[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Model-Predicted Prevalence Estimates of Obesity
Model-Predicted Prevalence Estimates of Obesity

Extreme (A) and moderate (B) obesity among adults by county, United States, 2012 Behavioral Risk Factor Surveillance System. Classes divided based on quantile breaks.

Figure 2.  Local Indicators of Spatial Association Cluster Maps for Obesity
Local Indicators of Spatial Association Cluster Maps for Obesity

Extreme (A) and (B) moderate obesity among adults by county in the contiguous United States, 2012. High-high indicates clustering of high prevalence; low-low, clustering of low prevalence; low-high, counties with low prevalence adjacent to those with high prevalence; and high-low, counties with high prevalence adjacent to those with low prevalence.

Table 1.  Weighed Prevalence Estimates of Extreme and Moderate Obesity by State, United States, BRFSS 2012a
Weighed Prevalence Estimates of Extreme and Moderate Obesity by State, United States, BRFSS 2012a
Table 2.  County-Level Estimates and Weighted State-Level Estimates of Prevalence of Extreme and Moderate Obesity Among Adults With Model-Predicted Estimates, United States, 2012 BRFSSa
County-Level Estimates and Weighted State-Level Estimates of Prevalence of Extreme and Moderate Obesity Among Adults With Model-Predicted Estimates, United States, 2012 BRFSSa
1.
Flegal  KM, Kruszon-Moran  D, Carroll  MD, Fryar  CD, Ogden  CL.  Trends in obesity among adults in the United States, 2005 to 2014.   JAMA. 2016;315(21):2284-2291. doi:10.1001/jama.2016.6458 PubMedGoogle ScholarCrossref
2.
Fryar  CD, Carroll  MD, Ogden  CL. Prevalence of overweight, obesity, and extreme obesity among adults aged 20 and over: United States, 1960-1962 through 2013-2014. National Center for Health Statistics; 2016.
3.
Hales  CM, Fryar  CD, Carroll  MD, Freedman  DS, Ogden  CL.  Trends in obesity and severe obesity prevalence in US youth and adults by sex and age, 2007-2008 to 2015-2016.   JAMA. 2018;319(16):1723-1725. doi:10.1001/jama.2018.3060 PubMedGoogle ScholarCrossref
4.
Pan  L, Freedman  DS, Gillespie  C, Park  S, Sherry  B.  Incidences of obesity and extreme obesity among US adults: findings from the 2009 Behavioral Risk Factor Surveillance System.   Popul Health Metr. 2011;9(1):56. doi:10.1186/1478-7954-9-56 PubMedGoogle ScholarCrossref
5.
World Health Organization. Obesity: preventing and managing the global epidemic. World Health Organization. Published 2000. Accessed June 8, 2017. https://www.who.int/nutrition/publications/obesity/WHO_TRS_894/en/
6.
Yang  L, Colditz  GA.  Prevalence of overweight and obesity in the United States, 2007-2012.   JAMA Intern Med. 2015;175(8):1412-1413. doi:10.1001/jamainternmed.2015.2405 PubMedGoogle ScholarCrossref
7.
Sturm  R, Hattori  A.  Morbid obesity rates continue to rise rapidly in the United States.   Int J Obes (Lond). 2013;37(6):889-891. doi:10.1038/ijo.2012.159PubMedGoogle ScholarCrossref
8.
Freedman  DS, Khan  LK, Serdula  MK, Galuska  DA, Dietz  WH.  Trends and correlates of class 3 obesity in the United States from 1990 through 2000.   JAMA. 2002;288(14):1758-1761. doi:10.1001/jama.288.14.1758 PubMedGoogle ScholarCrossref
9.
Keating  C, Backholer  K, Gearon  E,  et al.  Prevalence of class-I, class-II and class-III obesity in Australian adults between 1995 and 2011-12.   Obes Res Clin Pract. 2015;9(6):553-562. doi:10.1016/j.orcp.2015.02.004 PubMedGoogle ScholarCrossref
10.
Mendy  VL, Vargas  R, Cannon-Smith  G, Payton  M.  Overweight, obesity, and extreme obesity among Mississippi adults, 2001-2010 and 2011-2015.   Prev Chronic Dis. 2017;14:E49. doi:10.5888/pcd14.160554 PubMedGoogle Scholar
11.
Holt  JB, Zhang  X. Geospatial data methods for estimating population health outcomes. Webinar presented at: Council of State and Territorial Epidemiologists; August 2013; Atlanta, GA.
12.
Zhang  Z, Zhang  L, Penman  A, May  W.  Using small-area estimation method to calculate county-level prevalence of obesity in Mississippi, 2007-2009.   Prev Chronic Dis. 2011;8(4):A85.PubMedGoogle Scholar
13.
Jia  H, Muennig  P, Borawski  E.  Comparison of small-area analysis techniques for estimating county-level outcomes.   Am J Prev Med. 2004;26(5):453-460. doi:10.1016/j.amepre.2004.02.004 PubMedGoogle ScholarCrossref
14.
Pierannunzi  C, Xu  F, Wallace  RC,  et al.  A methodological approach to small area estimation for the Behavioral Risk Factor Surveillance System.   Prev Chronic Dis. 2016;13:E91. doi:10.5888/pcd13.150480 PubMedGoogle Scholar
15.
Song  L, Mercer  L, Wakefield  J, Laurent  A, Solet  D.  Using small-area estimation to calculate the prevalence of smoking by subcounty geographic areas in King County, Washington, Behavioral Risk Factor Surveillance System, 2009-2013.   Prev Chronic Dis. 2016;13:E59. doi:10.5888/pcd13.150536 PubMedGoogle Scholar
16.
Dwyer-Lindgren  L, Freedman  G, Engell  RE,  et al.  Prevalence of physical activity and obesity in US counties, 2001-2011: a road map for action.   Popul Health Metr. 2013;11:7. doi:10.1186/1478-7954-11-7 PubMedGoogle ScholarCrossref
17.
Centers for Disease Control and Prevention (CDC). The BRFSS Data User Guide. Published August 15, 2013. Accessed November 15, 2017. https://www.cdc.gov/brfss/data_documentation/pdf/userguidejune2013.pdf
18.
 American Community Survey Information Guide. United States Census Bureau; 2013.
19.
The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. EQUATOR Network. Accessed July 4, 2019. https://www.equator-network.org/reporting-guidelines/strobe/
20.
Slack  T, Myers  CA, Martin  CK, Heymsfield  SB.  The geographic concentration of US adult obesity prevalence and associated social, economic, and environmental factors.   Obesity (Silver Spring). 2014;22(3):868-874. doi:10.1002/oby.20502 PubMedGoogle ScholarCrossref
21.
Befort  CA, Nazir  N, Perri  MG.  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. doi:10.1111/j.1748-0361.2012.00411.x PubMedGoogle ScholarCrossref
22.
Talbot  TO. Geographic Aggregation Tool; R version 1.33, user guide. 2015. Environmental Health Surveillance Section, New York State Department of Health. Accessed April 3, 2020. https://cdn.ymaws.com/www.cste.org/resource/resmgr/scale/GAT_vR13_guide.pdf
23.
Zhang  X, Holt  JB, Lu  H,  et al.  Multilevel regression and poststratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the behavioral risk factor surveillance system.   Am J Epidemiol. 2014;179(8):1025-1033. doi:10.1093/aje/kwu018 PubMedGoogle ScholarCrossref
24.
Mukhopadhyay  P, McDowell  A. Small area estimation for survey data analysis using SAS Software. SAS Global Forum; 2011. Accessed December 15, 2018. http://support.sas.com/resources/papers/proceedings11/336-2011.pdf
25.
Brewer  CA, Pickle  L.  Evaluation of methods for classifying epidemiological data on choropleth maps in series.   Ann Assoc Am Geogr. 2002;92(4):662-681. doi:10.1111/1467-8306.00310 Google ScholarCrossref
26.
Anselin  L.  Local indicators of spatial association—LISA.   Geogr Anal. 1995;27(2):93-115. doi:10.1111/j.1538-4632.1995.tb00338.x Google ScholarCrossref
27.
Anselin  L, Syabri  I, Kho  Y. GeoDa: an introduction to spatial data analysis. Geogr Analysis. Accessed April 15, 2018. https://onlinelibrary.wiley.com/doi/epdf/10.1111/j.0016-7363.2005.00671.x
28.
Rao  JNK.  Inferential issues in model-based small area estimation: some new developments.   Stat Transit New Ser. 2015;16(4):491-510. doi:10.21307/stattrans-2015-029 Google ScholarCrossref
29.
Anselin  L, Syabri  I, Kho  Y.  GeoDa: an introduction to spatial data analysis.   Geogr Anal. 2006;38(1):5-22. doi:10.1111/j.0016-7363.2005.00671.x Google ScholarCrossref
30.
Adult obesity in the United States: the state of obesity. Accessed April 10, 2018. https://stateofobesity.org/adult-obesity/
31.
US Census Bureau QuickFacts. Arkansas. Accessed July 10, 2019. https://www.census.gov/quickfacts/fact/table/bentoncityarkansas,AR/IPE120217
32.
Michimi  A, Wimberly  MC.  Spatial patterns of obesity and associated risk factors in the conterminous US.   Am J Prev Med. 2010;39(2):e1-e12. doi:10.1016/j.amepre.2010.04.008 PubMedGoogle ScholarCrossref
33.
Lundeen  EA, Park  S, Pan  L, O’Toole  T, Matthews  K, Blanck  HM.  Obesity prevalence among adults living in metropolitan and nonmetropolitan counties—United States, 2016.   MMWR Morb Mortal Wkly Rep. 2018;67(23):653-658. doi:10.15585/mmwr.mm6723a1 PubMedGoogle ScholarCrossref
34.
Ogden  CL, Lamb  MM, Carroll  MD, Flegal  KM.  Obesity and socioeconomic status in adults: United States, 2005-2008.   NCHS Data Brief. 2010;(50):1-8.PubMedGoogle Scholar
35.
Shields  M, Connor Gorber  S, Tremblay  MS.  Estimates of obesity based on self-report versus direct measures.   Health Rep. 2008;19(2):61-76.PubMedGoogle Scholar
36.
Merrill  RM, Richardson  JS.  Validity of self-reported height, weight, and body mass index: findings from the National Health and Nutrition Examination Survey, 2001-2006.   Prev Chronic Dis. 2009;6(4):A121.PubMedGoogle Scholar
37.
Dutton  DJ, McLaren  L.  The usefulness of “corrected” body mass index vs. self-reported body mass index: comparing the population distributions, sensitivity, specificity, and predictive utility of three correction equations using Canadian population-based data.   BMC Public Health. 2014;14:430. doi:10.1186/1471-2458-14-430 PubMedGoogle ScholarCrossref
38.
Liechty  JM, Bi  X, Qu  A.  Feasibility and validity of a statistical adjustment to reduce self-report bias of height and weight in wave 1 of the Add Health study.   BMC Med Res Methodol. 2016;16(1):124. doi:10.1186/s12874-016-0227-y PubMedGoogle ScholarCrossref
39.
Drieskens  S, Demarest  S, Bel  S, De Ridder  K, Tafforeau  J.  Correction of self-reported BMI based on objective measurements: a Belgian experience.   Arch Public Health. 2018;76:10. doi:10.1186/s13690-018-0255-7 PubMedGoogle ScholarCrossref
40.
Flegal  KM, Carroll  MD, Kit  BK, Ogden  CL.  Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999-2010.   JAMA. 2012;307(5):491-497. doi:10.1001/jama.2012.39 PubMedGoogle ScholarCrossref
41.
Srebotnjak  T, Mokdad  AH, Murray  CJ.  A novel framework for validating and applying standardized small area measurement strategies.   Popul Health Metr. 2010;8:26. doi:10.1186/1478-7954-8-26 PubMedGoogle ScholarCrossref
42.
Kitahara  CM, Flint  AJ, Berrington de Gonzalez  A,  et al.  Association between class III obesity (BMI of 40-59 kg/m2) and mortality: a pooled analysis of 20 prospective studies.   PLoS Med. 2014;11(7):e1001673. doi:10.1371/journal.pmed.1001673 PubMedGoogle Scholar
43.
Why you should care about the Treat & Reduce Obesity Act (TROA). Obesity Action Coalition. Accessed February 23, 2020. https://www.obesityaction.org/community/news/access-to-care/you-should-care-treat-reduce-obesity-act-troa/
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Nutrition, Obesity, and Exercise
    May 8, 2020

    Use of Small-Area Estimates to Describe County-Level Geographic Variation in Prevalence of Extreme Obesity Among US Adults

    Author Affiliations
    • 1Center for Systems and Community Design, The City University of New York Graduate School of Public Health & Health Policy, New York, New York
    • 2CUNY Institute for Demographic Research, The City University of New York, New York, New York
    • 3Marxe School of Public and International Affairs, Baruch College, The City University of New York, New York, New York
    JAMA Netw Open. 2020;3(5):e204289. doi:10.1001/jamanetworkopen.2020.4289
    Key Points español 中文 (chinese)

    Question  What is the county-level prevalence of extreme obesity in the United States?

    Findings  In this cross-sectional study of adults in the United States, county-level prevalence of self-reported extreme obesity ranged from 1.3% to 15.7%. Several prominent clusters of high prevalence were identified, including in the Mississippi Delta and the Southeast.

    Meaning  Prevalence of extreme obesity appears to vary considerably by county; heterogeneity is obscured by available state-level prevalence estimates.

    Abstract

    Importance  The prevalence of extreme obesity continues to increase among adults in the US, yet there is an absence of subnational estimates and geographic description of extreme obesity. This shortcoming prevents a thorough understanding of the geographic distribution of extreme obesity, which in turn limits the ability of public health agencies and policy makers to target areas with a known higher prevalence.

    Objectives  To use small-area estimation to create county-level estimates of extreme obesity in the US and apply spatial methods to identify clusters of high and low prevalence.

    Design, Setting, and Participants  A cross-sectional analysis was conducted using multilevel regression and poststratification with data from the 2012 Behavioral Risk Factor Surveillance System and the US Census Bureau to create prevalence estimates of county-level extreme obesity (body mass index ≥40 [calculated as weight in kilograms divided by height in meters squared]). Data were included on adults (aged ≥18 years) living in the contiguous US. Analysis was performed from June 4 to December 28, 2018.

    Main Outcomes and Measures  Multilevel logistic regression models estimated the probability of extreme obesity based on individual-level and area-level characteristics. Census counts were multiplied by these probabilities and summed by county to create county-level prevalence estimates. Moran index values were calculated to assess spatial autocorrelation and identify spatial clusters of hot and cold spots. Estimates of moderate obesity were obtained for comparison.

    Results  Overall, the weighted prevalence of extreme obesity was 4.0% (95% CI, 3.9%-4.1%) and the prevalence of moderate obesity was 23.7% (95% CI, 23.4%-23.9%). County-level prevalence of extreme obesity ranged from 1.3% (95% CI, 1.3%-1.3%) to 15.7% (95% CI, 15.3%-16.0%). The Pearson correlation coefficient comparing model-predicted estimates with direct estimates was 0.81 (P < .001). The Moran index I score was 0.35 (P < .001), indicating spatial clustering. Significant clusters of high and low prevalence were identified. Hot spots indicating clustering of high prevalence of extreme obesity in several regions, including the Mississippi Delta region and the Southeast, were identified, as well as clusters of low prevalence in the Rocky Mountain region and the Northeast.

    Conclusions and Relevance  Substantial geographic variation was identified in the prevalence of extreme obesity; there was considerable county-level variation even in states generally known as having high or low prevalence of obesity. The results suggest that extreme obesity prevalence demonstrates spatial dependence and clustering and may support the need for substate analysis and benefit of disaggregation of obesity by group. Findings from this study can inform local and national policies seeking to identify populations most at risk from very high body mass index.

    Introduction

    As the obesity epidemic enters its fifth decade in the US, estimates indicate that the prevalence of obesity among adults has reached 37.7%.1 Furthermore, the population with class III, or extreme, obesity (body mass index [BMI] ≥40 [calculated as weight in kilograms divided by height in meters squared]), has increased from 2.9% in 1988-1994 to 7.7% in 2015-2016.2,3 The high rate of extreme adult obesity is likely to not only continue but to increase. The annual incidence of extreme obesity was most recently estimated to be 0.7%.4 While obesity overall has been the focus of extensive research, much less is known about extreme obesity. Given the increasing prevalence of extreme obesity, it is necessary to disaggregate obesity to better understand differences in both the epidemiologic factors and morbidity and mortality by class of obesity.5

    Existing research on extreme obesity describes variation in prevalence by sex, race/ethnicity, and other demographic factors.1,2,6-10 In contrast, patterns of geographic variation appear to be missing, which limits the ability of public health agencies and policy makers to target areas with a known higher prevalence.

    Currently, surveys designed and weighted for national- or state-level estimates are the primary source of data on obesity surveillance.11 State estimates, however, often mask the heterogeneity of health risk differences within communities because of regional differences in population age distribution and socioeconomic factors, such as race/ethnicity and poverty; this heterogeneity has been demonstrated in obesity research.12 Statistical methods for small-area estimation, ie, for providing prevalence estimates at a substate level, include multilevel regression and poststratification (MRP), which combines estimates from a multilevel prediction model and stratified population counts to estimate desired prevalence.13-15

    Small-area estimation methods have been used to create county-level prevalence estimates of obesity,12,16 but, to our knowledge, similar county-level estimates by class of obesity have never been published or publicly estimated. The aim of our study was to create prevalence estimates of county-level extreme obesity using MRP and assess spatial clustering of prevalence estimates. To facilitate comparison between extreme obesity and obesity in general, this analysis also calculated estimates for moderate obesity.

    Methods
    Data Sources

    We conducted a retrospective, cross-sectional observational study using the 2012 Behavioral Risk Factor Surveillance System (BRFSS) and data from the US Census Bureau. The BRFSS is an annual telephone survey of adults aged 18 and older conducted by the Centers for Disease Control and Prevention to monitor health-related behavior of the noninstitutionalized US population and was used to obtain individual-level variables used in the MRP prediction model. The cross-sectional survey is weighted to allow for direct national and state-level estimates.17 The 2012 survey was chosen because a county-level residence indicator is not available from 2013 onward.

    County-level population counts for each of the 3109 counties in the contiguous US were obtained from the 2010 US Census for the purpose of weighting estimates for the MRP. County-level covariates used in the MRP prediction model were obtained from the 2012 American Community Survey 5-year roll-up. The American Community Survey, conducted by the US Census Bureau, is an annual nationwide survey that provides detailed information about select social, economic, and housing characteristics of the US population.18 This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.19 The CUNY School of Public Health Human Research Protection Program indicated that institutional review board approval was not required because secondary analysis of publicly available and deidentified data is not human subjects research.

    Outcome Variable

    The primary outcome was county-specific prevalence estimates of extreme obesity. Body mass index was calculated from self-reported weight and height. Extreme obesity (yes/no) was defined as BMI greater than or equal to 40.0. A secondary outcome of moderate obesity (yes/no) was defined as BMI between 30.0 and 39.9. The public-use BRFSS data set excluded BMI for pregnant women (n = 2873). Individual-level covariates included sex (male/female), age group (18-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, and ≥80 years), and race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, Asian, Hawaii Native/Pacific Islander, American Indian/Alaska Native, other single race, and multiracial). County-level sociodemographic characteristics were the percentage of county residents living below the federal poverty level, the percentage of adults aged 25 years or older with a bachelor’s degree, and the percentage of county residents who lived in rural settings, since these factors are understood to be associated with obesity.20,21 County-level variables were categorized into quartiles, based on the national distribution. Finally, county-level census counts stratified by age group × sex × race/ethnicity were used in the poststratification step to obtain estimates. Cross-tabulated fields for age group × sex × race/ethnicity (208 groups) were extracted at the county level, and categorization for each demographic identifier was identical to categories used in the BRFSS survey data.

    Statistical Analysis

    Before model construction, counties missing any survey data owing to sampling variation and state-specific determination of substate sampling17 (878 of 3109 counties) were aggregated to larger countylike areas. Neighboring counties were joined using an iterative aggregation tool22 until a minimum of 5 survey observations were obtained, resulting in 2215 county or countylike areas (eMethods 1 in the Supplement), referred to henceforth as counties. Survey responses, population counts, and covariate data were summed across aggregated counties. County-level prevalence was then obtained through a 2-step process, as follows.

    First, a multilevel logistic regression model was fit to estimate the probability of extreme obesity using BRFSS data, accounting for both individual-level and county-level covariates.23 Random effects for county and state were included in the model to allow each to represent the association of county and state contextual effects with the outcome, accounting for between-area variation that cannot be explained with the inclusion of ancillary variables.24 Survey weights were rescaled by state to ensure accurate parameter SEs from model results23 and a weight statement was included in the model. Covariates that were not significant at P < .05 with 2-tailed testing were excluded and the model was refit. Educational level was the only county-level variable that was significantly associated with extreme obesity in the multivariable models. The fitted model was applied to estimate the average probability of extreme obesity for each cross classification of characteristics, county, and state (eMethods 2 in the Supplement).

    In the second step, parameter estimates were applied to corresponding census population counts to obtain the prevalence of extreme obesity weighted to the demographic characteristics of the county’s population. Model parameter estimates were summed for each of the 208 age groups × sex × race/ethnicity × county groupings and then multiplied by the population count for the corresponding age group × sex × race/ethnicity within the county (eMethods 3 and eFigure 1 in the Supplement). After weighting the predicted value by the actual subgroup proportion within each county for each of the aforementioned cells, all cells within a county were summed to produce the county-level prevalence estimates.23 After repeating these steps with moderate obesity as the outcome, choropleth maps were created for both extreme and moderate obesity estimates for counties, using 5 classes based on quantile breaks.25

    Overall spatial dependence for the prevalence of each obesity group was assessed using the global Moran index I test, which compares neighboring units across the whole study area to inform positive spatial autocorrelation or dispersion. A positive value of the Moran index I statistic (range, −1 to 1) indicates clustering, a negative value indicates dispersal, and 0 indicates complete spatial randomness (eMethods 4 in the Supplement). A P value was calculated for each indicator by running 999 permutations to create a distribution around the original index value.

    Local indicators of spatial association were used to identify areas with local spatial clustering (eMethods 4 in the Supplement).26,27 This method computes county-specific Moran index I statistics based on a local neighborhood of counties, which we defined through a first-order queen contiguity specification. Alternative specifications were assessed in sensitivity analyses (eFigure 2 in the Supplement). County-specific local Moran index I statistics were mapped to show significant clusters of obesity group prevalence, identifying hot spots (high-high [clustering of high prevalence]) and cold spots (low-low [clustering of low prevalence]), as well as counties classified as low-high, indicating counties with low prevalence adjacent to those with high prevalence, and the inverse scenario of high-low.

    Internal validation examined the degree to which model estimates correlated with known data estimates in 2 ways: first, model-predicted estimates created by MRP were compared with direct unweighted estimates from all counties with 100 or more survey observations14,23; second, county-level estimates were aggregated to yield state estimates using a population-weighted average and compared with direct state-level weighted survey estimates.28 Prevalence estimates by state were calculated accounting for the complex sampling design of the BRFSS. In both scenarios, descriptive statistics, including median, range, and interquartile range, were compared and Pearson correlation coefficients were computed to assess linear correlation between the 2 sets of estimates. Each of these methods was performed separately for moderate and extreme obesity.

    Data analysis was performed from June 4 to December 28, 2018, using SAS, version 9.4 (SAS Institute Inc), GeoDa, version 1.8.16.4,29 and QGIS, version 2.18.

    Results

    Overall, the weighted prevalence of extreme obesity was 4.0% (95% CI, 3.9%-4.1%) and the prevalence of moderate obesity was 23.7% (95% CI, 23.4%-23.9%). Weighted state prevalence of extreme obesity ranged from 2.5% (95% CI, 2.1%-2.9%) in Colorado to 6.1% (95% CI, 5.3%-6.9%) in Louisiana (Table 1). Prevalence of moderate obesity ranged from 18.0% (95% CI, 17.1%-18.9%) in Colorado to 28.9% (95% CI, 27.4%-30.4%) in Mississippi. Weighted prevalence estimates by demographic factor can be found in the eTable in the Supplement.

    County-level prevalence estimates of extreme obesity ranged from 1.3% (95% CI, 1.3%-1.3%) to 15.7% (95% CI, 15.3%-16.0%), with a median of 4.6% (95% CI, 4.5%-4.7%) (Figure 1A; eFigure 3 in the Supplement). The highest prevalence of extreme obesity was found among counties in Ohio, Arkansas, and Alabama. Counties with the lowest prevalence of extreme obesity were found in Colorado, California, and Massachusetts. States that showed the greatest variability of county-level prevalence estimates of extreme obesity included Ohio, where estimates ranged from 3.0% (95% CI, 2.9%-3.0%) to 15.7% (95% CI, 15.3%-16.0%), followed by Arkansas (from 2.5%; 95% CI, 2.4%-2.6% to 15.1%; 95% CI, 14.8%-15.5%) and South Carolina (from 2.0%; 95% CI, 2.0%-2.0% to 12.2%; 95% CI, 11.9%-12.4%). In California, the most populous state, prevalence estimates ranged from 1.7% (95% CI, 1.7%-1.8%) to 8.8% (95% CI, 8.5%-8.9%), and in Texas, the largest state areawise, estimates ranged from 2.5% (95% CI, 2.4%-2.6%) to 7.4% (95% CI, 7.2%-7.5%).

    The county-level prevalence of moderate obesity ranged from 13.3% (95% CI, 13.1%-13.4%) to 41.3% (95% CI, 41.0%-41.6%), with a median of 26.1% (95% CI, 25.9%-26.4%) (Figure 1B; eFigure 4 in the Supplement). The highest prevalence was found among counties in Louisiana, Mississippi, and Arkansas; the lowest prevalence was found in Colorado, New York, and California. States that showed the greatest variability of county-level prevalence estimates of moderate obesity included Idaho (from 14.7%; 95% CI, 14.6%-14.9%, to 37.5%; 96% CI, 37.2%-37.8%) followed by California (from 14.2%; 95% CI, 14.1%-14.4% to 34.6%; 95% CI, 34.4%-34.8%) and Louisiana (from 21.4%; 95% CI, 21.2%-21.5% to 41.3%; 95% CI, 41.0%-41.6%).

    Spatial Dependence and Clustering

    The global Moran index I statistic for extreme obesity was 0.35 (P < .001), indicating that the distribution of prevalence of extreme obesity was spatially autocorrelated. The Moran index I statistic for moderate obesity was similar to that for extreme obesity, at 0.38 (P < .001). Given that one would expect a high level of heterogeneity at the county level, having this much spatial dependence is considered meaningful.

    Statistically significant spatial clusters of extreme obesity are presented in Figure 2A. Of the 2215 counties, there were 208 hot-spot counties and 326 cold-spot counties. The largest cluster was in the Mississippi Delta region, composed of counties surrounding the Mississippi River in the Southeast and in parts of Texas. The second largest cluster was also in the Southeast, predominantly among counties in North Carolina and South Carolina and Virginia. Additional small hot spots were found among counties in Ohio’s rural northwest region and Kentucky.

    A broad swath of the western US and the Great Plains region showed cold spots of a lower prevalence of extreme obesity. The second predominant cold spot was located among counties in New England. There were many additional small cold spots, including within Washington, DC, and surrounding suburbs, the southern tip of Florida, and central Texas. Georgia also contained a small cold-spot cluster of lower prevalence of extreme obesity in the counties comprising the Atlanta metropolitan area.

    Statistically significant spatial clusters of moderate obesity are presented in Figure 2B. There were 262 hot-spot counties and 248 cold-spot counties. Similar to extreme obesity, the predominant cluster of moderate obesity was composed of counties in the southern states surrounding the Mississippi River. Unlike with extreme obesity, this cluster also included counties in the eastern part of Oklahoma. There were also 2 hot-spot clusters in west and south Texas. Similar to extreme obesity, there was a noticeable, but smaller, cluster of moderate obesity found in North Carolina, South Carolina, and Virginia. Additional small, less-dense clusters of moderate obesity were found in West Virginia, Ohio, Kentucky, and northern Michigan.

    Relative to extreme obesity, a smaller area of the country showed cold spots for moderate obesity. In addition to the dominant western cold spot in Montana down through Utah and Colorado, the second cold spot for moderate obesity was located among counties in New England and was denser than for extreme obesity. Similar to extreme obesity, there was also a cold spot in Washington, DC, and surrounding suburbs. The southern tip of Florida also showed a cluster of lower prevalence, and this cluster was larger for moderate obesity than for extreme obesity and also extended into central Florida. The only area in Minnesota that was a cold spot was the Minneapolis-St Paul metropolitan area; this finding is in contrast to the entire state of Minnesota forming the majority of an extreme obesity cold-spot cluster.

    Sensitivity Analysis

    There were 867 counties with 100 or more survey observations included in the first internal validation approach. The Pearson correlation coefficient, r, between direct unweighted prevalence estimates and model-predicted estimates was 0.81 (P < .001) for extreme obesity and 0.86 (P < .001) for moderate obesity (Table 2). The Pearson r value between weighted direct state-level estimates and model-predicted estimates based on aggregated county-level estimates (the second validation approach) was 0.99 (P < .001) for both extreme and moderate obesity.

    Discussion

    To our knowledge, this is the first study that estimated county-level prevalence of extreme obesity across the US. Results indicate that the prevalence of extreme obesity varied substantially among counties and states. The highest prevalence was most often found in counties in the southern US, while the lowest prevalence was most often in the northeastern and western regions. Mapping of prevalence showed similar overall patterns of extreme obesity and moderate obesity, with differences identified when the focus was more local. Estimates were found to be well validated in internal validity tests. Our findings appear to show that county-level prevalence of extreme obesity is spatially dependent in the contiguous US. We identified and located significant hot spots of extreme obesity, including 2 large hot spots in the southeastern states and a small cluster in Ohio. The prevalence of moderate obesity was also found to be spatially dependent, with hot spots similarly identified in the southeastern states as well as in Texas.

    Our findings provide information about variation in the prevalence of extreme obesity. While state estimates have a 2- to 3-fold range (from 2.5% to 6.1%), counties showed much greater variability, particularly at the higher end, with a 12-fold range of estimates (from 1.3% to 15.7%). In addition, there appears to be substantial variation of county-level prevalence within states, even in states consistently showing higher rates of obesity, such as Oklahoma and Kentucky, and those showing lower rates of obesity, such as California.30 These findings reinforce the importance of examining prevalence at a substate level to identify the areas with the greatest burden and need. County-level estimates allow for state and local health departments to more clearly identify jurisdictions where public health interventions and efforts targeting extreme obesity may be most crucial.

    In addition, our estimates appeared to successfully identify local differences. For example, the northwest tip of Arkansas shows a noticeably lower prevalence of extreme obesity than surrounding areas. This area, home to the Walmart world headquarters, has a different population structure than surrounding areas. Recent data indicate that, while 72% of adults in Arkansas are non-Hispanic white and 22% have a bachelor’s degree, 82% of those in the county where Walmart is located (Benton) are non-Hispanic white and more than 29% have a bachelor’s degree.31 The MRP estimates identified this difference through demographic data, leading to lower prevalence estimates.

    While there was substantial overlap of areas with low and high prevalence of both moderate and extreme obesity, there were also some notable differences. Parts of Florida and Maine indicated some of the highest prevalence rates of extreme but not moderate obesity. Texas and Central California had counties with a high prevalence of moderate but not extreme obesity.

    Hot spots identified for each obesity group were partially consistent with previous research of patterns among overall obesity, although earlier studies identified either smaller hot spots or additional ones.32 For example, Slack et al20 found a hot spot in an area that spanned the border between North Dakota and South Dakota that was not identified among either obesity group in this study. Still, findings from the present study seem to indicate that the hot spot in these states may be associated specifically with higher rates of extreme obesity; this area was identified as a high-low area in the local indicators of spatial association results for extreme obesity and was surrounded by a large cold spot. That is, there was an outlier county with high prevalence of extreme obesity surrounded primarily by counties with low prevalence.

    During model construction, rurality was not found to be significantly associated with extreme obesity in the adjusted model, which was unexpected given the general association found in the literature.21,33 One possibility for this finding is that the variation often identified by rurality was more strongly accounted for by other sociodemographic indicators. Still, cold spots of low prevalence were found in metropolitan regions of states traditionally identified to have high prevalence of obesity: Atlanta, Georgia, and metropolitan areas in Texas were small cold spots for extreme obesity and, to a lesser extent, moderate obesity. These findings are consistent with obesity research indicating lower rates in urban relative to rural areas.21

    Area-level poverty was also not found to be significantly associated with extreme obesity in the adjusted model, although differences by sex and race/ethnicity have been identified in the association between socioeconomic status and obesity, with a positive association found among women and a varying pattern of association among men of differing race and ethnicities.34 Future analyses using small-area estimates should explore extreme obesity prevalence stratified by sex.

    Limitations

    There were a few limitations of this study. The BRFSS relies on self-reported weight and height to calculate BMI, which has been demonstrated to underestimate BMI with a greater bias among those with obesity and showing variability by demographic factors, including race and ethnicity.35,36 Estimates are conservative with true rates of extreme and moderate obesity likely higher than those based on self-reports. Prevalence using surveyor-measured weight and height from 2011 to 2012 indicated a national prevalence of 6.4%.3 Several methods have been proposed for correcting self-reported BMI, yet there is no consensus on the best approach.37-39 Because individuals with high BMI are more likely to underreport weight, the issue of underestimation may be more severe in this analysis than in others focusing on general obesity, further highlighting the need for action. In addition, it would be ideal to assess trend over time and examine more recent estimates of extreme obesity. The most recently available national estimates using surveyor-measured weight and height found that extreme obesity increased 20% as of 2015-2016 (prevalence, 7.7%).3 At this time it is not possible to create estimates for BRFSS surveys after 2012 as more recent waves lack a county indicator in the publicly available data sets. However, differences in prevalence of extreme obesity by age, sex, and race/ethnicity—the main covariates used in the prediction model—have remained consistent over time,1-3,40 which indicates a continued relevance to variation in estimates obtained.

    A concern with small-area estimates is that it is difficult to externally validate findings, although correlation methods were used in this research to assess internal validity.41 Still, surveillance data that allowed for direct estimation would be preferable. Federal and local government agencies should devote resources to better track obesity and other obesity-related chronic health conditions.

    Despite these limitations, this study provides informative data about extreme obesity, which has been described as an integral component of the weight distribution in the US7 and has a deleterious effect on life expectancy that is comparable to the effect of cigarette smoking.42 Future analyses with access to more recent data could examine temporal changes in prevalence rates, particularly to monitor counties with above-average increases in rates and determine whether demographic changes in counties result in changes in prevalence rates.

    Conclusions

    The findings of this study suggest that MRP can be used with available individual-level and county-level indicators to generate county-level estimates of extreme and moderate obesity and that results show apparently substantial and informative variation in county-level prevalence rates. While this technique requires several steps, including data manipulation and an understanding of multilevel modeling, the estimates produced can be useful for understanding a more localized variation in prevalence rates. Extreme obesity poses a sizable health burden in many areas of the US and integrated, comprehensive, and wide-reaching solutions are needed for population-based approaches to prevention and treatment. These efforts should include national, state, and local (county and municipal) approaches aiming to implement a broad spectrum of policies, ranging from health promotion (eg, healthy eating and physical activity) to systemic approaches, such as the currently debated legislation regarding the Treat and Reduce Obesity Act, which seeks to expand insurance coverage for obesity treatment.43

    Back to top
    Article Information

    Accepted for Publication: March 1, 2020.

    Published: May 8, 2020. doi:10.1001/jamanetworkopen.2020.4289

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Mills CW et al. JAMA Network Open.

    Corresponding Author: Carrie W. Mills, PhD, MPH, Center for Systems and Community Design, The City University of New York Graduate School of Public Health & Health Policy, 55 W 125th St, New York, NY 10027 (mills.carrie@gmail.com).

    Author Contributions: Dr Mills had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: All authors.

    Acquisition, analysis, or interpretation of data: All authors.

    Drafting of the manuscript: Mills, Johnson, Balk, Wyka.

    Critical revision of the manuscript for important intellectual content: Mills, Johnson, Huang, Wyka.

    Statistical analysis: Mills, Johnson, Balk.

    Administrative, technical, or material support: Johnson, Huang.

    Supervision: Johnson, Balk, Wyka.

    Conflict of Interest Disclosures: None reported.

    References
    1.
    Flegal  KM, Kruszon-Moran  D, Carroll  MD, Fryar  CD, Ogden  CL.  Trends in obesity among adults in the United States, 2005 to 2014.   JAMA. 2016;315(21):2284-2291. doi:10.1001/jama.2016.6458 PubMedGoogle ScholarCrossref
    2.
    Fryar  CD, Carroll  MD, Ogden  CL. Prevalence of overweight, obesity, and extreme obesity among adults aged 20 and over: United States, 1960-1962 through 2013-2014. National Center for Health Statistics; 2016.
    3.
    Hales  CM, Fryar  CD, Carroll  MD, Freedman  DS, Ogden  CL.  Trends in obesity and severe obesity prevalence in US youth and adults by sex and age, 2007-2008 to 2015-2016.   JAMA. 2018;319(16):1723-1725. doi:10.1001/jama.2018.3060 PubMedGoogle ScholarCrossref
    4.
    Pan  L, Freedman  DS, Gillespie  C, Park  S, Sherry  B.  Incidences of obesity and extreme obesity among US adults: findings from the 2009 Behavioral Risk Factor Surveillance System.   Popul Health Metr. 2011;9(1):56. doi:10.1186/1478-7954-9-56 PubMedGoogle ScholarCrossref
    5.
    World Health Organization. Obesity: preventing and managing the global epidemic. World Health Organization. Published 2000. Accessed June 8, 2017. https://www.who.int/nutrition/publications/obesity/WHO_TRS_894/en/
    6.
    Yang  L, Colditz  GA.  Prevalence of overweight and obesity in the United States, 2007-2012.   JAMA Intern Med. 2015;175(8):1412-1413. doi:10.1001/jamainternmed.2015.2405 PubMedGoogle ScholarCrossref
    7.
    Sturm  R, Hattori  A.  Morbid obesity rates continue to rise rapidly in the United States.   Int J Obes (Lond). 2013;37(6):889-891. doi:10.1038/ijo.2012.159PubMedGoogle ScholarCrossref
    8.
    Freedman  DS, Khan  LK, Serdula  MK, Galuska  DA, Dietz  WH.  Trends and correlates of class 3 obesity in the United States from 1990 through 2000.   JAMA. 2002;288(14):1758-1761. doi:10.1001/jama.288.14.1758 PubMedGoogle ScholarCrossref
    9.
    Keating  C, Backholer  K, Gearon  E,  et al.  Prevalence of class-I, class-II and class-III obesity in Australian adults between 1995 and 2011-12.   Obes Res Clin Pract. 2015;9(6):553-562. doi:10.1016/j.orcp.2015.02.004 PubMedGoogle ScholarCrossref
    10.
    Mendy  VL, Vargas  R, Cannon-Smith  G, Payton  M.  Overweight, obesity, and extreme obesity among Mississippi adults, 2001-2010 and 2011-2015.   Prev Chronic Dis. 2017;14:E49. doi:10.5888/pcd14.160554 PubMedGoogle Scholar
    11.
    Holt  JB, Zhang  X. Geospatial data methods for estimating population health outcomes. Webinar presented at: Council of State and Territorial Epidemiologists; August 2013; Atlanta, GA.
    12.
    Zhang  Z, Zhang  L, Penman  A, May  W.  Using small-area estimation method to calculate county-level prevalence of obesity in Mississippi, 2007-2009.   Prev Chronic Dis. 2011;8(4):A85.PubMedGoogle Scholar
    13.
    Jia  H, Muennig  P, Borawski  E.  Comparison of small-area analysis techniques for estimating county-level outcomes.   Am J Prev Med. 2004;26(5):453-460. doi:10.1016/j.amepre.2004.02.004 PubMedGoogle ScholarCrossref
    14.
    Pierannunzi  C, Xu  F, Wallace  RC,  et al.  A methodological approach to small area estimation for the Behavioral Risk Factor Surveillance System.   Prev Chronic Dis. 2016;13:E91. doi:10.5888/pcd13.150480 PubMedGoogle Scholar
    15.
    Song  L, Mercer  L, Wakefield  J, Laurent  A, Solet  D.  Using small-area estimation to calculate the prevalence of smoking by subcounty geographic areas in King County, Washington, Behavioral Risk Factor Surveillance System, 2009-2013.   Prev Chronic Dis. 2016;13:E59. doi:10.5888/pcd13.150536 PubMedGoogle Scholar
    16.
    Dwyer-Lindgren  L, Freedman  G, Engell  RE,  et al.  Prevalence of physical activity and obesity in US counties, 2001-2011: a road map for action.   Popul Health Metr. 2013;11:7. doi:10.1186/1478-7954-11-7 PubMedGoogle ScholarCrossref
    17.
    Centers for Disease Control and Prevention (CDC). The BRFSS Data User Guide. Published August 15, 2013. Accessed November 15, 2017. https://www.cdc.gov/brfss/data_documentation/pdf/userguidejune2013.pdf
    18.
     American Community Survey Information Guide. United States Census Bureau; 2013.
    19.
    The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. EQUATOR Network. Accessed July 4, 2019. https://www.equator-network.org/reporting-guidelines/strobe/
    20.
    Slack  T, Myers  CA, Martin  CK, Heymsfield  SB.  The geographic concentration of US adult obesity prevalence and associated social, economic, and environmental factors.   Obesity (Silver Spring). 2014;22(3):868-874. doi:10.1002/oby.20502 PubMedGoogle ScholarCrossref
    21.
    Befort  CA, Nazir  N, Perri  MG.  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. doi:10.1111/j.1748-0361.2012.00411.x PubMedGoogle ScholarCrossref
    22.
    Talbot  TO. Geographic Aggregation Tool; R version 1.33, user guide. 2015. Environmental Health Surveillance Section, New York State Department of Health. Accessed April 3, 2020. https://cdn.ymaws.com/www.cste.org/resource/resmgr/scale/GAT_vR13_guide.pdf
    23.
    Zhang  X, Holt  JB, Lu  H,  et al.  Multilevel regression and poststratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the behavioral risk factor surveillance system.   Am J Epidemiol. 2014;179(8):1025-1033. doi:10.1093/aje/kwu018 PubMedGoogle ScholarCrossref
    24.
    Mukhopadhyay  P, McDowell  A. Small area estimation for survey data analysis using SAS Software. SAS Global Forum; 2011. Accessed December 15, 2018. http://support.sas.com/resources/papers/proceedings11/336-2011.pdf
    25.
    Brewer  CA, Pickle  L.  Evaluation of methods for classifying epidemiological data on choropleth maps in series.   Ann Assoc Am Geogr. 2002;92(4):662-681. doi:10.1111/1467-8306.00310 Google ScholarCrossref
    26.
    Anselin  L.  Local indicators of spatial association—LISA.   Geogr Anal. 1995;27(2):93-115. doi:10.1111/j.1538-4632.1995.tb00338.x Google ScholarCrossref
    27.
    Anselin  L, Syabri  I, Kho  Y. GeoDa: an introduction to spatial data analysis. Geogr Analysis. Accessed April 15, 2018. https://onlinelibrary.wiley.com/doi/epdf/10.1111/j.0016-7363.2005.00671.x
    28.
    Rao  JNK.  Inferential issues in model-based small area estimation: some new developments.   Stat Transit New Ser. 2015;16(4):491-510. doi:10.21307/stattrans-2015-029 Google ScholarCrossref
    29.
    Anselin  L, Syabri  I, Kho  Y.  GeoDa: an introduction to spatial data analysis.   Geogr Anal. 2006;38(1):5-22. doi:10.1111/j.0016-7363.2005.00671.x Google ScholarCrossref
    30.
    Adult obesity in the United States: the state of obesity. Accessed April 10, 2018. https://stateofobesity.org/adult-obesity/
    31.
    US Census Bureau QuickFacts. Arkansas. Accessed July 10, 2019. https://www.census.gov/quickfacts/fact/table/bentoncityarkansas,AR/IPE120217
    32.
    Michimi  A, Wimberly  MC.  Spatial patterns of obesity and associated risk factors in the conterminous US.   Am J Prev Med. 2010;39(2):e1-e12. doi:10.1016/j.amepre.2010.04.008 PubMedGoogle ScholarCrossref
    33.
    Lundeen  EA, Park  S, Pan  L, O’Toole  T, Matthews  K, Blanck  HM.  Obesity prevalence among adults living in metropolitan and nonmetropolitan counties—United States, 2016.   MMWR Morb Mortal Wkly Rep. 2018;67(23):653-658. doi:10.15585/mmwr.mm6723a1 PubMedGoogle ScholarCrossref
    34.
    Ogden  CL, Lamb  MM, Carroll  MD, Flegal  KM.  Obesity and socioeconomic status in adults: United States, 2005-2008.   NCHS Data Brief. 2010;(50):1-8.PubMedGoogle Scholar
    35.
    Shields  M, Connor Gorber  S, Tremblay  MS.  Estimates of obesity based on self-report versus direct measures.   Health Rep. 2008;19(2):61-76.PubMedGoogle Scholar
    36.
    Merrill  RM, Richardson  JS.  Validity of self-reported height, weight, and body mass index: findings from the National Health and Nutrition Examination Survey, 2001-2006.   Prev Chronic Dis. 2009;6(4):A121.PubMedGoogle Scholar
    37.
    Dutton  DJ, McLaren  L.  The usefulness of “corrected” body mass index vs. self-reported body mass index: comparing the population distributions, sensitivity, specificity, and predictive utility of three correction equations using Canadian population-based data.   BMC Public Health. 2014;14:430. doi:10.1186/1471-2458-14-430 PubMedGoogle ScholarCrossref
    38.
    Liechty  JM, Bi  X, Qu  A.  Feasibility and validity of a statistical adjustment to reduce self-report bias of height and weight in wave 1 of the Add Health study.   BMC Med Res Methodol. 2016;16(1):124. doi:10.1186/s12874-016-0227-y PubMedGoogle ScholarCrossref
    39.
    Drieskens  S, Demarest  S, Bel  S, De Ridder  K, Tafforeau  J.  Correction of self-reported BMI based on objective measurements: a Belgian experience.   Arch Public Health. 2018;76:10. doi:10.1186/s13690-018-0255-7 PubMedGoogle ScholarCrossref
    40.
    Flegal  KM, Carroll  MD, Kit  BK, Ogden  CL.  Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999-2010.   JAMA. 2012;307(5):491-497. doi:10.1001/jama.2012.39 PubMedGoogle ScholarCrossref
    41.
    Srebotnjak  T, Mokdad  AH, Murray  CJ.  A novel framework for validating and applying standardized small area measurement strategies.   Popul Health Metr. 2010;8:26. doi:10.1186/1478-7954-8-26 PubMedGoogle ScholarCrossref
    42.
    Kitahara  CM, Flint  AJ, Berrington de Gonzalez  A,  et al.  Association between class III obesity (BMI of 40-59 kg/m2) and mortality: a pooled analysis of 20 prospective studies.   PLoS Med. 2014;11(7):e1001673. doi:10.1371/journal.pmed.1001673 PubMedGoogle Scholar
    43.
    Why you should care about the Treat & Reduce Obesity Act (TROA). Obesity Action Coalition. Accessed February 23, 2020. https://www.obesityaction.org/community/news/access-to-care/you-should-care-treat-reduce-obesity-act-troa/
    ×