The Longitudinal Effect of Area Socioeconomic Changes on Obesity: a Longitudinal Cohort Study in the USA from 2003 to 2017

Abstract 

Despite several dimensions of area socioeconomic status (SES), past literature has been dominated by the use of area socioeconomic position. We examined the longitudinal effect of three area SES measures (i.e., socioeconomic position, inequality, and segregation) on obesity. Using longitudinal data from the Fragile Families & Child Wellbeing Study (N = 1493), we estimated a linear mixed model to examine the effect of three time-varying area SES measures on time-varying measures of objectively measured body mass index z-score (BMIz) from ages 5 years to 15 years. Findings showed that BMIz increased steadily over time (B = 0.02, 95% CI = 0.02, 0.03). A significant interaction between time and area socioeconomic position indicates that children in areas with higher socioeconomic position had a smaller increase in BMIz than those in low socioeconomic areas (B =  − 0.02, 95% CI =  − 0.02, − 0.01). A non-linear relationship of area income inequality with BMIz such that BMIz was higher as area income inequality was greater, but the effect diminishes in magnitude with a higher level of area income inequality (linear term: B = 0.07; quadratic term: B =  − 0.03). Area income segregation was associated with greater BMIz (B = 0.08, 95% CI = 0.03, 0.12). No time interaction effect was found for area income inequality and segregation. Results highlight a need for community health policy efforts and evidence-based interventions to address childhood obesity issues in low-SES areas.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11524-022-00681-z.

Introduction 

Childhood obesity is a multisystem condition associated with poor physical, psychological, and behavioral outcomes over the life course [1–4]. In the United States (US), childhood obesity prevalence has increased over the last four decades. The percentage of children ages 6–11 who had a body mass index at 95th percentile or above increased from 4% in 1971–1974 to 19% in 2015–2018 [5, 6]. Furthermore, a more rapid increase was observed in children from racial/ethnic minority and lower socioeconomic status households [7, 8], which results in rising social disparities in childhood obesity.

There is a growing recognition that residential socioeconomic environments play an important role in child health [9–11]. Area socioeconomic status (SES) has several dimensions: socioeconomic position, inequality, and segregation [12, 13]. While area socioeconomic position is operationalized as an absolute level of socioeconomic resources (e.g., census tract-level poverty rate and unemployment rate), area socioeconomic inequality indicates the distribution of socioeconomic resources among individuals within the residential area, such as income inequality. Area socioeconomic segregation represents a geographic difference in socioeconomic resources between a residential area and a larger geographic unit (e.g., the extent to which census tract-level income distributions differ from county-level overall income distributions) [14]. Despite the presence of different dimensions of area SES, past literature has been dominated by area socioeconomic position. To our knowledge, no study has investigated area socioeconomic inequality and segregation as a factor of childhood obesity. The present study extends previous research by examining the longitudinal effects of area socioeconomic position, inequality, and segregation on childhood obesity. While we measured area socioeconomic position at the census tract level, area income inequality and segregation were measured at the county level because social comparison is more likely to occur in areas large enough to contain social heterogeneity [15].

Area socioeconomic position has been investigated as an important factor of childhood obesity in the past two decades [16–32]. According to the neighborhood institutional resource model [33], residence in socioeconomically advantaged areas reduces the risk of child health problems through health-promoting support and resources such as social support, safety, a low level of physical hazards, and convenient access to healthy food options, parks, after-school programs, and community centers. A systematic review [34] identified four longitudinal studies and sixteen cross-sectional studies examining the association between area economic position and childhood obesity independent of family-level economic status. The review [34] showed that three of four longitudinal studies and ten of sixteen cross-sectional studies found a significant association between area economic position and childhood obesity.

Despite a burgeoning number of studies on the association between area socioeconomic position and childhood obesity, most studies have been cross-sectional [19–31] or short-term longitudinal designs [16–18], limiting the ability to understand the long-lasting effect of area socioeconomic position on obesity from early childhood to adolescence. Furthermore, little is known about whether area socioeconomic position impacts the trajectory of childhood obesity, which can be used as evidence to target children in socioeconomically disadvantaged neighborhoods for mitigating obesity. One longitudinal study conducted an interaction test between child age and area socioeconomic position and reported a stronger impact of area socioeconomic position on obesity for children aged 11–18 than those aged 2–10 [32]; however, to our best knowledge, no study has examined whether area socioeconomic position can change the trajectory of childhood obesity.

Although area socioeconomic inequality has been described as critical to health [35], there are theoretical inconsistencies in how area socioeconomic inequality affects health [33]. On one hand, area socioeconomic inequality is hypothesized to deteriorate health through greater psychological distress and lower levels of social trust, social support, and investments in social infrastructure [35–39]. On the other hand, area socioeconomic inequality may promote the health of low-income residents as a result of sharing the community with high-income residents. On average, higher-income residents attract health-promoting infrastructure and engage in healthier behaviors, which ultimately promotes low-income residents’ health [40]. This idea is aligned with the rationale of mixed-income housing developments [40, 41]. Empirical research shows a mix of positive, negative, and no associations between area income inequality and adult health [41–44] and child health behavior [45]. To date, no study has examined the effect of area socioeconomic inequality on childhood obesity, independent of family-level economic status.

While area socioeconomic inequality examines the degree of socioeconomic dispersion within an individual’s residential area in an aspatial manner, area socioeconomic segregation accounts for spatial patterning of income dispersion and examines whether low socioeconomic households are spatially concentrated in a given area of the county. Historical structural discrimination, racism, and housing market discrimination are related to area socioeconomic segregation [13, 46, 47]. A high level of socioeconomic segregation within the county is likely to lead to health-compromising environmental conditions, such as high crime rates [48, 49], which increase psychological stress and health-compromising behaviors [33]. Thus, in low-income segregated areas, low-income individuals may have adverse health outcomes not only resulting from their low economic status but also from their neighbors’ relatively lower economic status compared to that of surrounding neighborhoods [13]. Nowadays, an examination of area socioeconomic segregation is especially important as gentrification has recently increased a socioeconomic divide between gentrified and ungentrified areas; however, it has not been yet examined in childhood obesity literature.

In response to the need to explore contextual characteristics that may be targeted in childhood obesity prevention policies, this study aims to examine the longitudinal effect of area socioeconomic position at the census tract level, income inequality at the county level, and income segregation at the county level on childhood obesity over time. We explore the effect of three area SES measures on childhood obesity and if its impact changes the trajectory of childhood obesity over time.

Methods

Sample

This study used the data from the Fragile Families & Child Wellbeing Study (FFCWS). The FFCWS is a national, longitudinal study of children born in 20 US cities with 200,000 or more residents [50]. Baseline data were collected when the index child was born (1998–2000), and follow-up data collections continued when the child was 1 (1999–2001), 3 (2001–2003), 5 (2003–2006), 9 (2007–2010), and 15 years old (2014–2017). We used publicly available FFCWS data files from ages 3 through 15 and merged the data with restricted contextual data files containing area-level SES measures for each respondent. For restricted contextual data files, the FFCWS team geocoded residential addresses of mother and fathers when the child was age three, five, and nine using the 2000 geographical boundary definitions. Area SES measures were derived from the 2000 Decennial Census and federal income tax records covering the US population from 1996 to 2012 and corresponded to the residence of the biological mother or father living with the index child at the time of each data collection.

Among 3299 respondents who participated in the data collection when children were at age 3, we excluded 149 respondents who did not participate in any of the data collection at ages 5, 9, and 15. In addition, as our study focuses on child body mass index (BMI), we excluded respondents who did not participate in any of in-home examinations of child BMI at ages 5, 9, and 15 (n = 522). We also excluded those missing data for area SES measures (n = 401), household income-to-needs ratio (n = 25), and objectively measured child BMI at age 3 (n = 709), which results in 1493 respondents in our analytic sample. Compared with those who remained in the study, those who dropped out were more likely to be Hispanic and have a lower level of parental highest education. There were no significant differences in child sex, household income-to-needs ratio, area SES, and child BMI. This study was approved by the institutional review board of the University of Texas at Arlington.

Measures

Anthropometric Measures

Weight and height were measured by the interviewer at in-home visits at ages 3, 5, 9, and 15. The interviewer took two measurements of height and weight, and if a difference between the first two measurements was 2 or more centimeters (for height) and 2 or more pounds (for weight), a third measurement was taken. BMI was calculated based on measured weight (in kilograms) divided by measured height (in meters) squared and was then standardized for sex and age to calculate BMI z-score (BMIz). We used BMIz at ages 5, 9, and 15 as an outcome and BMIz at age 3 as a covariate. Of note, while all the children in the FFCWS were eligible for in-home anthropometric examinations at ages 3, 5, and 9, an in-home anthropometric examination at age 15 was conducted among a random sub-sample of eligible families due to the FFCWS’s limited budget. Of 3580 respondents at the 15-year study, an in-home examination was conducted only for 1090 children (30%). There were no significant differences in child BMIz at ages 3, 5, and 9 between missing and non-missing data of BMIz at age 15.

Area Socioeconomic Position

Area socioeconomic position at the census tract level was defined as a weighted average of socioeconomic characteristics in a census tract where the index child lived at the time of each data collection. We calculated area socioeconomic position by averaging standard scores of five census tract-level measures for each time point of data collection, including median household income, median value of housing units, the percentage of adults 25 years of age or older who had completed high school, the percentage of adults 25 years of age or older who had completed college, and the percentage of employed persons 16 years of age or older [51]. Area socioeconomic position was standardized within the FFCWS sample, with greater values of area socioeconomic position indicating more advantaged socioeconomic conditions at the census tract level.

Area Income Inequality

Area income inequality at the county level was calculated by Cherry and Hendren [52, 53]. Cherry and Hendren [52, 53] estimated the county-level Gini coefficient using data from federal income tax records for more than 40 million children and their parents between 1996 and 2012. The Gini coefficient measures the degree of income dispersion among households with children residing in the residential area and theoretically ranges between 0 (perfect equality) and 1 (perfect inequality). The Gini coefficients were standardized within the FFCWS sample, with greater values indicating greater income inequality at the county level.

Area Income Segregation

Area income segregation at the county level was calculated by Chetty and Hendren [52, 53]. Chetty and Hendren [52, 53] estimated the degree to which individuals below the xth percentile of the local household income distribution are segregated from those above the xth percentile across census tracts within each county, which is defined as a two-group segregation index. Then, the two-group segregation index was averaged across percentiles of local household income distribution. A high level of area income segregation indicates that low-income families are more likely to live in the same census tract within each county. Area income segregation was standardized within the FFCWS sample, with greater values indicating greater income segregation at the county level.

Covariates

Covariates are child sex, child-reported race/ethnicity, parental highest education, and household income-to-needs ratio. Child sex was obtained from the baseline assessment, child-reported race/ethnicity was obtained from the 15-year assessment, and parental highest education was obtained from the 3-year assessment. Household income-to-needs ratio was obtained from the 3, 5, and 9-year assessments, and it was considered time dependent to account for changes in family economic status over time.

Analysis

A linear mixed model was estimated with area SES measures (t-1) as primary exposures and BMIz (t) as the outcome in order to capture the longitudinal effect of area SES on child BMIz. A linear mixed model is fitted using maximum likelihood methods, which handle missing outcome data at a few time-points [54, 55]. Initially, we estimated the average trajectory of child BMIz with time (age in years) as a predictor. Within the first model, we assessed the correlation of variance of observations within an individual and different functional forms of time. The second model included individual- and family-level covariates. We also tested the influence of sex, race/ethnicity, parent highest education, and household income-to-needs ratio on the trajectories of BMIz over time (i.e., sex by time interaction, race/ethnicity by time interaction, parent highest education by time interaction, and household income-to-needs ratio by time interaction) and removed nonsignificant interaction terms. The third model included area SES measures one at a time while adjusting for the individual- and family-level covariates. Area SES measures were time dependent to account for respondents’ residential moves during the study period. Also, we accounted for change in area socioeconomic position, inequality, and segregation over time by using time-dependent measures. Within the third model, we initially included higher-ordered (up to cubic) forms of each area SES measure to capture potential non-linearity between area SES measures and child BMIz, and we removed nonsignificant terms. In the fourth model, to examine the effect of area SES measures on child BMIz over time, area SES-by-time interaction variables (i.e., area socioeconomic position by time interaction, area income inequality by time interaction, and area income segregation by time interaction) were added to the third model one at a time because we dropped out a significant percentage of respondents with missing data, as a supplementary analysis, we performed multiple imputation techniques to classify missing data in exposures and covariates and repeated the linear mixed modeling. Throughout the model fitting process, log likelihood, the Akaike information criterion (AIC), and the Bayesian information criterion (BIC) were assessed to provide evidence for the fit of the model, with higher log-likelihood value and lower AIC and BIC values indicating a better model fit [56]. We did not conduct a multilevel analysis at the census tract level and at the county level due to very low intraclass correlation coefficients (ICC) for child BMIz (ICCs < 0.01) [57]. Analyses were conducted using StataSE 16 software program.

Results

As shown in Table 1, the sample consisted of 1493 children, including 772 boys (52%) and 721 girls (48%). Forty-five percent of children self-identified their race/ethnicity as non-Hispanic Black/African American only and 20% self-identified as Hispanic. About 40% of children had parents with less than high school educations, and 13% of children were from families in poverty at age 3.

Table 1.

Sample characteristics (N = 1493) 

Characteristics	N (%)
Child sex
Boys	772 (51.7%)
Girls	721 (48.3%)
Child race/ethnicity
Non-Hispanic White only	261 (17.5%)
Non-Hispanic Black/African American only	675 (45.2%)
Hispanic	298 (20.0%)
Non-Hispanic other	100 (6.7%)
Missing	159 (10.7%)
Parental highest education level at year 3
Less than high school	196 (13.1%)
High school graduate and equivalent	422 (28.3%)
Some college	611 (40.9%)
College graduate and above	264 (17.7%)
Household income-to-needs ratio at year 3
0–99%	593 (39.7%)
100–199%	368 (24.7%)
200–299%	215 (14.4%)
300% and above	317 (21.2%)
Child body mass index z-score at year 3 (mean ± standard deviation)	0.53 ± 1.22
Child body mass index z-score at year 5 (mean ± standard deviation)	0.59 ± 1.10
Child body mass index z-score at year 9 (mean ± standard deviation)	0.70 ± 1.07
Child body mass index z-score at year 15 (mean ± standard deviation)	0.86 ± 1.01

Child race/ethnicity was self-reported at year 15

Table 2 shows results of linear mixed modeling. In the time-only model, the overall average predicted BMIz at age 5 was 0.61 (95% CI = 0.56, 0.67) and predicted BMIz increased by 0.02 (95% CI = 0.02, 0.03) every year. With the inclusion of individual- and family-level characteristics, boys had a lower BMIz than girls (B =  − 0.09, 95% CI =  − 0.17, − 0.01), and non-Hispanic Black children had higher BMIz than non-Hispanic White children (B = 0.16, 95% CI = 0.03, 0.30). With an increase of 1 unit in child BMIz at age 3, there was a 0.46 (95% CI = 0.42, 0.51) increase in BMIz. In addition, we observed a significant interaction effect of child BMIz at age 3 and time (B =  − 0.02, 95% CI =  − 0.02, − 0.01), indicating that the trajectory of BMIz from ages 5 through 15 differs by the level of BMIz at age 3. In other words, children with high BMIz at age 3 tend to have higher BMIz at age 5 than those with lower BMIz at age 3, but their increase in BMIz from ages 5 through 15 was less rapid than those with lower BMIz at age 3.

Table 2.

Effect estimates of area socioeconomic status on trajectories of child body mass index z-score (N = 1493)

Variable	Time-only model		Individual-level model		Area socioeconomic position model		Area income inequality model		Area income segregation model
	B	95% CI	B	95% CI	B	95% CI	B	95% CI	B	95% CI
Intercept (at age 5)	0.61	0.56, 0.67	0.37	0.20, 0.55	0.37	0.19, 0.55	0.42	0.24, 0.60	0.39	0.21, 0.57
Individual effects
Sex
Boys			− 0.09	− 0.17, − 0.01	− 0.09	− 0.17, − 0.01	− 0.08	− 0.17, 0.01	− 0.09	− 0.17, − 0.01
Girls			…		…		…		…
Race/ethnicity
Non-Hispanic White only			…		…		…		…
Non-Hispanic Black only			0.16	0.03, 0.30	0.16	0.03, 0.30	0.11	− 0.02, 0.25	0.13	− 0.01, 0.27
Hispanic			0.15	− 0.01, 0.30	0.15	− 0.01, 0.30	0.11	− 0.05, 0.26	0.12	− 0.03, 0.27
Non-Hispanic other			0.10	− 0.09, 0.30	0.11	− 0.09, 0.30	0.09	− 0.11, 0.28	0.10	− 0.10, 0.29
Missing			0.05	− 0.12, 0.23	0.05	− 0.12, 0.23	0.01	− 0.16, 0.19	0.03	− 0.14, 0.21
Parent highest education
Less than high school			…		…		…		…
High school or equivalent			− 0.04	− 0.18, 0.11	− 0.04	− 0.18, 0.11	− 0.03	− 0.18, 0.11	− 0.03	− 0.1, 0.11
Some college			− 0.08	− 0.22, 0.07	− 0.08	− 0.22, 0.07	− 0.06	− 0.20, 0.08	− 0.07	− 0.21, 0.07
College or above			− 0.18	-0.36, 0.00	− 0.18	− 0.36, 0.00	− 0.15	− 0.33, 0.03	− 0.17	− 0.35, 0.01
Household income to needs ratio
< 100%			…		…		…		…
100–199%			0.03	− 0.04, 0.11	0.03	− 0.05, 0.11	0.03	− 0.05, 0.11	0.03	− 0.05, 0.10
200–299%			0.04	− 0.06, 0.13	0.04	− 0.06, 0.13	0.04	− 0.06, 0.14	0.03	− 0.06, 0.13
300% and above			− 0.07	− 0.17, 0.04	− 0.07	− 0.17, 0.04	-0.06	− 0.16, 0.04	− 0.06	− 0.17, 0.04
Child BMIz at age 3			0.46	0.42, 0.51	0.47	0.42, 0.51	0.46	0.42, 0.51	0.46	0.42, 0.50
Area socioeconomic effects
Area socioeconomic position					0.03	− 0.02, 0.08
Area income inequality							0.07	0.03, 0.12
Area income inequality2 (quadratic term)							− 0.03	− 0.05, − 0.01
Area income segregation									0.08	0.03, 0.12
Time effects
Time (years since age 5)	0.02	0.02, 0.03	0.03	0.02, 0.04	0.03	0.02, 0.04	0.03	0.02, 0.04	0.03	0.02, 0.04
Child BMIz at age 3 × time			− 0.02	− 0.02, − 0.01	− 0.02	− 0.02, − 0.01	-0.02	− 0.02, − 0.01	− 0.02	− 0.02, − 0.01
Area socioeconomic position × time					− 0.02	− 0.02, − 0.01
Model fit statistics
Log likelihood	− 4102.74		− 3849.67		− 3847.23		− 3841.10		− 3843.84
Akaike information criterion	8217.47		7737.33		7736.47		7724.19		7727.67
Bayesian information criterion	8253.72		7852.12		7863.34		7851.06		7848.50

CI indicates confidence interval. BMIz indicates body mass index z-score

In the area socioeconomic position model, we added the measure of area socioeconomic position to the individual-level model. We found no significant association between area socioeconomic position and BMIz (B = 0.03, 95% CI =  − 0.02, 0.08); however, the interaction between time and area socioeconomic position was statistically significant (B =  − 0.02, 95% CI =  − 0.02, − 0.01; see Fig. 1), which indicates that area socioeconomic position impacts the trajectory of BMIz from ages 5 through 15. Children in high socioeconomic areas had a smaller increase in BMIz over time compared to those in lower socioeconomic areas.

Fig. 1 .

Predicted trajectories of child body mass index z-score over time, as shown in Table 2, area socioeconomic position model, setting all other variables within the model at their average value

In the area income inequality model, we added the measure of area income inequality to the individual-level model. Linear and quadratic terms of area income inequality were significantly associated with BMIz (linear term: B = 0.07, 95% CI = 0.03, 0.12; quadratic term: =  − 0.03, 95% CI =  − 0.05, − 0.01), which indicates non-linearity. In other words, child BMIz was higher as area income inequality is greater, but this effect diminished in magnitude with a higher level of area income inequality (see Fig. 2). When we included an interaction term between area income inequality and time, there was no significant interaction (results not shown).

Fig. 2.

Predicted child body mass index z-score at age 5 by area income inequality and income segregation, as shown in Table 2, area income inequality model, setting all other variables within the model at their average value

In the area income segregation model, we added the measure of area income segregation to the individual-level model. We observed that child BMIz at age 5 increased by 0.08 (95% CI = 0.03, 0.12) with an increase of 1 unit in area income segregation (see Fig. 2). With the inclusion of an interaction term between area income segregation and time, there was no significant interaction (results not shown). As a sensitivity analysis, we re-analyzed linear mixed models after multiple imputation and found similar results (see Supplementary Table 1).

Discussion

The present study examined the effect of area socioeconomic position, income inequality, and income segregation on BMIz from early childhood to adolescence in the US. Our findings provided evidence of overall increase in child BMIz over time. For example, in the time-only model, BMIz increased 0.20 units over a 10-year period (= 0.02 × 10). Regarding the effect of area socioeconomic position on BMIz, children living in low socioeconomic areas showed a greater increase in BMIz over time in comparison to those in higher socioeconomic areas. For example, over a 10-year period, children in low socioeconomic areas (i.e., 2 standard deviation (SD) below mean) had a BMIz increase by 0.50 (= [0.03 +  − 0.01 ×  − 2] × 10), whereas those in high socioeconomic areas (i.e., 2 SD above mean) had a BMIz increase by 0.10 (= [0.03 +  − 0.01 × 2] × 10) (see Fig. 1). The BMIz difference is clinically meaningful given a difference of 0.15 BMIz units is generally considered as clinically significant although there is no clear cut off [58, 59]. In addition, we found that higher area income inequality and segregation were significantly associated with greater BMIz as a main effect, indicating that those in more unequal or segregated areas had greater BMIz at age 5 (see Fig. 2). Specifically, children living in areas with high-income inequality (i.e., 2 SD above mean) had 0.12 units higher BMIz (= [0.05 × 4] + [− 0.02 × 4]) at age 5 than those in areas with low-income inequality (i.e., 2 SD below mean). Children in areas with high-income segregation (i.e., 2 SD above mean) had 0.16 units higher BMIz (= 0.04 × 4) at age 5 than those in areas with low-income segregation (i.e., 2 SD below mean), which could be considered as clinically meaningful. Our findings are consistent with past studies reporting a significant cross-sectional or short-term longitudinal association between area socioeconomic position and childhood obesity [16, 18, 21–23, 25–27, 31, 32] and reporting a significant association between area income inequality and child health behaviors [45] and adult health [41, 43, 44], independent of individual demographic characteristics and family SES. Our findings indicate that there is the need for health promotion policies and programs in low-SES areas to advance contextual equity and subsequently child health equity.

Our findings showed that the association between area SES and child BMIz differs by measures of area SES. Specifically, while area socioeconomic position had a significant interaction with time, area income inequality and segregation had only a direct effect on BMIz without an interaction with time, which indicates that area income inequality and segregation impacts BMIz at the baseline (at age 5) but not the trajectory of BMIz over time. The underpinning of our observed differences in the effect of three area SES measures on BMIz is not clear. An explanation is that area income inequality and segregation capture a degree to which residents feel relatively deprived or disadvantaged in comparison with neighbors or surrounding neighborhoods, which might influence parental psychological distress and parenting behaviors in early childhood and subsequently childhood obesity [33, 35–37]. The explanation is aligned with past literature reporting that those in early childhood are impacted indirectly by neighborhood environments as young children interact with their neighborhoods indirectly through their parents [60]. On the other hand, the absolute level of neighborhood socioeconomic resources may be impactful from early childhood through adolescence as children interact directly with their neighborhood contexts and experienced disadvantaged socioeconomic circumstances in their neighborhood [60]. Future research is needed to test a mediating mechanism by which different area SES measures impact childhood obesity, which helps policymakers and practitioners further understand the impact of area SES on childhood obesity and how best to address them. Also, further investigation on the effect of developmental stage on the area SES-obesity association may inform policymakers and practitioners to consider disparate childhood obesity prevention programs by developmental stage. For example, for early childhood, health policies and programs may need to address factors associated with income inequality and segregation, such as relative deprivation and distress, whereas for middle childhood and adolescence, they may need to focus on improving health-promoting resources in areas with low socioeconomic position.

Limitations

This study has several limitations. First, despite being a national study, the FFCWS is not a nationally representative sample of US children so it is unclear whether these findings can be generalized to other populations in the US. Second, as mentioned above, this study did not test a mediating mechanism potentially underlying the effects of area SES on childhood obesity. Further research is warranted to test potential mediators such as physical activity, dietary habits, sleep, parenting skills, parental stress, and built and social environments. Third, we used geographically defined administrative boundaries (i.e., counties and census tracts) to characterize area SES measures, which can result in spatial misclassification [61]. Further research is needed to examine whether the effect of area SES on child BMIz is different depending on neighborhood definitions (e.g., administrative units, home-based spatial buffer, Global Positioning Systems daily path buffers). Also, we obtained data from 2000 Decennial Census for area socioeconomic position and area income inequality. Also, data from federal income tax records in 1996–2012 were obtained for area income inequality. The difference in data collection periods between area SES measures might result in heterogeneity of the results depending on area SES measures. Lastly, due to the FFCWS’s limited budget, only a random sub-sample of FFCWS respondents were eligible for an in-home anthropometric examination at age 15. Although linear mixed modeling could handle missing outcome at a few time-points using maximum likelihood methods, missing data may potentially bias the results.

Conclusions

This study’s findings showed the significant effect of area socioeconomic position, income inequality, and income segregation on BMIz (at baseline or its trajectory over time). Our results suggest a need for support of low-SES areas in childhood obesity prevention programs. Also, the results highlight that each dimension of area SES needs to be considered to better understand the impact of area SES on child BMIz. Future research is needed to examine a mediating mechanism by which different area SES measures impact childhood obesity, which informs policymakers and practitioners of best practices for developing community-based childhood obesity prevention programs.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This research project is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R03HD101752. The original FFCWS was supported by Award Numbers R25HD074544, P2CHD058486, and R01HD036916 awarded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.