Geographic differences in the magnitude of black‐white disparities in having obesity

Abstract Background Obesity disparities in the United States are well documented, but the limited body of research suggests that geographic factors may alter the magnitude of these disparities. A growing body of evidence has identified a “rural mortality penalty” where morbidity and mortality rates are higher in rural than urban areas, even after controlling for other factors. Black‐White differences in health and mortality are more pronounced in rural areas than in urban areas. Objective Therefore, the purpose of this study was to explore how rural‐urban status and region moderate Black‐White health disparities in obesity. Methods Data were abstracted from the 2012 Behavioral Risk Factor Surveillance System, with the sample being restricted to Black and White respondents (n = 403,231). Respondents’ county of residence was linked to US Census information to obtain the county‐level Index of Relative Rurality (IRR) and Census division. Crude and adjusted logistic regression models were utilized to assess the magnitude of Black‐White disparities in having obesity (yes/no) by IRR quartile and by Census division. Results Overall, Black‐White differences in obesity were wider in rural than in urban counties, with a significant linear trend (p < 0.001). Furthermore, when stratified by US Census division, results revealed that disparities were significantly wider in rural than urban areas for respondents living in the Middle Atlantic and South Atlantic divisions. In contrast, the association was reversed for the remaining divisions (New England, East North Central, West North Central, Mountain, and Pacific), where the magnitude of the Black‐White difference was the largest in urban areas. Conclusion Findings highlight the need to understand and account for critical place‐based factors that exacerbate racial obesity disparities to develop and maximize the effectiveness of policies and programs designed to reduce racial inequalities and improve population health.

For example, de facto racial segregation is one way in which structural racism manifests.Segregation contributes to differential opportunities for employment, education quality and access, and homeownership. 30 can also create obesogenic environments and influence individual behaviors through the implementation of policies, planning, and zoning that can limit the availability of healthy food options and opportunities for exercise and physical activity due to a lack of investment and safety concerns. 31In addition to residential segregation, systemic racism impacts health behaviors that lead to higher levels of obesity and other adverse health outcomes through other pathways.These include but are not limited to, unfair lending practices and related barriers to home ownership and taxation, biased policing, and voter suppression. 32asuring how these manifestations of structural racism impact health and health behaviors is challenging, but important, and deserves further attention.That said, one recurring, plausible mechanism through which structural racism impacts health is the sustained physical toll of concentrated and psychosocial stressors across the lifespan, 33 which both directly impacts health, as well as individual health behaviors that can promote obesity.
In addition to those factors mentioned above, the overall magnitude of Black-White differences in obesity prevalence also varies substantially by geography. 34There is increasing recognition that place-based factors, such as neighborhood, community, and state characteristics, contribute to obesity rates and Black-White differences in obesity rates.These factors include, but are not limited to, neighborhood disadvantage, 35 segregation, 36,37 and access to healthy foods. 38For example, a study of a well-integrated, lowincome neighborhood found that racial disparities are reduced somewhat when community-level socioeconomic factors, such as income and education, are equalized. 39re broadly, other research suggests that policies, even those not directly aimed at affecting health outcomes, likely contribute to structural racism and racial disparities in obesity. 40One such example is zoning policies, which influence where business can be built in a community.Such policies can impact the availability of food, green space, and other services impact obesity. 40,41Additionally, state and community tax policies can alter the relative costs and ease of obtaining healthful versus unhealthful foods, with the availability of healthful foods more likely in more affluent, White neighborhoods. 42These zoning policies, land use regulations, and other financial incentives have been the focal point of considerable discussion in addressing the systematic causes of Black-White differences in obesity and other population health measures. 43other community attribute, rural-urban status, has also been shown to influence population health outcomes.Studies have identified a "rural mortality penalty", 44 where mortality rates are higher in rural than urban areas of the US, even after controlling for other factors, including SDH. 45 This rural mortality penalty extends to other health outcomes, including preventive health behaviors, 46 COVID-19 outcomes, 47 cancer screenings 48 and drug overdoses. 49wever, the driving forces behind the observed rural mortality penalty are not well understood.Fundamental cause theory suggests that rural areas may be more likely to provide or promote the underlying social conditions that give rise to more proximal causes of death and morbidity such as high poverty, unemployment, and lower education. 50Although rural areas tend to have higher levels of green space and better air and water quality, they may be less likely to have a built environment conducive to physical activity and recreation. 51Cultural aspects of rural areas may also play a role.
For example, food preparation methods and celebrations and events revolving around unhealthy foods are more common in rural areas of the US. 52,53Furthermore, evidence suggests that people living with obesity living in rural areas are more likely to perceive themselves as healthy and adopt fatalistic beliefs about weight and health compared to those living in non-rural areas. 54However, the individual contributions of these and other attributes of rural environments to the rural mortality penalty are unknown and deserve further study.
There is growing evidence that this "rural mortality penalty" not only contributes to higher levels of premature mortality and morbidity in rural areas but may also exacerbate other types of health problems.Recent studies have shown that Black-White differences in health and mortality are more pronounced in rural areas than urban areas.Black-White differences in mortality are widest in rural areas, and those differences have widened over time. 55Similar results have been found with obesity: rural areas have the highest rates of obesity, but also Black-White differences in the rates of having obesity are largest in rural areas. 56Furthermore, a 2021 study 57 found critical regional differences in the rural health and mortality penalty in the US, where rural mortality rates are worse in the rural south compared to other rural areas.However, to date, no research has examined how Black-White differences in obesity vary by both rural-urban status and US region.Therefore, the objective of this study was to examine regional differences in how the magnitude of Black-White differences in obesity vary by ruralurban status using a large, nationally representative sample of US adults.We hypothesized that the magnitude of Black-White differences would be greatest in rural areas, particularly in the South and Midwest.

| Data
This study is a secondary data analysis of the 2012 Behavioral Risk Factor Surveillance System (BRFSS), the largest network of healthrelated telephone surveys administered by the Centers for Disease Control and Prevention.The BRFSS collects data from US community dwelling residents 18+ years of age in all 50 states, as well as Guam and Puerto Rico, regarding their demographics, selfreported health-related behaviors, use of preventive services, and other health-related information, and guides planning and prevention efforts at the state and federal levels. 58Over 400,000 interviews with BRFSS respondents aged 18 and older are conducted annually.
This study used the 2012 BRFSS sample, the most recent year in which the respondent's place of residence (county) was collected.
The 2012 BRFSS included 475,687 total respondents, with response rates of 49.1% and 35.3% for landlines and cell phones, respectively. 59The analytic sample for the current study was restricted to respondents who provided information on height and weight and all other key study variables that were living in the contiguous US (lower 48 states), and who responded "White" or "Black" as their preferred race.The county of residence was not available in the 2012 BRFSS for residents of Alaska and Hawaii.The resultant sample size was 403,231 (84.8%).Each respondent was linked to area-level data from the 2010 US Census 60 and American Community Survey 5-year estimates 61 via county Federal Information Processing Standard code.

| Outcome variable
The main outcome variable of interest was obesity (yes vs. no).
Obesity status was determined by BMI, which was calculated using self-reported height and weight.Respondents with a BMI of 30 kg/ m 2 or above were classified as having obesity. 62

| Predictor variables
The main predictor variables were race, US Census division, and rural-urban status.Race was obtained and categorized in several ways in the BRFSS data.Respondents were asked "Which of these groups best represents your race?" Possible responses were "White", "Black or African American", "Asian", "Native Hawaiian or Other Pacific Islander", or American Indian/Alaska Native".The analytical sample was restricted to respondents reporting their race as being either "Black or African American" or "White".S1 (adapted from 63 ).
Rural-urban status was obtained through the Index of Relative Rurality (IRR), a continuous, composite measure of four measures of rural-urban characteristics-population size, population density, percent urban population, and distance to nearest metropolitan area. 64,657][68][69][70][71] For analysis, all US counties were categorized into IRR quintiles with a quintile of 1 being the most rural and a quintile of 5 being the most urban. 70,71

| Covariates and complex sampling
Other variables of interest included in this analysis were respondents' age (in years), sex (female, male), annual income category (<$25,000, $25,000-49,999, ≥$50,000, missing/not available), currently employed for pay (yes, no), education (bachelor's degree or higher vs. less than bachelor's degree), currently married (yes, no), and current smoker (yes, no).The BRFSS data set included the Centers for Disease Control and Prevention's analytic sample weights, which were used in all analyses to account for differences in sampling and response probabilities, in accordance with BRFSS guidelines. 59eighted generalized linear models with a logistic link function were used to assess the magnitude of the difference between Black and White respondents with respect to obesity by US Census division and IRR quintile.Four sets of models were obtained.First, Black-White differences were examined in the entire sample, both unadjusted and including covariates (age, sex, income, employment, marital status, and smoking status).Second, Black-White differences were stratified by IRR quintile and then by US Census division.Third, Black-White differences were then modeled for each IRR quintile-US Census division category.For the second and third sets of models, obesity (yes/no) was modeled both unadjusted and adjusted for covariates for which a set of propensity scores were initially created based on the covariates using the complete data set.These propensity scores were used to address confounding without using excessive degrees of freedom using the covariates above as predictors (age, sex, income, employment status, marital status, and smoking status).Respondents were then ranked according to their estimated propensity score and were stratified into subsets based on decile of the propensity score for the analysis in these models. 72stly, interaction terms were incorporated for each US Census division to assess the potential for monotonic trends in the association between the IRR quintile and the magnitude of the association comparing Blacks to Whites in having obesity using race*IRR interaction terms.For all models, the model fit was evaluated using pseudo-R-squared and Akaike Information Criteria statistics.SAS

| RESULTS
Descriptive statistics for the sample overall and by race (Black vs.   S2).

White) are shown in
Overall, within each US Census division, Black respondents were significantly more likely than White respondents to have obesity.Table 2 also indicates whether there was a monotonic association or linear trend between rural-urban status and the strength of the association between race and obesity within each Census division.Overall, and for the two Census divisions (Middle Atlantic and South Atlantic), the strength of the race-obesity association was significantly higher in rural areas than in urban areas.However, for five Census divisions (New England, East North Central, West North Central, Mountain, and Pacific), the strength of the race-obesity association was significantly higher in urban areas than in rural areas.

Odds ratios of having obesity comparing
The association was not statistically significant for both the East and West South Central divisions.
The results of this study validate previous findings identifying Black-White differences in obesity with respect to geography.Study results determined that within all US Census divisions, the differences in the prevalence of obesity between Blacks and Whites were statistically significant and that this difference was fairly consistent, with the difference ranging from 49% in the Middle Atlantic division to 79% in the Pacific Division.These findings are consistent with previous research, showing that not only did the overall levels of having obesity vary by US region, but also the magnitude of differences between Black and White respondents. 73The current study found that the prevalence of obesity was lowest in the Northeast region (New England and Middle Atlantic US Census divisions), where White T A B L E 1 Frequencies [N (weighted %)] for analytic sample from the Behavioral Risk Factor Surveillance System (BRFSS), 2012.respondents had a higher prevalence of obesity compared to Black respondents.This study also showed that, after adjusting for demographics and other health issues, Black respondents had higher rates of obesity in all other regions.

Overall
The present study's findings extend previous research by highlighting both rural-urban differences in the overall prevalence of having obesity, as well as variability in the magnitude of the Black-White difference in the prevalence of obesity. 74Studies using a dichotomous measure of rural-urban status found that rural residents were more likely to have obesity than urban residents, after adjusting for demographic and social factors. 10,12,74Another study found a similar rural-urban difference in overall levels of obesity, and determined that health behaviors, such as dietary quality and physical inactivity, were also lower in rural areas compared to more urban areas. 75The present study's main contribution is to illustrate critical geographic and regional differences in the associations between rural-urban status and the magnitude of the Black-White difference in obesity.In five of the nine US Census divisions-New England, East North Central, West North Central, Mountain, and Pacific-the magnitude of Black-White differences in obesity was significantly larger in urban areas than in rural areas.However, the association was reversed when examining the sample as a whole and in the Middle Atlantic and South Atlantic divisions.Here, the magnitude of Black-White differences in obesity was significantly larger in rural areas than in urban areas.There is considerable previous literature indicating regional differences in the US for various health outcomes.
These include COVID-19 cases and deaths, 77 cardiovascular disease, 78-80 cancer, 81,82 drug overdoses, 83 and general health. 84One study examined four diseases-cancer, stroke, cardiovascular disease, and chronic obstructive pulmonary disease (COPD)-and found that the factors that predicted the overall level of disease varied substantially by region. 85yond obesity, a more limited set of studies has examined variability in the magnitude of Black-White differences by geography for other health outcomes, such as pre-term birth rates 86 and coronary heart disease (CHD). 90Although Black women experienced higher levels of pre-term birth, there was substantial variability in rates based on both place of residence (US region) and the type of community (rural vs. urban).There were notable regional differences, as well: the magnitude of the Black-White difference in pre-term birth rates was significantly higher in the Northeast, South, and Midwest regions compared to the West. 86In a study examining temporal changes in mortality from CHD by US division, 90 CHD  mortality decreased in all divisions for both Black and White populations.However, the temporal rate of decline was substantially faster for White populations than for Black populations in four US divisions (Pacific, Mid-Atlantic, East North Central, and West North Central).The approach and findings of the present study expand upon these studies and highlight important and highly nuanced geographic differences in the prevalence of obesity by race that would otherwise be masked if assessing those racial differences for the country as a whole.
There are numerous potential explanations for why these observed associations and differences occurred.One such explanation is regional differences in the manifestation of structural racism across the country.By definition, structural racism is based on system-wide, large-scale cultural, socioeconomic, and political forces that result in health inequities. 27,29However, the manifestation of structural racism and how populations experience it with respect to health outcomes may vary regionally. 87It could be inferred that racism in the regions in which obesity disparities were highest in rural areas (e.g., Middle Atlantic and South Atlantic) may be more pronounced in rural areas than in urban areas.These factors may, in turn, create more obesogenic environments in rural areas, particularly for Black populations, which may experience greater levels of stress and lower levels of security due to those long-term cultural, social, and political factors. 88In other areas, such as New England, racism may be more strongly experienced in urban areas, particularly in historically Black communities, which may have been neglected through decades of harmful policies 33 and cultural factors, such as "White flight" from urban areas and de facto segregation. 89More research is needed to identify and address the root causes of these stark geographic differences in Black-White disparities in obesity by rural-urban status.
The findings of the present study should be interpreted in the context of several important limitations.First, as the data were cross-sectional, causation cannot be assessed.Second, the sample had a greater proportion of respondents from the most urban quintiles of US counties, which limited statistical power in the most rural counties (Q1).Third is the "modifiable area unit problem", which identifies critical spatial variability with respect to the size and physical layout of counties across the country, which can result in statistical bias of model-based estimates. 90,91Fourth, the level of spatial aggregation for rural-urban status was done at the county level.Previous research suggests that geographic associations may vary by the level of spatial aggregation utilized. 92,93Relatedly, the analysis did not account for potential spatial autocorrelation because not all counties were represented in the BRFSS sample.
Next, rural-urban status is a complex and multi-dimensional characteristic. 94Although the measure used to characterize rural-urban status, the IRR, 64 considers four aspects of rural-urban status and has been validated in prior studies, [66][67][68][69][70][71] there may be other factors associated with rurality that were not included in this measure.
Furthermore, analyses were restricted to non-Hispanics to reduce the potential for bias stemming from self-reports and misclassification of Hispanic ethnicity, 90 so the findings cannot be extended to those identifying as Hispanic.Another consideration is that the data were from 2012, and obesity rates have likely increased since the data were collected, in addition to other changes in SDH during that period.Data from 2012 were used because it was the most recent year in which the BRFSS included county-level geographic identifiers, which were used to spatially link each subject to Census data used in the study.Lastly, only a limited set of confounders were included in the multivariable analysis, and the associations between the confounders and the main exposure variable were evaluated in the entire sample.This approach leads to two limitations: residual confounding and the potential for the associations between each of the confounders and the main exposure variable (race) to vary by region.
Despite these limitations, the study has a number of notable strengths.First, the study is the first to evaluate the magnitude of Black-White differences in having obesity individually and jointly by rural-urban status and geographic region using a large, nationally representative dataset.The observed geographic variability in the associations between rural-urban status and Black-White differences in obesity underscores the critical need to consider the geographic context when creating and implementing health policies and programs designed to reduce such differences and promote health equity.
To maximize the effectiveness of any such policies and programs to reduce disparities and promote health equity, the specific, arealevel causes and contextual factors of such disparities must be understood and addressed.Those causes and contextual factors may differ across regions, so one-size-fits-all approaches may be limited in effectiveness.
Another important strength is the assessment of the association between rural-urban status and obesity across US divisions using rural-urban status as an ordinal, five-level variable and considering both monotonic and non-monotonic associations.5][96][97] Results of the present study suggest that although there may be overall trends toward better health outcomes in rural or urban areas, depending upon the region, many such associations were muted in the most rural and remote areas, resulting in a J-shaped association.More research is needed to validate these findings and determine what potentially modifiable factors contribute to these non-monotonic associations.
Overall, the study findings corroborate the vast majority of previous literature showing that Black-White differences in obesity are pervasive and persistent across geographies.However, the magnitude of those differences and the overall prevalence of obesity varied substantially by both rural-urban status and region.Additional research is needed to better understand the community-level drivers of these associations at the local level and provide a more comprehensive understanding of why these differences occur.Mixed methods approaches using qualitative information from key community stakeholders could provide insights into critical cultural, socioeconomic, and environmental factors that may vary from place to place.Findings from such research could uncover highly influential Effective measures to target those inequities may require a deeper, systematic understanding and addressing of these micro-level factors that likely vary by geography.For example, efforts to create healthier, less obesogenic built environments by promoting physical activity (e.g., parks, green space, sidewalks, etc.) should consider other factors that could inhibit physical activity, such as crime rates or perceived safety, or traffic patterns, 98 which themselves may vary by neighborhood or even by block. 99The results of this study suggest that efforts to reduce obesity and obesity-related health disparities that may work well in one area may not necessarily work in another seemingly similar area due to these underlying factors.Such broadbrush, "one-size-fits-all" efforts may be strengthened by understanding the unique needs of different populations in diverse settings across the country and adapting those efforts to meet those needs of populations at the local level.
Descriptive statistics were obtained for all study variables, including weighted means and standard deviations or medians and interquartile ranges for all continuous and discrete variables, and weighted frequencies and percentages for all categorical variables.Bivariate associations were obtained through the use of chi squared tests, t-tests, Wilcoxon rank sum tests, and Pearson and Spearman correlation coefficients, depending on the variable.Differences in the distribution of respondents by US Census division and rural-urban status (IRR quintile) were explored using crosstabulations and descriptive graphs.Black-White differences were tabulated across all 518 -COHEN ET AL. study variables, including the outcome of interest (having obesity) and geographic indicators (US Census division and IRR quintile).
version 9.4 (Cary, NC) and IBM SPSS version 28 (Armonk, NY) were used to analyze the data.Statistical significance was set at p < 0.05 for all analyses.
The distribution of respondents by IRR quintile varied by US Census division (Figure1).The percentage of respondents residing in the most urban quintile (Q5) ranged from 41.7% in the East South Central division to 74.3% in the Pacific division.The percentage living in the most rural quintile (Q1) ranged from 0.2% in the Middle Atlantic division to 8.0% in the Mountain division.Weighted percentages of obesity by race, US Census division, and IRR quintile are shown in Figure 2.For each division and level of ruralurban status, Black respondents consistently had higher percentages of having obesity compared to their White counterparts (Panel A).However, respondents living in the Mountain division tended to have the lowest rates of obesity and the smallest Black-White differences, regardless of the IRR quintile.Overall, the percentage of respondents having obesity increased with increasing rurality among both Black and White respondents (Panel B).Adjusted odds ratios (OR) of obesity within each IRR quintile compared to Whites in the most urban quintile (Q5) are shown in Supplemental Figure S1.The association between obesity and rural-urban status varied by Census division (Figure 2, Panel C).For example, for both Black and White respondents, there was an upside-down J-shaped association between obesity and the IRR quintile for some of the divisions, including the East North Central, West North Central, and the South Atlantic.These trends indicate that the prevalence of obesity was lowest in the most rural and most urban areas, and highest in the areas of intermediate rural-urban status.However, in the East South Central division, the association was monotonic, indicating that obesity increased with increasing rurality.Furthermore, there was substantial variation in both the overall level of obesity and the magnitude of the Black-White difference in obesity by both division and IRR quintile.The weighted and adjusted OR comparing Black to White respondents odds of having obesity overall, and jointly by US Census division and IRR quintile are shown in Table 2.For instance, the 1.86 odds ratio for the Pacific division in the most urban IRR quintile (Q5) indicates that among those in the most urban counties living in the Pacific division (California, Oregon, and Washington), Black respondents were 86% more likely to have obesity than their White counterparts, after adjusting for covariates (Supplemental Table Black respondents to White respondents ranged from 1.51 (95% CI 1.49, 1.53) in the Middle Atlantic division to 1.79 (95% CI 1.76, 1.82) in the Pacific division.In all US counties, regardless of division, the strength of the association between race and obesity was lowest (OR 1.41, 95% CI 1.34, 1.48) in the most rural quintile (Q1), and highest in the second-most rural quintile (Q2) (OR 1.87, 95% CI 1.82, 1.93).

F I G U R E 1
Percentage of respondents within each US Census division from each of the Index of Relative Rurality quintiles.COHEN ET AL.Furthermore, the present study was the first to examine Black-White differences in obesity by rural-urban status.Overall, the Black-White differences in the likelihood of having obesity were significantly larger in rural areas than in urban areas.It should be noted that among the most rural respondents (Q1), the magnitude of the Black-White difference was actually the smallest of all the IRR quintiles (OR 1.41), indicating a J-shaped association when examining the entire sample.Previous research suggests that despite a general trend toward worse health outcomes in rural areas, the most rural and remote populations may have a lower risk of obesity56 and improved (increased) rates of cancer survivorship76 compared to other populations living in less rural areas.The present study also found this J-shaped association between rural-urban status and obesity prevalence as well as the magnitude of Black-White differences in obesity.

E 2 2
Prevalence of obesity by race (Black and White) by US Census division (Panel A), by Index of Relative Rurality quintile (Panel B), and jointly by US Census division and Index of Relative Rurality quintile (Panel C).Prevalence of obesity and odds ratios (OR) (with 95% confidence intervals) of obesity comparing Black respondents to White respondents (referent group) by region, overall and by Index of Relative Rurality quintile (Q1-Q5) a .

524 -
COHEN ET AL.local attributes that may help explain variability in these findings on a finer geographic scale.Creating effective policies, programs, and interventions needs to consider the community factors such as culture, social determinants, and environmental factors that give rise to these complex patterns of health inequities.The results of this research underscore the notion that the community-based factors that promote or impede health and health disparities in obesity likely differ from place to place.

Table 1 .
Just over one-quarter of the sample (27.0%) had obesity, with 25.6% of White respondents and 38.4% of Black respondents having obesity.The mean age of the analytic sample was 56.3 years, and most of the sample was female (60.2%), and the Mountain (13.3%) divisions.All associations comparing Black and White respondents were statistically significant (p < 0.001).