Evaluating the Contribution of the Built Environment on Obesity Among New York State Students

Abstract

Objectives. One third of school-aged children in New York State (NYS) are overweight or obese, with large geographic disparities across local regions. We used NYS student obesity surveillance data to assess whether these geographical variations are attributable to the built environment. Method. We combined NYS Student Weight Status Category Reporting System 2010-2012 data with other government publicly available data. Ordinary least squares regression models identified key determinants of school district–level student obesity rates for elementary and middle/high schools. Geographical weighted regression models explored spatial variations in local coefficients of the built environment predictors. Results. From ordinary least squares models, higher farmers’ market density was only significantly associated with lower obesity rates among elementary school students (b = −0.116; p < .01). Higher fast-food restaurant density was significantly associated with higher obesity rates (b = 0.014; p < .05), and higher land use mix was only significantly associated with lower obesity rates (b = −0.054; p < .01) among middle/high school students. In geographical weighted regression analyses, the inverse association between market density and obesity rates among elementary school students was more pronounced in the eastern portion of the state. The relationship between higher fast-food restaurant density and higher obesity rates among middle/high school students was found in the southeastern portion of the state. Conclusions. Different patterns of food consumption may explain varying determinants of obesity between younger and older students. Regional variations in local associations between the built environment variables and obesity may suggest differences in how healthy food sources are accessed locally.

Keywords

built environment childhood obesity food environment geographical weighted regression health disparity physical activity environment

Impact Statement

The high prevalence of obesity among New York’s school-aged students and geographic disparities have been documented, but the contribution of the built environment in explaining these disparities is not yet known. To address this research gap, we ran spatial econometric models using statewide student weight status surveillance data combined with other publicly available data, finding that the features of the built environment that were associated with childhood obesity differ between grades and that the strength of associations differed across the state. These findings provide support for tailored public health policy interventions to different age groups and local regions, and highlight the importance of maintaining robust surveillance systems for important public health issues such as childhood obesity.

Obesity among children and adolescents is a critical public health issue in the United States. From the 1980s to 2012, the percentages of obese children and adolescents increased from 7% to 18%, and from 5% to about 21%, respectively (National Center for Health Statistics, 2012; Ogden, Carroll, Kit, & Flegal, 2014). In New York State (NYS) outside New York City (NYC), 17.6% of school-aged children were obese from 2010 to 2012 (New York State Department of Health [NYSDOH], 2013). Annually, obesity-related treatments and expenditures in NYS have cost Medicaid more than $4.3 billion, and $7.5 billion to private health insurance and Medicare (Office of the State Comptroller, 2012). Consequently, childhood obesity is a critical focus area in the NYS Prevention Agenda (NYSDOH, 2016).

The NYSDOH established the student weight surveillance system in 2008 to monitor long-term obesity trends among children and adolescents in NYS schools. Using these data, the NYSDOH (2013) documented the high prevalence of obesity and regional geographic disparities. These data recently received considerable attention when news outlets highlighted these findings, thereby causing some schools to change their local school wellness policies (Waldman, 2013). A better understanding of factors contributing to geographic disparities in obesity rates is needed to identify tailored public health interventions to mitigate the childhood obesity epidemic.

The Built Environment and Obesity

We adapted the socioecological framework to better understand the contribution of the built environment on obesity among NYS students (Figure 1; Institute of Medicine, 2005). Obesity among students is determined by students’ individual factors as well as their social networks, built environment settings, and broader macro-level environments. Social networks, such as families and peer groups, have been associated with eating behaviors and physical activity among children and adolescents (Gruber & Haldeman, 2009; Salvy, De La Haye, Bowker, & Hermans, 2012). Built environment settings dictate individuals’ access to healthy foods, health resources, and facilities that promote physical activity and healthy behaviors (Institute of Medicine, 2005). More distal macro-level environmental factors, such as government policies, health care systems, and economic systems, can also influence how people live (Story, Kaphingst, Robinson-O’Brien, & Glanz, 2008).

Figure 1.

Socioecological model of obesity among school-aged children.

The built environment affects individual and community health outcomes by defining barriers and opportunities to engage in healthy behaviors. Some features of the built environment, such as land use diversity and street connectivity, have been associated with increased physical activity (Cervero & Kockelman, 1997; Frank et al., 2006; Wang, Wen, & Xu, 2013). Mixed land use may promote more active living by allowing residents to live close to multiple destinations without relying on motorized vehicles (Brown et al., 2009). Areas with better connectivity promote more walking by making available more and shorter routes (Doyle, Kelly-Schwartz, Schlossberg, & Stockard, 2006; Wang et al., 2013). Despite the theoretical models, empirical evidence on the associations among environmental conditions, physical activity, and obesity rates is still inconclusive (Durand, Andalib, Dunton, Wolch, & Pentz, 2011; Galvez, Pearl, & Yen, 2010).

Other built environment features relate to food access. The geographical distribution of food stores, services, and other entities where food may be obtained is associated with health outcomes including obesity (Danhong, Jaenicke, & Volpe, 2016). Supermarkets and large grocery stores sell more healthy food options and at lower prices than independent and smaller grocery stores (Lee, 2012). Farmers’ markets provide alternative sources of healthy food for low-income communities (Larsen & Gilliland, 2009; Ruelas, Iverson, Kiekel, & Peters, 2012). Fast-food restaurants are important food sources for many American households, who spend almost half of their food budgets on food away from home (Todd, Mancino, & Lin, 2010). Regular fast-food consumption among children is associated with obesity (Lee, 2012).

We aimed to better understand whether the built environment can explain disparities in childhood obesity rates. We combined student weight surveillance data with education, built environment, food resource outlets, and other data to examine geographical variations in local associations between the built environment and obesity among NYS students. This study contributes to the growing literature examining the effects of neighborhood factors on childhood obesity.

Method

Study Design

We used an ecological study design to investigate the relationships between the food and physical activity environments and obesity rates among students in selected school grade levels from 2010 to 2012. We combined the NYSDOH surveillance data, the Student Weight Status Category Reporting System (SWSCRS), with other publicly available data from state and federal government agencies (see the appendix for additional information on data sources and their data collection methodologies). The unit of analysis was school district, with the study population comprised of all school districts excluding NYC, which is exempt from participating in the SWSCRS.

Outcome Variables

The outcome, obesity rates among school-aged children, was operationalized as the proportions of obese students among all reportable grades within the school districts, by grade group. Students whose body mass index are at or above the 95th percentile of the sex- and age-specific values are considered obese. We reported obesity rates separately for elementary (prekindergarten/kindergarten, 2nd grade, and 4th grade) and middle/high schools (7th and 10th grades). The SWSCRS reports the proportions of obese students, aggregated by school district, based on schools’ reports on student counts in each weight status category by grade group and sex. The data were obtained from legally mandated student health certificates completed by medical providers.

Food Environment Variables

Three food environment measures (supermarket, fast-food restaurant, and farmers’ market densities) were operationalized as the number of vendors per square mile. We used similar approach employed by Chen, Florax, Snyder, and Miller (2010) to create a census of food stores and services by obtaining the names and addresses for supermarkets, fast-food restaurants, and farmers’ markets from the NYS Department of Agriculture (NYSDOA) and Markets Food Retail Food Store Database, the NYSDOH Food Services Establishment Inspection Report, and the NYS Department of Agriculture Farmers Markets Directory. Their addresses were geocoded using the NYS Street Address Mapping database and Google’s geocoding service. Locations with missing longitude and latitude were excluded.

We defined supermarkets as large corporate-owned chained food stores, distinguished grocery stores, or smaller non–corporate-owned food stores (Morland, Diez Roux, & Wing, 2006). The list of supermarket names was created from multiple online sources, with a final list of 105 supermarkets names (Appendix, Table A2). We also added keywords such as “market,” “grocery,” “mart,” and “food market” to identify nonchain supermarkets. We identified 2,373 supermarkets.

We used a similar approach employed by Currie, DellaVigna, Moretti, and Pathania (2010) to define fast-food restaurants. We used the 15 most popular fast-food restaurant chains in 2012 based on Technomic’s report on the chains’ total sales in 2011 to identify fast-food restaurants (McConnell & Bhasin, 2012; Appendix, Table A3), yielding 2,399 establishments.

Physical Activity Environment Variables

We included two measures of the physical activity environment: built environment diversity and connectivity. We used the Environmental Protection Agency Smart Location Database (SLD) to create these measures (Ramsey & Bell, 2014). The built environment diversity was operationalized as land use mix and measured using land use entropy (Cervero & Kockelman, 1997), which calculates the evenness of land use distribution in a given geographical unit (Brown et al., 2009). We used the employment and household entropy variable from the SLD, which was computed from all employments within the five-tier employment categories and the number of occupied housing units within a geographical unit (Ramsey, 2014). It ranges between zero (single land use environment) and one (perfectly mixed environment).

Built environment connectivity was operationalized using the street intersection density variable from the SLD. It is calculated as the weighted sum of intersections accessible by motorized vehicles and pedestrian-oriented intersections with limited access for motorized vehicles per square mile (Ramsey, 2014).

Census blocks did not perfectly correspond with school districts, with some blocks spanning multiple districts. For census blocks spanning multiple districts, the proportions of blocks in each school district were used as weights when assigning each block’s value to corresponding districts.

Control Variables

We controlled for factors that might confound the relationship between the built environment variables and obesity. Childhood poverty adjusted for the increased likelihood that children and adolescents from low-socioeconomic families are overweight and obese (Drewnowski, Rehm, Kao, & Goldstein, 2009; Singh, Siahpush, & Kogan, 2010). We operationalized it as the proportion of students receiving free or reduced-price lunch, from the 2012 NYS Education Department School Report Card.

Racial composition accounted for community-level racial and ethnic composition as a risk factor for obesity (Caprio et al., 2008; Kirby, Liang, Chen, & Wang, 2012). We operationalized it as the proportions of Black and Hispanic students, from the 2012 NYS Education Department School Report Card.

To account for possible differences in obesity risk factors by urbanicity, we included dummy variables for city and rural/town school districts, with suburban as the reference category. This adjusts for some exposures, such as industrial pollution, congestion, and motor vehicle accidents, being more common in urban environments (Vlahov & Galea, 2002). We classified urbanicity using the school district locale codes from the National Center for Education Statistics.

Statistical Analyses

Ordinary least squares (OLS) models examined associations between the built environment variables and school district–level obesity rates and provided the global baseline models for the subsequent geographical weighted regression (GWR) analyses. We ran separate models for the elementary and middle/high school grades and estimated two separate models per outcome: with and without the urbanicity control variables. Models with urbanicity variables served as robustness checks for the parameter estimates from the original models before proceeding to GWR analyses.

The GWR assessed whether local associations vary geographically by allowing regression coefficients to differ across geographic units (Brunsdon, Fotheringham, & Charlton, 1998). A Monte Carlo randomization test for significant spatial variability of GWR parameter estimates assessed the justification for the GWR models (Brunsdon et al., 1998). The GWR models were calibrated with an adaptive bisquare kernel function and a “nearest neighbor” approach, with each local point estimated using data points from the nearest neighbors (Brunsdon et al., 1998). The optimum numbers of the nearest neighbors were obtained by minimizing the corrected Akaike information criterion statistics (Yu, 2006). We generated maps to display spatial distributions of local coefficients for selected variables with significant spatial variations of local coefficients.

We performed all statistical analyses in R (Version 3.3.0), using the linear model (lm) function (OLS), and GWmodel package (GWR). We implemented multiple imputation (m = 10) to handle missing values in outcome variables using expectation–maximization with bootstrapping algorithm in the AMELIA II package. We determined there was a high probability that values were missing at random using a nonparametric test of homoscedasticity in the MissMech package (Jamshidian, Jalal, & Jansen, 2014). Imputation did not change the pattern or significance of the regression results.

Human Subjects

All data were available publicly under an open license and therefore exempt from institutional review board review.

Results

Descriptive Statistics

Table 1 reports school district characteristics from 2010 to 2012. Of the 680 school districts, 637 and 609 school districts reported their student obesity rates for elementary and middle/high grade groups, respectively. The percentage of obese students in elementary and middle/high grade groups across school districts ranged from 1.3% to 46.4% and from 3.8% to 64.3%, respectively. Fast-food restaurants and supermarkets were more common than farmers’ markets (0.291 and 0.297 per square mile, compared to 0.006 per square mile, respectively). Only 3.5% of school districts were classified as city districts, and rural or town districts accounted for about 60% of the total.

Table 1.

Characteristics of New York State School Districts, 2010-2012.

Measure	All school districts
Proportion of obese elementary school students, M (SD)^a,b	0.18 (0.06)
Proportion of obese middle/high school students, M (SD)^b,c	0.20 (0.075)
Proportion of students receiving free/reduced price lunch, M (SD)	0.32 (0.184)
Proportion of black students, M (SD)	0.05 (0.096)
Proportion of Hispanic students, M (SD)	0.07 (0.107)
Farmers’ market density, M (SD)	0.01 (0.056)
Supermarket density, M (SD)	0.30 (0.747)
Fast-food restaurant density, M (SD)	0.29 (0.582)
Land use mix, M (SD)	0.63 (0.16)
Intersection density, M (SD)	22.24 (34.06)
City location, N (%)	24 (3.53%)
Suburban location, N (%)	240 (35.29%)
Rural or town location, N (%)	416 (61.18%)
No. of school districts, N (%)	680 (100%)

For the 2010-2012 reporting year, the median overall completeness was 81% for district total (interquartile range, 83% for middle/high schools and 79% for elementary schools). The Family Educational Rights and Privacy Act grants parents and guardians the right to request the exclusion of their children’s weight status data from reporting. In addition to this data suppression, there may be other sources of missing data such as absenteeism. ^bThe reported percentages were limited to the 637 districts with one or more elementary schools. The average number of reported obese students per school district was 76.7 (range = 5-1,294, SD = 103.4 students). ^cThe reported percentages were limited to the 609 districts with one or more middle/high schools. The average number of reported obese students per school district was 76.7 (range = 5 to 912, SD = 58.7).

Linear Regression Results

From the OLS regressions, the proportion of students receiving free or reduced price lunch was a consistent predictor of obesity in both grade levels (Table 2). The proportion of Black students was not significantly associated with obesity rates, whereas the proportion of Hispanic students was only positively associated with obesity rate in elementary school grades (b = 0.084; p < 0.01). Rural districts had higher obesity rates compared to city and suburban districts but only for elementary school students (b = 0.015; p < 0.05).

Table 2.

Parameter Estimates of the Relationships Between the Built Environment Variables and Proportion of Obese Students in New York State, 2010-2012.

Variable	Elementary school grades, B (robust SE)		Middle/high school grades, B (robust SE)
Proportion of free/reduced-price lunch students	0.173 (0.016)***	0.164 (0.018)***	0.240 (0.016)***	0.236 (0.016)***
Proportion of black students	0.007 (0.032)	0.024 (0.031)	−0.032 (0.040)	−0.010 (0.040)
Proportion of Hispanic students	0.084 (0.026)**	0.083 (0.026)**	−0.027 (0.028)	−0.031 (0.029)
Farmers’ market density	−0.116 (0.027)***	−0.098 (0.029)***	0.002 (0.035)	0.024 (0.034)
Supermarket density	−0.003 (0.004)	−0.003 (0.004)	0.002 (0.007)	0.002 (0.007)
Fast-food restaurant density	−0.000 (0.004)	−0.001 (0.004)	0.015 (0.006)**	0.014 (0.006)*
Land use mix	−0.021 (0.016)	−0.011 (0.016)	−0.054 (0.020)**	−0.044 (0.021)*
Intersection density	−0.000 (0.0001)	0.00 (0.0001)	−0.000 (0.0001)*	−0.000 (0.0002)
City^a		−0.013 (0.010)		−0.022 (0.014)
Rural/town^a		0.015 (0.006)*		0.012 (0.007)
Constant	0.129 (0.013)***	0.113 (0.013)***	0.160 (0.015)***	0.146 (0.017)***
Observations	677	677	669	669
R ²	0.325	0.325	0.385	0.391
Adjusted R²	0.317	0.317	0.378	0.382

Suburban school district is the reference category.

p < .05. **p < .01. ***p < .001.

The built environment variables had different effects by grade level. Among the food environment variables, higher farmers’ market density was significantly associated with lower obesity rates in elementary school grades (b = −0.116; p < .001). A higher fast-food restaurant density was significantly associated with obesity among middle/high school students (b = 0.015; p < .05). Among the physical activity environment variables, only land use mix was significantly associated with lower obesity rates among middle/high school students (b = −0.054; p < .01). These coefficients slightly changed but remained significant even after city and rural/town variables were included in the models.

Geographically Weighted Regression Results

Table 3 reports the summary of estimated local coefficients and the results of the Monte Carlo tests for spatial variability of parameter estimates for the built environment variables of interest. The farmers’ market density was the only built environment variable that had a significant inverse association with the obesity rate in the elementary school student model, and this association varied statistically across school districts (Table 3). The fast-food restaurant density was the only variable in the middle/high school student models that had a significant association with the obesity rate, and this association also varied statistically across school districts (Table 3). We only mapped significant local coefficients of these two variables, as they were significant predictors with significant spatial variability in their local parameter estimates. Figure 2 shows that the inverse local associations between farmer’s market density and obesity among elementary school students varied statewide, and were only statistically significant in Nassau-Suffolk County, Hudson Valley, and the northeastern region. For middle/high school students, significant local coefficients of fast-food restaurant density variable were concentrated in the Capital Region, Hudson Valley, and Nassau-Suffolk counties (Figure 3).

Table 3.

Geographically Weighted Regression of the Built Environment Variables on Proportion of Obese Students in New York State, 2010-2012.

Variable	OLS, B (robust SE)	Local coefficient estimates^a					Monte Carlo spatial variability test^b
Variable	OLS, B (robust SE)	Minimum	1st Quartile	Mdn	3rd Quartile	Maximum	Monte Carlo spatial variability test^b
Outcome variable: Proportion of obese elementary school students
Farmers’ market density	−0.116 (0.027)***	−0.142	−0.138	−0.119	−0.020	0.540	Nonstationary***
Supermarket density	−0.003 (0.004)	−0.008	−0.006	0.001	0.011	0.028	Nonstationary***
Fast-food restaurant density	−0.000 (0.004)	−0.030	−0.013	−0.007	0.003	0.004	Nonstationary***
Land use mix	−0.021 (0.016)	−0.035	−0.028	−0.008	0.001	0.008	Stationary
Intersection density	−0.000 (0.0001)	−0.001	−0.000	−0.000	0.000	0.000	Nonstationary**
Constant	0.129 (0.013)***	0.118	0.124	0.127	0.130	0.133	Stationary
Outcome variable: Proportion of obese middle/high school students
Farmers’ market density	0.002 (0.035)	−0.488	−0.026	−0.004	0.040	0.629	Nonstationary**
Supermarket density	0.002 (0.007)	−0.005	−0.002	0.0003	0.012	0.033	Stationary
Fast-food restaurant density	0.015 (0.006)**	−0.052	−0.009	0.017	0.019	0.058	Nonstationary***
Land use mix	−0.054 (0.020)**	−0.074	−0.056	−0.050	−0.045	−0.026	Stationary
Intersection density	−0.000 (0.0001)*	−0.000	−0.000	−0.000	−0.000	−0.000	Stationary
Constant	0.160 (0.015)***	0.140	0.149	0.162	0.171	0.186	Stationary

Note. OLS = ordinary least squares. Both models included socioeconomic control variables.

The results of geographical weighted regression analyses are local parameter estimates for all given data points (geographical units). These columns report the distribution of local parameter estimates for all geographical units. ^bThis is a test for significant spatial variability of geographical weighted regression models’ parameter estimates. It tests against the null hypothesis that the coefficients or parameter estimates are stationary or that there is no significant variation in the local coefficients across observations. Significant results indicate that the coefficients vary and justify the use of geographically weighted regression models.

p < .05. **p < .01. ***p < .001.

Figure 2.

Local parameter estimates of farmers’ market density on proportion of obese students in elementary school grades in New York State, 2010-2012.

Figure 3.

Local parameter estimates of fast-food restaurant density on proportion of obese students in middle/high school grades in New York State, 2010-2012.

Discussion

We combined the NYS student obesity surveillance data with other publicly available state and federal data in novel ways to investigate the relationship between the built environment and obesity among NYS students. Obesity prevalence varied dramatically across the state, from 1.3% to 46.4% among elementary students and from 3.8% to 64.3% among middle/high school students. Farmers’ market density and land use mix were associated with lower obesity rates, and fast-food restaurant density was associated with higher obesity rates, even after controlling for sociodemographic factors and urbanicity. However, these associations differed by grade: Higher farmers’ market density was associated only with lower obesity rates among elementary school students, higher land use mix and intersection density were associated with reduced obesity among middle/high school students, and higher fast-food restaurant density was associated with higher obesity rates among middle/high school students only. The GWR models documented geographical variations in the associations between food environment predictors and obesity: The inverse local associations between farmers’ market density and obesity in elementary school students were concentrated in the northeastern, Hudson Valley, and Nassau-Suffolk regions, and the positive local associations between fast-food restaurant density among middle/high school students were more pronounced in the Capital District region.

Many findings from the OLS regressions were consistent with previous literature on childhood and adolescent obesity. The proportion of students receiving free or reduced-price lunch was a consistent predictor of obesity for all grades and confirms other studies documenting higher obesity rates among socioeconomically disadvantaged households (Drewnowski et al., 2009; Singh et al., 2010). In terms of race/ethnicity variables, we only found significant association between the proportion of Hispanic students and higher obesity rates among elementary school students. Our ability to detect associations between obesity and race/ethnicity variables was limited because less than 5% of school districts had substantial proportions of Black students. Similarly, our ability to detect associations between supermarket density and obesity was limited due to the small number of school districts with relatively high supermarket densities.

Different determinants between younger and older students might be attributed to different food consumption patterns. In particular, higher farmers’ market density was associated only with lower obesity rates among younger students, whereas fast-food restaurant density was associated only with higher obesity rates among older students. One potential explanation is that middle/high school students are more mobile, thereby less reliant on meals provided at home and able to purchase fast food themselves. One study found that children from kindergarten through the fourth grade were less likely to eat meals at restaurants (Paleti, Copperman, & Bhat, 2011). Moreover, fast-food consumption was higher among adolescents than younger children (French, Story, Neumark-Sztainer, Fulkerson, & Hannan, 2001). However, the farmers’ market density variable might also capture other unobserved school district characteristics, such as high proportions of more educated households with children. One study suggested that farmers’ markets tend to locate in areas with high concentrations of younger and more educated populations comprising households with children (Berning, Bonanno, & Etemaadnia, 2013). Limited access to food markets for individuals who may be at greatest risk for chronic disease, such as obesity, represents an important health equity issue (Jones & Bhatia, 2011).

Our study also contributes to the childhood obesity literature by exploring spatial patterns of local associations between food environments and obesity. The GWR analyses identified important regional differences. The significant local associations between farmers’ market and fast-food restaurant density and obesity were concentrated in eastern NYS, particularly in the Capital District metropolitan region. One potential explanation for this spatial pattern is that farmers’ markets may not be important sources of fresh produce in agricultural regions. Another possible explanation is that farmers’ markets may not be used by lower income residents unless they are easily accessible via public transportation or accept food stamps. Additional research is needed to understand the causes of these differences.

Strengths

This study has a number of strengths. Unlike most previous studies using small area study designs, this study is statewide, improving the generalizability of findings. Spatial econometric methods allowed us to explore geographical differences in local associations. The study findings provide support for the need to implement tailored public health interventions to address the childhood obesity epidemic, although further individual-level analyses are required. Using school district as the unit of analysis reflects social and economic segregation of neighborhoods, driven by income inequality and households’ choice of residence (Owens, 2016). Social and economic segregation across school districts can affect the obesogenic environment that may influence weight status among children.

Limitations

This study has several limitations. The built environment predictors might suffer from measurement errors. Unlike most studies on food environment that used the North American Industry Classification System, our data did not have this information, requiring us to rely on establishment brand names. This approach might exclude local and nonchain food outlets and service establishments, thereby underestimating the total number. Due to variations in brand names, we could not include convenience stores in our analysis, which have been associated with obesity among children and adolescents (Galvez et al., 2009; Ohri-Vachaspati, Lloyd, DeLia, Tulloch, & Yedidia, 2013). For the physical activity environment variables, aggregating from the census block to school district level may have introduced some measurement errors. Excluding NYC may limit generalizability, as its densely populated environment is distinct from other urban environments. This exclusion may have caused misspecification in GWR models, leading to possibly biased local coefficients in surrounding areas. Significant associations between the built environment variables and student obesity rates require careful interpretation due to possible ecological fallacy resulting from the study’s ecological design. Obesity among some students might be caused by factors other than exposure to specific built environment features.

Implications for Public Health Policy and Practice

Our study contributes to the childhood obesity literature by exploring different determinants of obesity between younger and older students, between rural and urban communities, and across geographical regions. These results suggest that one-size-fits-all community-level interventions may be less effective. For instance, a metropolitan area may require specific policy interventions that target different features of the built environment. In other regions, policy interventions targeting various issues associated with poverty may be more beneficial.

More broadly, this study highlights the value of maintaining robust surveillance systems for important public health issues such as childhood obesity. As part of its government transparency initiative, the NYSDOH, the U.S Department of Health and Human Services, and other state and federal agencies have devoted substantial efforts to publishing open data that are publicly accessible without use and distribution rights (Manyika et al., 2013). Using open data resources can eliminate the need to purchase expensive proprietary built environment and other data. This study therefore also serves as a use case for how open data resources can be combined creatively to generate new insights about important public health issues.

This study advances public health practice in two ways. First, the quantitative findings increase our understanding of the complex nature of the childhood obesity epidemic, providing evidence that features of the built environment may have different effects regionally and by age. Second, we illustrate the potential uses of publicly available data for public health research, as this analysis would have been costlier and more difficult to perform before the development of open data portals. Efforts to improve the availability and accessibility of health data may facilitate high-quality research to inform better health policies.

Footnotes

Appendix

Table A3.

Most Popular Fast-Food Restaurant Chains in 2012.

Rank	Name
1	McDonald’s
2	Subway
3	Starbucks
4	Wendy’s
5	Burger King
6	Taco Bell
7	Dunkin’ Donuts
8	Pizza Hut
9	Kentucky Fried Chicken
10	Chick-Fil-A
11	Sonic Drive-Ins
12	Domino’s Pizza
13	Panera Bread
14	Arby’s
15	Jack in the Box

Note. The fast-food restaurant chains’ names were from the Technomic’s report (McConnell & Bhasin, 2012).

Acknowledgements

The authors are grateful to the New York State Department of Health Bureau of Chronic Disease Evaluation and Research staff and to Ashley Fox and Gang Chen from the University at Albany for comments on an early draft.

Authors’ Note

The views expressed in the article are those of the authors and not the New York State Department of Health.

Declaration of Conflicting Interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: IB oversees the Student Weight Status Category Reporting System data that were used in this analysis.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Robert Wood Johnson Foundation’s Public Health Services & Systems Research Program (Grant ID#71597 to EGM and GSB). AD received support from the Fulbright Scholarship. EGM and GSB were the co–principal investigators for a grant from the Robert Wood Johnson Foundation’s Public Health Services & Systems Research Program (Grant ID#71597) from which this study was funded.

References

Berning

Bonanno

Etemaadnia

(2013, August). Farmers’ markets location determinants: An empirical analysis in New England. Paper presented at the Agricultural & Applied Economics Association’s 2013 AAEA & CAES joint annual meeting, Washington, DC. Retrieved from http://ageconsearch.umn.edu/bitstream/150593/2/Berning-Bonanno-Etemaadnia-AAEA%202013.pdf

Brown

B. B.

Yamada

Smith

K. R.

Zick

C. D.

Kowaleski-Jones

Fan

J. X.

(2009). Mixed land use and walkability: Variations in land use measures and relationships with BMI, overweight, and obesity. Health & Place, 15, 1130-1141. doi:10.1016/j.healthplace.2009.06.008

Brunsdon

Fotheringham

Charlton

(1998). Geographically weighted regression-modelling spatial non-stationarity. Journal of the Royal Statistical Society. Series D (The Statistician), 47, 431-443.

Caprio

Daniels

S. R.

Drewnowski

Kaufman

F. R.

Palinkas

L. A.

Rosenbloom

A. L.

Schwimmer

J. B.

(2008). Influence of race, ethnicity, and culture on childhood obesity: Implications for prevention and treatment: A consensus statement of Shaping America’s Health and the Obesity Society. Diabetes Care, 31, 2211-2221. doi:10.2337/dc08-9024

Cervero

Kockelman

(1997). Travel demand and the 3Ds: Density, diversity, and design. Transportation Research Part D: Transport and Environment, 2, 199-219. doi:10.1016/S1361-9209(97)00009-6

Chen

Florax

R. J. G. M.

Snyder

Miller

C. C.

(2010). Obesity and access to chain grocers. Economic Geography, 86, 431-452. doi:10.1111/(ISSN)1944-8287/issues

Currie

DellaVigna

Moretti

Pathania

(2010). The effect of fast food restaurants on obesity and weight gain. American Economic Journal: Economic Policy, 2(3), 32-63. doi:10.1257/pol.2.3.32

Danhong

Jaenicke

E. C.

Volpe

R. J.

(2016). Food environments and obesity: Household diet expenditure versus food deserts. American Journal of Public Health, 106, 881-888 888p. doi:10.2105/AJPH.2016.303048

Doyle

Kelly-Schwartz

Schlossberg

Stockard

(2006). Active community environments and health: The relationship of walkable and safe communities to individual health. Journal of the American Planning Association, 72(1), 19-31. doi:10.1080/01944360608976721

10.

Drewnowski

Rehm

Kao

Goldstein

(2009). Poverty and childhood overweight in California Assembly districts. Health & Place, 15, 631-635. doi:10.1016/j.healthplace.2008.09.008

11.

Durand

C. P.

Andalib

Dunton

G. F.

Wolch

Pentz

M. A.

(2011). A systematic review of built environment factors related to physical activity and obesity risk: Implications for smart growth urban planning. Obesity Reviews, 12(5), e173-e182. doi:10.1111/j.1467-789X.2010.00826.x

12.

Frank

L. D.

Sallis

J. F.

Conway

T. L.

Chapman

J. E.

Saelens

B. E.

Bachman

(2006). Many pathways from land use to health: Associations between neighborhood walkability and active transportation, body mass index, and air quality. Journal of the American Planning Association, 72(1), 75-87. doi:10.1080/01944360608976725

13.

French

S. A.

Story

Neumark-Sztainer

Fulkerson

J. A.

Hannan

(2001). Fast food restaurant use among adolescents: Associations with nutrient intake, food choices and behavioral and psychosocial variables. International Journal Obesity and Related Metabolic Disorder, 25(12), 1823-1833.

14.

Galvez

M. P.

Hong

Choi

Liao

Godbold

Brenner

(2009). Childhood obesity and neighborhood food store availability in an inner city community. Academic Pediatrics, 9, 339-343. doi:10.1016/j.acap.2009.05.003

15.

Galvez

M. P.

Pearl

Yen

I. H.

(2010). Childhood obesity and the built environment: A review of the literature from 2008-2009. Current Opinion in Pediatrics, 22(2), 202-207. doi:10.1097/MOP.0b013e328336eb6f

16.

Gruber

K. J.

Haldeman

L. A.

(2009). Using the family to combat childhood and adult obesity. Preventing Chronic Disease, 6(3), A106.

17.

Institute of Medicine. (2005). Preventing childhood obesity: Health in the balance. Washington, DC: National Academies Press.

18.

Jamshidian

Jalal

Jansen

(2014). Missmech: An R package for testing homoscedasticity, multivariate normality, and missing completely at random (MCAR). Journal of Statistical Software, 56(6). Retrieved from https://escholarship.org/uc/item/51x8q0nn

19.

Jones

Bhatia

(2011). Supporting equitable food systems through food assistance at farmers’ markets. American Journal of Public Health, 101, 781-783. doi:10.2105/AJPH.2010.300021

20.

Kirby

J. B.

Liang

Chen

H.-J.

Wang

(2012). Race, place, and obesity: The complex relationships among community racial/ethnic composition, individual race/ethnicity, and obesity in the United States. American Journal of Public Health, 102, 1572-1578 1577p. doi:10.2105/AJPH.2011.300452

21.

Larsen

Gilliland

(2009). A farmers’ market in a food desert: Evaluating impacts on the price and availability of healthy food. Health & Place, 15, 1158-1162. doi:10.1016/j.healthplace.2009.06.007

22.

Lee

(2012). The role of local food availability in explaining obesity risk among young school-aged children. Social Science & Medicine, 74, 1193-1203. doi:10.1016/j.socscimed.2011.12.036

23.

Manyika

Chui

Groves

Farrell

Kuiken

S. V.

Doshi

E. A.

(2013). Open data: Unlocking innovation and performance with liquid information. Retrieved from http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/open-data-unlocking-innovation-and-performance-with-liquid-information

24.

McConnell

Bhasin

(2012, July 12). Ranked: The most popular fast food restaurants in America. Retrieved from http://www.businessinsider.com/the-most-popular-fast-food-restaurants-in-america-2012-7?op=1

25.

Morland

Diez Roux

A. V.

Wing

(2006). Supermarkets, other food stores, and obesity. American Journal of Preventive Medicine, 30, 333-339. doi:10.1016/j.amepre.2005.11.003

26.

National Center for Health Statistics. (2012). Health, United States, 2011: With special feature on socioeconomic status and health. Retrieved from https://www.cdc.gov/nchs/data/hus/hus11.pdf

27.

New York State Department of Health. (2013). Information for Action # 2013-5: One-third of school-age children in New York State (exclusive of New York City) are overweight or obese. Retrieved from https://www.health.ny.gov/statistics/prevention/injury_prevention/information_for_action/docs/2013-05_ifa_report.pdf

28.

New York State Department of Health. (2016). Prevention agenda 2013-2018: New York State’s health improvement plan. Retrieved from https://www.health.ny.gov/prevention/prevention_agenda/2013-2017/index.htm

29.

Office of the State Comptroller. (2012). Soaring health care costs highlight need to address childhood obesity. Retrieve from https://www.osc.state.ny.us/reports/obesity_and_child_obesity_10_23_12.pdf

30.

Ogden

C. L.

Carroll

M. D.

Kit

B. K.

Flegal

K. M.

(2014). Prevalence of childhood and adult obesity in the United States, 2011-2012. Journal of the American Medical Association, 311, 806-814. doi:10.1001/jama.2014.732

31.

Ohri-Vachaspati

Lloyd

DeLia

Tulloch

Yedidia

M. J.

(2013). A closer examination of the relationship between children’s weight status and the food and physical activity environment. Preventive Medicine, 57, 162-167. doi:10.1016/j.ypmed.2013.05.009

32.

Owens

(2016). Inequality in children’s contexts: Income segregation of households with and without children. American Sociological Review, 81, 549-574. doi:10.1177/0003122416642430

33.

Paleti

Copperman

R. B.

Bhat

C. R.

(2011). An empirical analysis of children’s after school out-of-home activity-location engagement patterns and time allocation. Transportation, 38, 273-303. doi:10.1007/s11116-010-9300-2

34.

Ramsey

(2014). Smart Location Database: Version 2.0 user guide. Retrieved from https://www.epa.gov/sites/production/files/2014-03/documents/sld_userguide.pdf

35.

Ramsey

Bell

(2014). The Smart Location Database: A nationwide data resource characterizing the built Environment and destination accessibility at the neighborhood scale. Cityscape, 16, 145-162. Retrieved from http://www.huduser.org/periodicals/cityscpe/prev_iss/cspast.html

36.

Ruelas

Iverson

Kiekel

Peters

(2012). The role of farmers’ markets in two low income, urban communities. Journal of Community Health, 37, 554-562. doi:10.1007/s10900-011-9479-y

37.

Salvy

S.-J.

De La Haye

Bowker

J. C.

Hermans

R. C. J.

(2012). Influence of peers and friends on children’s and adolescents’ eating and activity behaviors. Physiology & Behavior, 106, 369-378.

38.

Singh

G. K.

Siahpush

Kogan

M. D.

(2010). Rising social inequalities in U.S. childhood obesity, 2003–2007. Annals of Epidemiology, 20, 40-52. doi:10.1016/j.annepidem.2009.09.008

39.

Story

Robinson-O’Brien

Glanz

(2008). Creating healthy food and eating environments: Policy and environmental approaches. Annual Review of Public Health, 29, 253-272.

40.

Todd

J. E.

Mancino

Lin

B.-H.

(2010). The impact of food away from home on adult diet quality. Washington, DC: U.S. Department of Agriculture, Economic Research Service.

41.

Vlahov

Galea

(2002). Urbanization, urbanicity, and health. Journal of Urban Health, 79(Suppl. 1), S1-S12. doi:10.1093/jurban/79.suppl_1.S1

42.

Waldman

(2013, July 16). State reports obesity rates in area school districts. Local News, Times Union. Retrieved from http://www.timesunion.com/local/article/State-reports-obesity-rates-in-area-school-4667336.php

43.

Wang

Wen

(2013). Population-adjusted street connectivity, urbanicity and risk of obesity in the U.S. Applied Geography, 41, 1-14. doi:10.1016/j.apgeog.2013.03.006

44.

D.-L.

(2006). Spatially varying development mechanisms in the greater Beijing area: A geographically weighted regression investigation. Annals of Regional Science, 40, 173-190.