528; doi: 10.2105/ajph.89.10.1522

26.

Yusuf

, Dongarwar

, Yusuf

, et al. Social determinants of overweight and obesity among children in the United States. Int J MCH AIDS, 2020; 9(1):22–33; doi: 10.21106/ijma.337

27.

Hearst

, Martin

, Rafdal

, et al. Early childhood development and obesity risk-factors in a multi-ethnic, low-income community: Feasibility of the ‘five hundred under five’ social determinants of health pilot study. Health Educ. J, 2013; 72(2):203–215; doi: 10.1177/0017896912437553

28.

Hawkins

, Gillman

, Rifas-Shiman

, et al. The linked century study: Linking three decades of clinical and public health data to examine disparities in childhood obesity. BMC Pediatr, 2016; 16(1):32–32; doi: 10.1186/s12887-016-0567-0

29.

Larson

, Russ

, Crall

, et al. Influence of multiple social risks on children’s health. Pediatrics, 2008; 121(2):337–344; doi: 10.1542/peds.2007-0447

30.

Anderson

, Fatima

, Shah

, et al. Income and neighbourhood deprivation in relation to obesity in urban dwelling children 0–12 years of age: A cross-sectional study from 2013 to 2019. J Epidemiol Community Health, 2022; 76(3):274–280; doi: 10.1136/jech-2021-216455

31.

Fyfe-Johnson

, Reid

, Jiang

, et al. Social determinants of health and body mass index in American Indian/Alaska native children. Child Obes, 2022; 19(5):341–352; doi: 10.1089/chi.2022.0012

32.

Cleveland

III , Espinoza

, Holzhausen

, et al. The impact of social determinants of health on obesity and diabetes disparities among Latino communities in Southern California. BMC Public Health, 2023; 23(1):37; doi: 10.1186/s12889-022-14868-1

33.

Peyer

, Welk

, Bailey-Davis

, et al. Relationships between county health rankings and child overweight and obesity prevalence: A serial cross-sectional analysis. BMC Public Health, 2016; 16:404; doi: 10.1186/s12889-016-3091-0

34.

Brochier

, Messmer

, Wexler

, et al. A cross-sectional study of relationships between social risks and prevalence and severity of pediatric chronic conditions. BMC Pediatr, 2023; 23(1):115; doi: 10.1186/s12887-023-03894-6

35.

Wen

, Kleinman

, Gillman

, et al. Childhood body mass index trajectories: Modeling, characterizing, pairwise correlations and socio-demographic predictors of trajectory characteristics. BMC Med Res Methodol, 2012; 12:12–:38; doi: 10.1186/1471-2288-12-38

36.

Baidal

JAW

, Taveras

. Childhood obesity: Shifting the focus to early prevention. Arch Pediatr Adolesc Med, 2012; 166(12):1179–1181; doi: 10.1001/2013.jamapediatrics.358

37.

Mattsson

, Maher

, Boland

, et al. Group-based trajectory modelling for BMI trajectories in childhood: A systematic review. Obes Rev, 2019; 20(7):998–1015; doi: 10.1111/obr.12842

38.

Mottalib

, Jones-Smith

, Sheridan

, et al. Subtyping patients with chronic disease using longitudinal BMI patterns. IEEE J Biomed Health Inform, 2023;PP; doi: 10.1109/JBHI.2023.3237753

39.

Dos Santos

, Picoito

, Nunes

, et al. Early individual and family predictors of weight trajectories from early childhood to adolescence: Results from the millennium cohort study. Front Pediatr, 2020; 8:417; doi: 10.3389/fped.2020.00417

40.

Kwon

, Janz

, Letuchy

, et al. Association between body mass index percentile trajectories in infancy and adiposity in childhood and early adulthood. Obesity (Silver Spring), 2017; 25(1):166–171; doi: 10.1002/oby.21673

41.

Montazeri

, Vrijheid

, Martinez

, et al. Maternal metabolic health parameters during pregnancy in relation to early childhood BMI trajectories. Obesity (Silver Spring), 2018; 26(3):588–596; doi: 10.1002/oby.22095

42.

Rzehak

, Oddy

, Mearin

, et al. WP10 working group of the Early Nutrition Project. Infant feeding and growth trajectory patterns in childhood and body composition in young adulthood. Am J Clin Nutr, 2017; 106(2):568–580; doi: 10.3945/ajcn.116.140962

43.

Garden

, Marks

, Simpson

, et al. Body mass index (BMI) trajectories from birth to 11.5 years: Relation to early life food intake. Nutrients, 2012; 4(10):1382–1398; doi: 10.3390/nu4101382

44.

Chen

T-A

, Baranowski

, Moreno

, et al. Obesity status trajectory groups among elementary school children. BMC Public Health, 2016; 16(1):526; doi: 10.1186/s12889-016-3159-x

45.

Balistreri

, Van Hook

. Trajectories of overweight among US school children: A focus on social and economic characteristics. Matern Child Health J, 2011; 15(5):610–619; doi: 10.1007/s10995-010-0622-7

46.

Nau

, Schwartz

, Bandeen-Roche

, et al. Community socioeconomic deprivation and obesity trajectories in children using electronic health records. Obesity (Silver Spring), 2015; 23(1):207–212; doi: 10.1002/oby.20903

47.

Nummi

, Hakanen

, Lipäinen

, et al. A trajectory analysis of body mass index for Finnish children. J Appl Stat, 2014; 41(7):1422–1435; doi: 10.1080/02664763.2013.872232

48.

Péneau

, Giudici

, Gusto

, et al. Growth trajectories of body mass index during childhood: Associated factors and health outcome at adulthood. J Pediatr, 2017; 186:64–71.e1; doi: 10.1016/j.jpeds.2017.02.010

49.

Slining

, Herring

, Popkin

, et al. Infant BMI trajectories are associated with young adult body composition. J Dev Orig Health Dis, 2013; 4(1):56–68; doi: 10.1017/S2040174412000554

50.

Hernandez

, Marcell

, Garcia

, et al. Predictors of favorable growth patterns during the obesity epidemic among us school children. Clin Pediatr (Phila), 2015; 54(5):458–468; doi: 10.1177/0009922815570579

51.

Jansen

, Mensah

, Nicholson

, et al. Family and neighbourhood socioeconomic inequalities in childhood trajectories of BMI and overweight: Longitudinal study of Australian children. PLoS One, 2013; 8(7):e69676; doi: 10.1371/annotation/f7e5e1f3-77f6-4c56-b0ba-53b54a86df14

52.

Forrest

, Margolis

, Bailey

, et al. PEDSnet: A National Pediatric Learning Health System. J Am Med Inform Assoc, 2014; 21(4):602–606; doi: 10.1136/amiajnl-2014-002743

53.

WHO, W. M. G. R. S. G. WHO child growth standards based on length/height, weight and age. Acta Paediatr Suppl, 2006; 450:76–85; doi: 10.1111/j.1651-2227.2006.tb02378.x

54.

Centers for Disease Control and Prevention. Childhood overweight and obesity. 2021.

55.

Mattsson

, Murray

, Hawkes

, et al. Body mass index trajectories in the first 5 years and associated antenatal factors. Front Pediatr, 2021; 9:622381; doi: 10.3389/fped.2021.622381

56.

Khalsa

, Li

, Rausch

, et al. Early childhood growth trajectories in a Medicaid population. Pediatr Obes, 2022; 17(9):e12918; doi: 10.1111/ijpo.12918

57.

Giles

, Whitrow

, Davies

, et al. Growth trajectories in early childhood, their relationship with antenatal and postnatal factors, and development of obesity by age 9 years: Results from an Australian birth cohort study. Int J Obes (Lond), 2015; 39(7):1049–1056; doi: 10.1038/ijo.2015.42

58.

Ziyab

, Karmaus

, Kurukulaaratchy

, et al. Developmental trajectories of body mass index from infancy to 18 years of age: Prenatal determinants and health consequences. J Epidemiol Community Health, 2014; 68(10):934–941; doi: 10.1136/jech-2014-203808

59.

U.S. Census Bureau. Census geocoder documentation. 2018.

60.

Data for a diverse and equitable future. Child opportunity index (coi). 2021.

61.

Agency for Toxic Substances and Disease Registry. Cdc/atsdr svi fact sheet. 2021.

62.

Robert Graham Center. Social deprivation index (sdi). 2021.

63.

Moritz

, Bartz-Beielstein

. Imputets: Time series missing value imputation in R. R J, 2017; 9(1):207.

64.

Jung

, Wickrama

. An introduction to latent class growth analysis and growth mixture modeling. Social & Personality Psych, 2008; 2(1):302–317; doi: 10.1111/j.1751-9004.2007.00054.x

65.

Nagin

, Tremblay

. Analyzing developmental trajectories of distinct but related behaviors: A group-based method. Psychol Methods, 2001; 6(1):18–34; doi: 10.1037/1082-989x.6.1.18

66.

Nagin

. Analyzing developmental trajectories: A semiparametric, group-based approach. Psychol Methods, 1999; 4(2):139–157; doi: 10.1037/1082-989X.4.2.139

67.

Muthén

, Muthén

. Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcohol Clin Exp Res, 2000; 24(6):882–891; doi: 10.1111/j.1530-0277.2000.tb02070.x

68.

Liu

, Liu

, Frongillo

, et al. Body mass index trajectories during infancy and pediatric obesity at 6 years. Ann Epidemiol, 2017; 27(11):708–715.e1; doi: 10.1016/j.annepidem.2017.10.008

69.

Stout

, Espel

, Sandman

, et al. Fetal programming of children’s obesity risk. Psychoneuroendocrinology, 2015; 53:29–39; doi: 10.1016/j.psyneuen.2014.12.009

70.

Koning

, Hoekstra

, Jong

E. D

, et al. Identifying developmental trajectories of body mass index in childhood using latent class growth (mixture) modelling: Associations with dietary, sedentary and physical activity behaviors: A longitudinal study. BMC Public Health, 2016; 16(1):1128; doi: 10.1186/s12889-016-3757-7

71.

Liu

, Shelton

, Eldred-Skemp

, et al. Early exposure to cumulative social risk and trajectories of body mass index in childhood. Child Obes, 2019; 15(1):48–55; doi: 10.1089/chi.2018.0116

72.

Daniels

, Lê-Scherban

, Auchincloss

, et al. Longitudinal associations of neighborhood environment features with pediatric body mass index. Health Place, 2021; 71:102656; doi: 10.1016/j.healthplace.2021.102656

73.

Daniels

, Schinasi

, Auchincloss

, et al. The built and social neighborhood environment and child obesity: A systematic review of longitudinal studies. Prev Med, 2021; 153:106790; doi: 10.1016/j.ypmed.2021.106790

74.

Singh

, Siahpush

, Kogan

. Rising social inequalities in us childhood obesity, 2003–2007. Ann Epidemiol, 2010; 20(1):40–52; doi: 10.1016/j.annepidem.2009.09.008

75.

Miranda

, Edwards

, Anthopolos

, et al. The built environment and childhood obesity in Durham, North Carolina. Clin Pediatr (Phila), 2012; 51(8):750–758; doi: 10.1177/0009922812446010

76.

Evans

, Fuller-Rowell

, Doan

. Childhood cumulative risk and obesity: The mediating role of self-regulatory ability. Pediatrics, 2012; 129(1):e68–73–e73; doi: 10.1542/peds.2010-3647

77.

Gundersen

, Mahatmya

, Garasky

, et al. Linking psychosocial stressors and childhood obesity. Obes Rev, 2011; 12(5):e54–63; doi: 10.1111/j.1467-789X.2010.00813.x

78.

Dixon

, Peña

M-M

, Taveras

. Lifecourse approach to racial/ethnic disparities in childhood obesity. Adv Nutr, 2012; 3(1):73–82; doi: 10.3945/an.111.000919

79.

Kumanyika

. Environmental influences on childhood obesity: Ethnic and cultural influences in context. Physiol Behav, 2008; 94(1):61–70; doi: 10.1016/j.physbeh.2007.11.019

80.

, Goran

, Kaur

, et al. Developmental trajectories of overweight during childhood: Role of early life factors. Obesity (Silver Spring), 2007; 15(3):760–771; doi: 10.1038/oby.2007.585

81.

Pryor

, Tremblay

, Boivin

, et al. Developmental trajectories of body mass index in early childhood and their risk factors: An 8-year longitudinal study. Arch Pediatr Adolesc Med, 2011; 165(10):906–912; doi: 10.1001/archpediatrics.2011.153

82.

Eagle

, Sheetz

, Gurm

, et al. Understanding childhood obesity in america: Linkages between household income, community resources, and children’s behaviors. Am Heart J, 2012; 163(5):836–843; doi: 10.1016/j.ahj.2012.02.025

83.

Huffman

, Kanikireddy

, Patel

. Parenthood—a contributing factor to childhood obesity. Int J Environ Res Public Health, 2010; 7(7):2800–2810; doi: 10.3390/ijerph7072800

84.

Chen

, Escarce

. Family structure and childhood obesity, early childhood longitudinal study—kindergarten cohort. Prev Chronic Dis, 2010; 7(3):A50.

85.

Rossen

. Neighbourhood economic deprivation explains racial/ethnic disparities in overweight and obesity among children and adolescents in the USA. J Epidemiol Community Health, 2014; 68(2):123–129; doi: 10.1136/jech-2012-202245

86.

Grow

HMG

, Cook

, Arterburn

, et al. Child obesity associated with social disadvantage of children’s neighborhoods. Soc Sci Med, 2010; 71(3):584–591; doi: 10.1016/j.socscimed.2010.04.018

87.

Mohammed

, Habtewold

, Birhanu

, et al. Neighbourhood socioeconomic status and overweight/obesity: A systematic review and meta-analysis of epidemiological studies. BMJ Open, 2019; 9(11):e028238; doi: 10.1136/bmjopen-2018-028238

88.

Gable

, Lutz

. Household, parent, and child contributions to childhood obesity. Fam Relat, 2000; 49(3):293–300; doi: 10.1111/j.1741-3729.2000.00293.x

89.

Roux

AVD

. Neighborhoods and health: What do we know? What should we do? Am J Public Health, 2016; 106(3):430–431; doi: 10.2105/AJPH.2016.303064

90.

Kirby

, Kaneda

. Neighborhood socioeconomic disadvantage and access to health care. J Health Soc Behav, 2005; 46(1):15–31; doi: 10.1177/002214650504600103

91.

Kim

, Kawachi

. Perceived neighborhood social cohesion and preventive healthcare use. Am J Prev Med, 2017; 53(2):e35–e40; doi: 10.1016/j.amepre.2017.01.007

92.

Latham

, Williams

. Does neighborhood disorder predict recovery from mobility limitation? findings from the health and retirement study. J Aging Health, 2015; 27(8):1415–1442; doi: 10.1177/0898264315584328

93.

Ross

, Mirowsky

. Neighborhood disorder, subjective alienation, and distress. J Health Soc Behav, 2009; 50(1):49–64; doi: 10.1177/002214650905000104

94.

Booth

, Teixeira

, Zuberi

, et al. Barrios, ghettos, and residential racial composition: Examining the racial makeup of neighborhood profiles and their relationship to self-rated health. Soc Sci Res, 2018; 69:19–33; doi: 10.1016/j.ssresearch.2017.10.002

95.

Schroeder

, McCormick

, Perez

, et al. The role and impact of community health workers in childhood obesity interventions: A systematic review and meta-analysis. Obes Rev, 2018; 19(10):1371–1384; doi: 10.1111/obr.12714

96.

Skelton

, Palakshappa

, Moore

, et al. Community engagement and pediatric obesity: Incorporating social determinants of health into treatment. J Clin Transl Sci, 2020; 4(4):279–285; doi: 10.1017/cts.2019.447

97.

Soltero

, O’Connor

, Thompson

, et al. Opportunities to address obesity disparities among high-risk Latino children and adolescents. Curr Obes Rep, 2021; 10(3):332–341; doi: 10.1007/s13679-021-00445-x

98.

Cadzow

, Chambers

, Sandell

. School-based obesity intervention associated with three year decrease in student weight status in a low-income school district. J Community Health, 2015; 40(4):709–713; doi: 0.1007/s10900-015-9989-0

99.

Carroll-Scott

, Gilstad-Hayden

, Rosenthal

, et al. Associations of neighborhood and school socioeconomic and social contexts with body mass index among urban preadolescent students. Am J Public Health, 2015; 105(12):2496–2502; doi: 10.2105/AJPH.2015.302882

100.

Vazquez

, McBride

, Hess

, et al. Examining school and neighborhood effects of socioeconomic status on childhood obesity in the US. Int J Environ Res Public Health, 2022; 19(10):5831; doi: 10.3390/ijerph19105831

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.96 MB