Rural–Urban Disparities in Obesity Prevalence Among Working Age Adults in the United States: Exploring the Mechanisms

Abstract

Purpose:

Higher prevalence rates of overweight and obesity in rural America have been consistently reported, but sources of these disparities are not well known. This study presented patterns and mechanisms of these disparities among working age Americans.

Design:

Cross-sectional study.

Setting:

United States of America.

Participants:

The study included 10 302 participants of the 2003-2008 National Health and Nutrition Examination Survey (NHANES) who were 20 to 64 years old, not pregnant, and with a body mass index ranging from 18.5 to 60.

Measures:

Individual-level data were from NHANES including age, gender, race/ethnicity, immigrant status, education, and family income. The outcomes were prevalence of obesity and prevalence of overweight and obesity combined. Neighborhood data were constructed from the 2000 US Census providing tract-level information on family median income and built environmental features and from the 2006 ESRI ArcGIS 9.3 Data DVD providing tract-level park location information.

Analysis:

Geographic information system (GIS) methods were used to create a measure of spatial distance to local parks capturing park accessibility. Random intercept logistic and ordinal logit regression analyses were performed.

Findings:

Multivariate regression results showed that the odds of obesity was higher in rural areas compared to urban areas (odds ratio = 1.358, P < .001) net of demographic controls and that this gap was largely attributable to individual educational attainment and neighborhood median household income and neighborhood built environmental features. After controlling for these hypothesized mediators, the elevated odds associated with rural residence was reduced by nearly 94% and rendered statistically insignificant.

Conclusions:

In this nationally representative cross-sectional sample, rural–urban obesity disparities were large and explained by rural–urban educational differences at the individual level and economic and built environmental differences at the neighborhood level.

Keywords

obesity built environment park rural urban mechanisms NHANES

According to the 2010 US Census, 19.3% of the American population lived in rural areas, distributed over 95% of the country’s land mass.¹ Rural–urban health disparities in terms of both morbidity and mortality have been routinely reported in the United States.^2

–5 Previous work tended to focus on the health-care system and access-to-care barriers as potential sources of the observed poorer health outcomes among rural residents compared to their urban counterpart in the United States.⁶ More recently, research and policy attention has turned to population-based, preventive approaches for improving the physical, mental, and social well-being of rural residents.⁷ In this vein, lifestyle factors are suspected as a key contributor to rural health disadvantages,^5,8,9 among which higher obesity prevalence rates in rural areas have drawn increasing attention in recent years.

Two earlier nationally representative studies using data from the 1994 to 1996 and 2000 to 2001 Behavioral Risk Factor Surveillance System (BRFSS) and the 1998 National Health Interview Survey have documented higher self-reported obesity prevalence rates in rural areas compared to those in urban areas.^10,11 A recent study has revisited this issue and confirmed higher obesity prevalence in rural America, using objectively measured obesity data from the 2005 to 2008 National Health and Nutrition Examination Survey (NHANES),¹² finding that the higher rural obesity prevalence remains significant after a host of sociodemographic and lifestyle factors are controlled.

Mechanisms for higher rural obesity prevalence rates have been proposed but have not been well examined. Authors have speculated that rural residents’ individual and community characteristics jointly mediate the link between rural residence and higher obesity prevalence.¹³ At the individual level, rural residents are more likely to live in poverty,^14,15 have lower educational attainment,^14,15 and follow less healthy lifestyles such as consuming less nutritious diets than urban residents.^12,16,17 Poverty, low education, and unhealthy lifestyle have all been linked to higher obesity risk,^18

–23 making them likely sources of rural areas’ higher obesity rates.

However, despite observed differences in dietary intake between rural and urban populations, some studies have not observed differences in total energy intake by rurality.^12,17 Studies have also struggled to observe consistent differences in energy expenditure according to rurality perhaps due to difficulties in measuring physical activity. Recent evidence based on nationally representative data showed that the direction and significance of rural–urban difference in physical activity varied by the method of physical activity measurement, plausibly due to that rural residents generally spend more time in low-intensity household physical activity but less time in high-intensity physical activity.²⁴ Therefore, the role of diet and physical activity in explaining rural–urban obesity disparity seems more elusive and difficult to assess than that of socioeconomic factors.

At the community or neighborhood level, socioeconomic status (SES),^13,25 access to fast foods and/or healthy food choices,^16,26 and physical activity environmental features such as availability of public transportation, land use pattern, park accessibility, and walkability^16,27
–29 have all been linked to individual-level obesity risks over and above individual-level risk factors in multilevel studies. Nonetheless, the relationships between these environmental factors and rural residence are complex. A recent study showed that rural areas had a greater number of supermarkets, playgrounds, swimming pools, and public open space per 1000 persons than urban areas, but that greater travel distances were required to reach these amenities in rural areas.³⁰ Whether neighborhood social and physical features can help explain rural–urban obesity disparity needs to be further explored.

Few studies have directly examined these mediating effects with 1 exception. Bennett and colleagues¹³ analyzed data from the 2005 BRFSS and found the observed rural–urban obesity disparities were largely attributable to a range of individual sociodemographic and behavioral factors and county-level food and health-care environmental features. This study did not directly examine the mediating role of county-level poverty, relied on self-report of height and weight to characterize obesity status, and used county-level variables to capture community contexts which may mask within-county variations in obesogenic environmental exposures.

Given this background, the main purpose of the current study was to examine rural–urban disparities in overweight and obesity and explore individual- and neighborhood-level mediators that may offer some explanations for these disparities. Our research advances the previous work in 3 important ways. First, we used objectively measured height and weight data collected in a nationally representative survey to identify overweight and obesity categories. Second, we used census tracts as the main geographic unit to capture more immediate residential environments than what county-level data could offer. And third, we simultaneously considered both individual-level and neighborhood-level mediators of weight-related urban–rural disparities.

Methods

Data

The study used data from the 2003 to 2008 NHANES. The NHANES is a continuing study designed to assess the health and nutritional status of adults and children in the United States, using a complex, multistage, probability sampling design to select participants who are representative of the civilian, noninstitutionalized US population.³¹ The NHANES is unique in that it combines interviews and physical examinations. We focused on working age adults who were 20 to 64 years old, as the health implications for obesity may differ for children and older adults. To make references for working age population under normal weight and health conditions, we excluded pregnant women and those with body mass index (BMI) greater than 60 (considered outliers). We also excluded underweight cases to focus on the contrast between excessive weight and normal weight. After further dropping cases missing covariate information, the final individual-level analytical sample included 10 302 observations with 8188 urban respondents and 2114 rural respondents. Neighborhood-level data were constructed from the 2000 US Census providing information on tract income and built environmental features and the 2006 ESRI ArcGIS 9.3 Data DVD³² providing park location information.

The individual-level NHANES data were merged with the neighborhood-level data using the 2000 Census tract identifiers by staff of the Research Data Center at National Center for Health Statistics, Centers for Disease Control and Prevention (CDC). The merged data were made available for us to analyze with both individual and census tract identifiers excluded. The study was declared exempt by the institutional review board of University of Utah.

Dependent Measure

Weight status

Height and weight were collected in a mobile examination center using standardized protocols (http://www.cdc.gov/nchs/data/nhanes/meccomp.pdf). Overweight was defined as BMI ≥25 kg/m² and <30 kg/m². Obesity was defined as BMI ≥30 kg/m². The reference group was the respondents of normal weight (BMI ≥18.5 and <25) when the odds of overweight was examined; the reference group was normal weight and overweight categories combined when the odds of obesity was examined.

Independent Measures

Individual sociodemographic variables

Age (continuously measured), gender (male vs female), race/ethnicity (non-Hispanic blacks, Hispanics, others vs non-Hispanic whites), immigrant status (US-born vs foreign-born), educational attainment (high school graduates or equivalent, some college, bachelor’s degree or higher vs less than high school), and income-to-needs ratio (continuous) were included in the analysis. Income-to-needs ratio was computed by NHANES staff using annual family income divided by the federal poverty threshold for the appropriate family size, location, and year (https://www.census.gov/hhes/www/poverty/about/overview/measure.html). Age, gender, race/ethnicity, and immigrant status were controlled as confounding factors. Education and income-to-needs ratio were hypothesized to be individual-level mediators of rural–urban disparities in overweight and obesity.

Tract-level rural–urban residence

We used 2000 Rural–Urban Commuting Area (RUCA) codes to determine rurality, a measure developed by the US Department of Agriculture to capture tract-level rurality.¹ The primary 10-category RUCA codes were dichotomized into rural (RUCA = 4-10, areas with a population of less than 50 000 people) and urban (RUCA = 1-3, areas with a population of at least 50 000 people). The RUCA codes were geo linked to NHANES by staff at the CDC Research Data Center.

Other tract-level variables

Four neighborhood variables were hypothesized to be contextual mediators of rural–urban disparities in obesity. Median household income was included to measure neighborhood SES. Three additional neighborhood variables examined built environmental features of walkability, land use patterns, and spatial park accessibility. Walkability and land use patterns were captured by percentage driving for an hour or more per day to work (ie, prevalence of the long commuting residents in the neighborhood) and percentage walking, biking, or taking public transportation to work (ie, prevalence of residents taking active transportation to work) among the residents 16 and over who work outside of home. These 2 variables were constructed from the 2000 Census data. Park accessibility was measured using a recently published method to calculate census tract park accessibility.³³ Following this method, we identified 7 parks closest to a census block’s centroid and calculated average distance from the census block group’s centroid to each of these 7 parks. We then aggregated the block group-level average distance to parks to the census tract level weighted by the block group-level population and the 7 parks’ sizes. A key advantage of this new method of measuring neighborhood park accessibility is that it takes into account the uneven population distributions across areas and different park sizes. Details of this method have been published elsewhere.³³

Analyses

T tests (for continuous outcomes) or χ² (for categorical outcomes) tests were performed to compare urban and rural respondents’ individual and neighborhood characteristics. The geographic information system methods were used to construct a measure of neighborhood park accessibility. All the analyses were performed using the combined sample weights for 2003 to 2004, 2005 to 2006, and 2007 to 2008 to provide nationally representative results. Six random intercept models were fit to examine the patterns and mechanisms of rural–urban disparities in the 2 weight status outcomes, overweight and obesity, for which the same modeling strategy was adopted. Model 1 was the baseline model including rural–urban residence and all the individual-level controls. Compared to model 1, model 2 added education as an individual-level mediator, and model 3 added income-to-needs ratio as another individual-level mediator. Including all the individual-level variables, model 4 added neighborhood median household income. Model 5 added the 3 built environmental variables capturing walkability, land use patterns, and park accessibility; income was excluded this time due to its insignificant effect in the previous model (model 4). The last model, model 6, included all the previously significant variables at both the individual and neighborhood level. The software SAS version 9.2 was used to conduct the statistical analyses.

To examine the mediating effects, we adopted the most popular way to express effect size for mediation through informal descriptors, such as complete or partial mediation.³⁴ A complete mediation occurs when the effect of X (eg, rural–urban residence) on Y (eg, obesity) completely disappears when M (eg, education) is added as a predictor of Y. We also calculated the percentage reduction in the elevated odds of overweight/obesity associated with rural residence to express the practical importance of each hypothesized mediator for the key relation of interest in this study—rural–urban residence and the odds of excess body weight, following previous work.³⁵

Results

Table 1 shows the weighted descriptive statistics for the whole sample as well as for urban and rural subsamples. Obesity prevalence rate was higher in rural areas (41%) compared to that in urban areas (32%). Rural residents were also older, more likely to be US-born, and less likely to have received a bachelor’s or higher degree. In terms of neighborhood contexts, rural residents tended to live in neighborhoods with lower median household income; greater distance to the nearest 7 public parks; lower prevalence of residents walking, biking, or taking public transportation to work; and higher percentage of residents driving ≥1 hour per day to work. These differences were all significant at the P < .001 level. There was no difference in gender distribution by rural–urban residence.

Table 1.

Sample Statistics.

	Total Sample	Urban	Rural
	Mean, %	Mean, %	Mean, %
Individual level
Obese (defined as BMI ≥30)^a	34.36	32.33	40.77
Overweight (defined as BMI ≥25 and <30)^a	33.36	34.16	30.83
Normal weight (defined as BMI ≥18.5 and <25)^a	32.28	33.51	28.39
Age, mean, years^a	41.28	40.74	42.95
Gender: male, %	50.35	50.23	50.73
Race and ethnicity^a
Non-Hispanic white	68.88	63.72	85.12
Non-Hispanic black	11.85	13.81	5.68
Hispanics	13.30	15.72	5.70
Other	5.97	6.75	3.49
US-born, %^a	54.87	51.72	64.79
Education^a
Less than high school	16.63	16.23	17.89
High school graduates or equivalent	25.02	22.65	32.46
Some college	32.27	31.85	33.58
Bachelor’s degree or higher	26.03	29.19	16.07
Family income-to-needs ratio^a	3.12	3.21	2.83
Neighborhood level
Tract-level median household income (in the unit of US$1000)^a	84.54	81.13	93.74
Tract-level distance to the closest seven parks (in the unit of mile)^a	7.44	3.62	19.47
Tract-level percentage walking, biking, or taking public transportation to work^a	7.13	8.40	3.12
Tract-level percentage driving for at least 1 h/d to work among aged 16 and above^a	7.86	7.71	8.33
Sample size	10 302	8188	2114

Abbreviation: BMI, body mass index.

^a P < .001.

Table 2 presents the results from the 2-level random intercept logistic regression analyses of obesity (in odds ratio [OR]). Controlling for age, gender, race/ethnicity, and immigrant status, the baseline model showed that in the United States rural residents had 35.8% greater odds of obesity than urban residents. When separately examined, education (model 2) and family income (model 3) were both significant covariates; but when examined together (model 4), the significance of family income disappeared and the education effect remained. The neighborhood variables were mostly significant all showing effects in expected directions (models 4 and 5); that is, neighborhood median household income and walkability were negative covariates, and distance to parks and prevalence of long-commuting residents were positive ones for the odds of obesity.

Table 2.

Multilevel Logistic Regression Odds Ratios of Obesity Prevalence.^a

Variables	1	2	3	4	5	6
Rural vs urban	1.358^b	1.302^b	1.328^b	1.224^b	1.105	1.022
Age	1.017^b	1.018^b	1.017^b	1.018^b	1.017^b	1.017^b
Male	0.720^b	0.720^b	0.724^b	0.721^b	0.719^b	0.722^b
Non-Hispanic white	Ref	Ref	Ref	Ref	Ref	Ref
Non-Hispanic black	1.725^b	1.672^b	1.690^b	1.601^b	1.732^b	1.645^b
Hispanic	1.375^b	1.324^b	1.331^b	1.286^b	1.332^b	1.284^b
Other	0.676^b	0.681^b	0.667^b	0.671^b	0.704^c	0.698^c
US-born	1.166^c	1.156^c	1.177^c	1.153^c	1.161^c	1.160^c
Individual-level mediators
College or higher education
Less than high school		Ref		Ref	Ref	Ref
High school graduates or equivalent		1.122^d		1.145^e	1.106	1.134^e
Some college		1.121^d		1.158^e	1.105^d	1.152^e
Bachelor’s degree or higher		0.771^b		0.824^c	0.765^b	0.838^c
Income-to-needs ratio			0.968^e	1.001
Neighborhood-level mediators
Tract-level median household income (US$1000)				0.994^b		0.992^b
Tract-level distance to the closest seven parks (10 miles)					1.008^c	1.006^e
Tract-level % walking, biking, or taking public transportation to work (10%)					0.993^c	0.990^b
Tract-level % driving for at least 1 h/d to work among aged 16 and above (10%)					1.006	1.010^c
Percentage reduction in the rural–urban odds ratio		−15.6%	−8.38%	−37.43%	−70.67%	−93.85%

^aSample size = 10 302.

^b P < .001.

^c P < .01.

^d P < .10.

^e P < .05.

Model 6, the final model, included all the significant covariates in the final full-model analysis, showing that older age (OR = 1.017, P < .01), male gender (OR = 0.722, P < .001), non-Hispanic blacks (OR = 1.645, P < .001), Hispanics (OR = 1.284, P < .001), other group (OR = 0.698, P < .001), and US birthplace (OR = 1.160, P < .01) were individual-level significant factors of obesity. In addition, education had a curvilinear relation with obesity; compared to those without a high school diploma, respondents who had graduated from high school but had not received a bachelor’s degree were more likely to be obese (both ORs were significantly greater than 1), whereas college graduates were less likely to be obese (OR = 0.838; P < .001). At the neighborhood level, household median income (OR = 0.992, P < .001); distance to local parks (OR = 1.006, P < .05); prevalence of walking, biking, or taking active transportation to work (OR = 0.990, P < .001); and prevalence of long auto-commuting residents (OR = 1.010, P < .01), all remained significant in the final model.

The last row in this table presents the percentage reduction in the elevated odds of obesity associated with rural residence, signifying the mediating processes underlying the rural–urban obesity disparities. Adding education to the baseline model, the rural OR decreased from 1.358 to 1.302 (from model 1 to model 2), indicating that the increased odds of obesity for rural residents compared to urban residents was reduced by 15.6% ([30.2-35.8]/35.8) due to education. When income was added to model 1, the rural OR decreased from 1.358 to 1.328, a reduction of 8.38% in the elevated odds of obesity associated with rural residence (from model 1 to model 3). From model 1 to model 4, multilevel SES variables were added, including individual education, family income, and neighborhood median household income, corresponding to a 37.43% reduction in the elevated odds of obesity associated with rural residence. From model 1 to model 5, with the addition of individual education and neighborhood built environmental variables, the rural–urban residence and obesity link was attenuated by 70.67% and rendered insignificant. From model 1 to model 6, a complete mediation of the 5 mediators (individual education plus the 4 neighborhood variables) seemed to occur, exhibiting a 93.85% reduction in the elevated odds of obesity among rural residents compared to their counterpart and an OR vis-à-vis rural-urban residence (1.022) very close to 1 (indicating no association).

Table 3 presents results from the 2-level random intercept ordinal logit regression analyses of overweight and obesity combined prevalence. The general patterns of the main and mediating effects, in terms of both the effect size and direction, were very similar to those presented in Table 2. These findings thus seem to be robust to different ways of categorizing body weight as the outcome variable.

Table 3.

Multilevel Logistic Regression Odds Ratios of Overweight and Obesity Combined Prevalence.^a

Variables	1	2	3	4	5	6
Rural vs urban	1.311^b	1.271^b	1.302^b	1.221^b	1.076	1.015
Age	1.022^b	1.023^b	1.022^b	1.023^b	1.022^b	1.023^b
Male	0.916^c	0.917^c	0.918^c	0.918^c	0.919^c	0.924^c
Non-Hispanic white	Ref	Ref	Ref	Ref	Ref	Ref
Non-Hispanic black	1.733^b	1.722^b	1.723^b	1.657^b	1.758^b	1.691^b
Hispanic	1.597^b	1.594^b	1.582^b	1.552^b	1.583^b	1.540^b
Other	0.657^b	0.661^b	0.654^b	0.660^b	0.693^b	0.688^b
US-born	1.145^d	1.134^d	1.148^d	1.121^d	1.135^d	1.133^d
Individual-level mediators
College or higher education
Less than high school		Ref		Ref	Ref	Ref
High school graduates or equivalent		1.165^d		1.169^d	1.141^c	1.162^d
Some college		1.182^d		1.188^d	1.157^d	1.193^b
Bachelor’s degree or higher		0.847^d		0.862^c	0.838^d	0.895^e
Family income-to-needs ratio			0.991	1.016	1.016
Neighborhood-level mediators
Tract-level median household income (US$1000)				0.996^d		0.995^b
Tract-level distance to the closest seven parks (10 miles)					1.007^d	1.006^c
Tract-level % walking, biking, or taking public transportation to work (10%)					0.992^b	0.989^b
Tract-level % driving for at least 1 h/d to work among aged 16 and above (10%)					1.008^d	1.011^d
Percentage reduction in the rural–urban odds ratio		−12.86%	−2.89%	−28.94%	−75.56%	−95.18%

^aSample size = 10 302.

^b P < .001.

^c P < .05.

^d P < .01.

^e P < .10.

Discussion

The main purpose of this study was to explore patterns and mechanisms of rural–urban disparities in obesity among working age adults in the United States. We found that the higher prevalence of excess body weight in rural areas was explained by individual-level SES (particularly education) as well as neighborhood-level variables including built environmental features and median household income. These results clearly show the importance of neighborhood contexts in contributing to rural–urban disparities in obesity.

To the best of our knowledge, this study is the first to directly examine individual- and neighborhood-level mediators of overweight/obesity disparities between rural and urban adult Americans, using objectively measured height and weight to calculate BMI and categorize overweight/obesity status. A recent study also used NHANES data to report rural–urban obesity disparities which were found to be larger than what had been previously reported based on self-reported height and weight¹²; but the study did not examine potential mediators explaining this phenomenon. Another study specifically examined the role of county-level persistent poverty in rural–urban and within-rural obesity disparities, but it did not capture local environmental contexts.¹³ Clearly, more studies need to be done to investigate why rural residents consistently exhibit higher overweight/obesity prevalence compared to their urban counterparts in the United States.

In our analyses, we included the most basic sociodemographic factors such as age, gender, race/ethnicity, immigrant status, education, and family income but not behavioral factors such as diet and physical activity under the following considerations. First, as noted earlier, findings reported in previous studies are mixed as to how diet^12,17 and physical activity²⁴ may vary according to rural–urban residence. Empirically, it is not a clear-cut case that rural residents have greater energy intake and less energy expenditure thereby experiencing higher obesity prevalence than urban residents. It can be expected that any systematic environmental effect on obesity would operate via either diet or physical activity or some sorts of combination of both, given the energy balance etiology of obesity.³⁶ However, previous studies found weak mediating roles of caloric intake and physical activity in contributing to obesity disparities. For example, 1 study, using the 2003 to 2006 NHANES data, found that neither total caloric intake nor total moderate–vigorous physical activity could explain obesity disparities by ethnicity and immigrant status.³⁷ This puzzling finding may have resulted from the inherent difficulties in accurately measuring diet and physical activity variables. In NHANES, caloric intake was captured by information from a 2-day dietary intake diary, which was inevitably subject to response and recall bias. Although objective physical activity data were collected using accelerometer in the 2003 to 2006 NHANES, these measures were not free of bias or error. Approximately 40% of the 2003 to 2004 and 2005 to 2006 continuous NHANES sample were missing accelerometer data, and a considerable number of cases needed to be dropped from the analysis due to invalid accelerometer readings or too much time of not wearing the device.³⁸ Therefore, while we acknowledge that energy balance–related factors such as diet and physical activity are likely to be the ultimate determinants of body weight; for the purpose of this study, excluding diet and physical activity from the analysis seems to be a reasonable strategy.

One noteworthy side finding regarding individual-level factors is that education but not family income showed a strong association with overweight/obesity and explained a greater portion of the rural–urban residence and overweight/obesity association than family income. For health behaviors, education has been found to act as the most consistent predictor compared to other main SES indicators such as income, employment, and occupation—probably via its unique positive impact on psychosocial pathways such as beliefs about personal control and lifestyle-related self-efficacy.³⁹ By contrast, evidence on the income and obesity link is mixed based on studies conducted in developed countries⁴⁰; the magnitude and direction of the income–obesity relationship in the United States seem to vary according to gender and race–ethnicity.⁴¹ These complex forces may have contributed to the weaker effect of family income relative to that of education observed in this study.

In addition, it needs to be pointed out that NHANES income measure was a very simplified crude measure subject to measurement error. By contrast, neighborhood income was arguably a more sophisticated income measure as it took into consideration government transfer. In this study, we found neighborhood median income was a more consistently significant covariate of the odds of excess body weight than family income, a finding in line with previously published evidence based on studies conducted in developed countries.⁴⁰

In any event, the strong neighborhood economic effect confirms the importance of contextual material resources on individuals’ body weight status buttressing the recent attention on the obesity and poverty link.^42,43 Moreover, higher income neighborhoods may also be advantaged in terms of having greater stock of social capital, less environmental stress, and lower crime rates.^44,45 In addition to neighborhood median household income, neighborhood accessibility to local parks, walkability, and land use patterns were all significant covariates of individual odds of overweight/obesity. They also jointly helped to explain the rural–urban disparities. These 3 variables all tapped neighborhood physical activity related features influencing individuals’ energy expenditure.

Several important study limitations need to be discussed. First and foremost, the study is cross-sectional and findings on the associations cannot disentangle causal directions. Selection bias is inevitable in this study design. Second, the measurements of neighborhood built environments were crude. For example, land use pattern and walkability were not directly measured but captured using proxy indicators such as proportion of workers walking, biking, or taking public transportation to work. That said, the external validity of these census-based built environmental variables has been confirmed for objectively measured moderate to vigorous physical activity.⁴⁶ Third, some important individual-level factors were omitted. For example, health beliefs and self-efficacy related to physical activity or diet were not included in the analyses due to data limitation. They are not only likely determinants of body weight status in their own right but also plausible mediators explaining urban–rural disparities in overweight and obesity due to area-based weight-related perceptional and normative differences. And fourth, important neighborhood factors such as food environments were not examined in this study. In fact, the relationship between neighborhood food environment and obesity has been questioned due to inconsistent findings from local or regional studies^47
–49 and overall concerns with the food environmental data quality.⁵⁰ It is challenging to assess the food environment and obesity link because of families’ nonlocal shopping and travel patterns. Although nationally representative studies are advantaged in the wide coverage and broader study generalizability compared to local or regional studies, they are also constrained by limited quality environmental data covering the entire country.

A take-home message from this study is that rural–urban overweight/obesity disparities are large and explained by neighborhood economic and built environment as well as individual socioeconomic (particularly education) differences across rural and urban residents. Among these mediators, built environmental factors are arguably more modifiable than others. Neighborhood income and individual-level education clearly influence lifestyle and weight status but presumably entail complex solutions to change. Based on findings from this study, public policy efforts in reducing rural–urban obesity disparities need to consider targeting rural physical activity environments as a focal point for intervention. Perhaps 1 way to reduce overweight/obesity risks for rural residents is to modify rural areas’ built environment to make them more inviting to physical activity. Local built environmental initiatives such as building well-equipped public parks and safe walk tracks⁵¹ may be 1 viable venue promoting physical activity among rural residents who, especially those of working age, are nowadays doing less physically demanding field work compared to previous generations of farmers thereby losing high-caloric expenditure opportunities through agricultural work. Future research should tap into other potential mediators of rural–urban obesity disparities such as beliefs and norms at the individual level and food environments at the neighborhood level. Representative longitudinal data are needed to better study and understand the patterns and mechanism of rural–urban overweight/obesity disparities in the United States and make evidence-based intervention recommendations to effectively mitigate these disparities.

Footnotes

Acknowledgment

We thank Xingyou Zhang who provided insight and expertise that greatly assisted the GIS data construction.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the National Institute of General Medical Sciences of the National Institutes of Health under award number R01CA140319-01A1.

References

U.S. Department of Agriculture Economic Research Service . Rural-urban Commuting Area Codes. https://www.ers.usda.gov/data-products/rural-urban-commuting-area-codes/. Accessed January 16, 2017.

Cossman

James

Cosby

Cossman

. Underlying causes of the emerging nonmetropolitan mortality penalty. Am J Public Health. 2010;100(8):1417–1419.

Eberhardt

Ingram

Makuc

. Urban and Rural Health Chartbook. Health, United States, 2001. Hyattsville, Maryland: National Center for Health Statistics; 2011.

Mainous

AGI

Kohrs

. A comparison of health status between rural and urban adults. J Comm Health. 1995;20(5):423–431.

McIntosh

Sobal

. Rural eating, diet, nutrition, and body weight. In: Glasgow

Morton

Johnson

, eds. Critical Issues in Rural Health. Ames, IA: Blackwell; 2004:113–126.

Downey

. Rural populations and health: Determinants, disparities, and solutions. Prev Chronic Dis. 2013;10:130097.

Crosby

Wendel

Vanderpool

Casey

, eds. Rural Populations and Health: Determinants, Disparities, and Solutions. San Francisco, CA: Jossey-Bass; 2012.

Brumby

Kennedy

Chandrasekara

. Alcohol consumption, obesity, and psychological distress in farming communities – an Australian study. J Rural Health. 2013;29(3):311–319.

Martin

Kirkner

Mayo

Matthews

Larry

Hebert

. Urban, rural, and regional variations in physical activity. J Rural Health. 2005;21(3):239–244.

10.

Jackson

Doescher

Jerant

Hart

. A national study of obesity prevalence and trends by type of rural county. J Rural Health. 2005;21(2):140–148.

11.

Patterson

Moore

Probst

Shinogle

. Obesity and physical inactivity in rural America. J rural Health. 2004;20(2):151–159.

12.

Befort

Nazir

Perri

. Prevalence of obesity among adults from rural and urban areas of the United States: findings from NHANES (2005-2008). J Rural Health. 2012;28(4):392–397.

13.

Bennett

Probst

Pumkam

. Obesity among working age adults: the role of county-level persistent poverty in rural disparities. Health Place. 2011;17(5):1174–1181.

14.

Council of Economic Advisers. Strengthening the Rural Economy – The Current State of Rural America. https://www.whitehouse.gov/administration/eop/cea/factsheets-reports/strengthening-the-rural-economy/the-current-state-of-rural-america. Accessed January 16, 2017.

15.

Fisher

. Why is U.S. poverty higher in nonmetropolitan than in metropolitan areas? Growth Change. 2007;38(1):55–76.

16.

Bove

Olson

. Obesity in low-income rural women: qualitative insights about physical activity and eating patterns. Women Health. 2006;44(1):57–78.

17.

Champagne

Bogle

McGee

. Dietary intake in the lower Mississippi delta region: results from the Foods of our Delta Study. J Am Diet Assoc. 2004;104(2):199–207.

18.

Cohen

Rehkopf

Deardorff

Abrams

. Education and obesity at age 40 among American adults. Soc Sci Med. 2013;78:34–41.

19.

de Franca

Gigante

Olinto

. Binge eating in adults: prevalence and association with obesity, poor self-rated health status and body dissatisfaction. Public Health Nutr. 2014;17(4):932–938.

20.

Hill

Wyatt

Peters

. Energy balance and obesity. Circulation. 2012;126(1):126–132.

21.

Lee

Andrew

Gebremariam

Lumeng

Lee

. Longitudinal associations between poverty and obesity from birth through adolescence. Am J Public Health. 2014;104(5):e70–e76.

22.

Marmorstein

Iacono

Legrand

. Obesity and depression in adolescence and beyond: reciprocal risks. Int J Obes. 2014;38(7):906–911.

23.

Prosper

Moczulski

Qureshi

. Obesity as a predictor of self-rated health. Am J Health Behav. 2009;33(3):319–329.

24.

Fan

Wen

Kowaleski-Jones

. Rural-urban differences in objective and subjective measures of physical activity: findings from the National Health and Nutrition Examination Survey (NHANES) 2003-2006. Prev Chronic Dis. 2014;11:E141.

25.

Wen

Maloney

. Neighborhood socioeconomic status and BMI differences by immigrant and legal status: Evidence from Utah. Econ Hum Biol. 2014;12:120–131.

26.

Larson

Story

Nelson

. Neighborhood environments: disparities in access to healthy foods in the U.S. Am J Prev Med. 2009;36(1):74–81.

27.

Lovasi

Hutson

Guerra

Neckerman

. Built environments and obesity in disadvantaged populations. Epidemiol Rev. 2009;31:7–20.

28.

Wang

Wen

. Population-adjusted street connectivity, urbanicity and risk of obesity in the U.S. Appl Geogr. 2013;41:1–14.

29.

Wen

Kowalesi-Jones

. The built environment and risk of obesity in the United States: racial-ethnic disparities. Health Place. 2012;18(6):1314–1322.

30.

Thornton

Crawford

Cleland

Timperio

Abbott

Ball

. Do food and physical activity environments vary between disadvantaged urban and rural areas? Findings from the READI Study. Health Promot J Austr. 2009;23(2):153–156.

31.

Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. 2016. http://www.cdc.gov/nchs/nhanes.htm. Accessed January 16, 2017.

32.

ArcGIS [computer program]. Version 9.3. Redlands, CA: ESRI; 2006.

33.

Zhang

Holt

. Modeling spatial accessibility to parks: a national study. Int J Health Geogr. 2011;10:31.

34.

Mathieu

Taylor

. Clarifying conditions and decision points for mediational type inferences in organizational behavior. J Organ Behav. 2006;27:1031–1056.

35.

Boardman

Saint Onge

Rogers

Denney

. Race differentials in obesity: the impact of place. J Health Soc Behav. 2005;46(3):229–243.

36.

Weinsier

Hunter

Heini

Goran

Sell

. The etiology of obesity: relative contribution of metabolic factors, diet, and physical activity. Am J Med. 1998;105(2):145–150.

37.

Wen

Kowaleski-Jones

Fan

. Ethnic-immigrant disparities in total and abdominal obesity in the US. Am J Health Behav. 2013;37(6):807–818.

38.

Fan

Brown

Hanson

Kowaleski-Jones

Smith

Zick

. Moderate to vigorous physical activity and weight outcomes: does every minute count? Am J Health Promot. 2013;28(1):41–49.

39.

Mirowsky

Ross

Reynolds

. Links between social status and health status. In: Bird

Conrad

Fremont

, eds. Handbook of Medical Sociology. Upper Saddle River, NJ: Prentice Hall; 2000:47–67.

40.

McLaren

. Socioeconomic status and obesity. Epidemiol Rev. 2007;29:29–48.

41.

Chang

Lauderdale

. Income disparities in body mass index and obesity in the United States, 1971-2002. Arch Intern Med. 2005;165(18):2122–2128.

42.

Drewnowski

. The role of energy density. Lipids. 2003;38(2):109–115.

43.

Drewnowski

Specter

. Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr. 2004;79(1):6–16.

44.

Kawachi

Berkman

. Neighborhoods and Health. New York, NY: Oxford University Press; 2003.

45.

Sampson

Morenoff

Gannon-Rowley

. Assessing “neighborhood effects: social processes and new directions in research. Annu Rev Sociol. 2002;28:443–478.

46.

Brown

Smith

Hanson

Fan

Kowaleski-Jones

Zick

. Neighborhood design for walking and biking: physical activity and body mass index. Am J Prev Med. 2013;44(3):231–238.

47.

Sturm

. School and residential neighborhood food environment and diet among California youth. Am J Prev Med. 2012;42(2):129–135.

48.

Dubowitz

Ghosh-Dastidar

Eibner

. The women’s health initiative: the food environment, neighborhood socioeconomic status, BMI, and blood pressure. Obesity. 2012;20(4):862–871.

49.

Lee

. The role of local food availability in explaining obesity risk among young school-aged children. Soc Sci Med. 2012;74(8):1193–1203.

50.

Liese

Colabianchi

Lamichhane

. Validation of 3 food outlet databases: completeness and geospatial accuracy in rural and urban food environments. Am J Epidemiol. 2010;172(11):1324–1333.

51.

Sugiyama

Shibata

Koohsari

. Neighborhood environmental attributes and adults’ maintenance of regular walking. Med Sci Sports Exerc. 2015;47(6):1204–1210.