Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
To examine changes in state-specific obesity and overweight prevalence among US children and adolescents between 2003 and 2007.
Temporal cross-sectional analysis of the 2003 and 2007 National Survey of Children's Health data.
The 50 states and the District of Columbia.
A total of 46 707 and 44 101 children aged 10 to 17 years in 2003 and 2007, respectively.
Main Outcome Measures
Prevalence and odds of obesity and overweight, differentials in which were examined by bivariate and logistic regression analyses.
In 2007, 16.4% of US children were obese and 31.6% were overweight. The prevalence of obesity varied substantially across the states, with Mississippi having the highest prevalence (21.9%) and Oregon the lowest prevalence (9.6%). Overweight prevalence varied from a low of 23.1% for children in Utah to a high of 44.5% for children in Mississippi. Between 2003 and 2007, obesity prevalence increased by 10% for all US children and by 18% for female children, declined by 32% for children in Oregon, and doubled among female children in Arizona and Kansas. Children in Illinois, Tennessee, Kentucky, West Virginia, Georgia, and Kansas had more than twice the adjusted odds of being obese than children in Oregon. Individual, household, and neighborhood social and built environmental characteristics accounted for 45% and 42% of the state variance in childhood obesity and overweight, respectively.
Substantial geographic disparities in childhood obesity and overweight exist, with an apparent shift toward higher prevalence in 2007 for several states. Marked geographic disparities indicate the potential for considerable reduction in US childhood obesity.Published online May 3, 2010 (doi:10.1001/archpediatrics.2010.84).
During the past 3 decades, there has been a dramatic increase in childhood obesity in the United States.1- 4 The rate has more than tripled, and the current prevalence remains high among children across most age, sex, racial/ethnic, and socioeconomic groups.2- 4 In light of these trends, childhood obesity continues to be viewed as a major public health problem in the United States.1- 4
Although sex and racial/ethnic inequalities are presented on a routine basis,2,3 geographic disparities in US childhood obesity are less well examined and trends in geographic disparities are even rarer or nonexistent.5 Lack of geographic analysis is owing to the fact that the sample sizes of the surveys routinely used to estimate childhood obesity prevalence, such as the National Health and Nutrition Examination Surveys (NHANES) and the National Longitudinal Study of Adolescent Health, are far too small to permit state-specific estimates.2,3,5- 7 The National Survey of Children's Health (NSCH) is an exception, however. It has a large sample size, a unique sampling design that facilitates state-based analyses, and extensive sociodemographic and health data that allow estimation of childhood obesity rates at the state level while taking into account geographic differences in social and behavioral risk factors.4,5,8,9 Moreover, the recent availability of the NSCH at 2 time points, 2003 and 2007, allows us to examine changes in state-level disparities in childhood obesity over time.8,9
Although analyses at the national level are important in understanding the extent of sex, racial/ethnic, and socioeconomic disparities, geographic analysis is crucial in that it allows individual states to know how they are performing on a key health indicator (such as childhood obesity) relative to other states and to the nation as a whole.5 Such information is vital to state-based health surveillance efforts and to the development and implementation of state-based policies and programs to tackle the growing epidemic of childhood obesity. Second, documenting disparities between the states with the lowest and highest rates can tell us the extent to which reductions in obesity prevalence can be achieved.5 Third, a temporal analysis should help identify states that not only have high current prevalence of childhood obesity but have also experienced substantial recent increases in prevalence, indicating the need for urgent action for obesity prevention and control. Fourth, a temporal analysis could also help identify states with less than expected increases in obesity prevalence given their demographic and socioeconomic composition, thereby allowing the effect assessment of certain obesity prevention programs.
This study extends our previous analysis of the 2003 NSCH by examining geographic disparities in childhood obesity over time.5 Specifically, this study (1) provides latest estimates of and changes in obesity and overweight prevalence among children and adolescents aged 10 to 17 years in the 50 states and the District of Columbia (DC) using 2 large, nationally representative samples of US children in 2003 and 2007 and (2) examines the extent of state disparities in childhood obesity and overweight after adjusting for individual-level socioeconomic and behavioral characteristics and neighborhood social and environmental characteristics.
The data for this study came from the 2003 and 2007 NSCH.8- 11 The 2003 and 2007 surveys were both conducted by the Centers for Disease Control and Prevention's National Center for Health Statistics, with funding and direction from the Health Resources and Services Administration's Maternal and Child Health Bureau.10,11 The purpose of the 2 surveys was to provide national and state-specific prevalence estimates for a variety of children's health and well-being indicators.10,11
The 2007 NSCH was a telephone survey conducted between April 2007 and July 2008, while the 2003 NSCH was conducted between January 2003 and July 2004.10,11 The 2007 survey had a total sample size of 91 642 children from birth through 17 years including a sample of about 1800 children per state.10 The total sample size of the 2003 NSCH was 102 353.4,11 In the 2 surveys, a random-digit-dial sample of households with children younger than 18 years was selected from each of the 50 states and DC. One child was selected from all children in each identified household to be the subject of the survey. Interviews were conducted in English, Spanish, and 4 Asian languages. The respondent was the parent or guardian who knew most about the child's health status and health care. The interview completion rate, measuring the percentage of completed interviews among known households with children, was 66.0% in 2007 and 68.8% in 2003.10,11 Substantive and methodological details of the 2 surveys are described elsewhere.10,11 The National Center for Health Statistics Research Ethics Review Board approved all data collection procedures.
For the present analysis, the dependent variable was the percentage or odds of obesity and overweight defined as body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) at or above the sex- and age-specific 85th and 95th percentile BMI cutoff points from the 2000 Centers for Disease Control and Prevention growth charts, respectively.1- 5,12 Note that the overweight category (BMI ≥85th percentile) includes obese children (BMI ≥95th percentile). Analysis of obesity and overweight differentials was carried out for 44 101 children aged 10 to 17 years in 2007 and for 46 707 children aged 10 to 17 years in 2003. Information on BMI was not available for children younger than 10 years in the 2007 public-use NSCH data set.10
The child's state of residence was the primary covariate of interest. This included the 50 states and DC. Besides state of residence, we used, based on previous research, the following variables as covariates: age, sex, race/ethnicity, household composition, metropolitan/nonmetropolitan residence, household poverty status measured as a ratio of family income to poverty threshold, neighborhood social conditions as measured by perceived neighborhood safety, presence of garbage or litter in the neighborhood, poor housing, built environmental factors such as neighborhood access to sidewalks, parks/playgrounds, recreation centers, community centers, or boy’s/girls' clubs, television viewing time, recreational computer use, and physical activity.1,4,5,13- 17
Less than 4% of the observations for children aged 10 to 17 years had missing BMI data in 2007; these were excluded from the analysis. The likelihood of missing BMI data was higher for younger children aged 10 to 11 years, Hispanic children, non–English speaking households, and low-income households but it did not differ significantly by sex, household composition, and state of residence. Approximately 9% of the observations had missing income data, which were imputed by a multiple imputation technique.10 For all other covariates, there were few missing cases, which were excluded from the multivariate regression models, yielding an effective sample size of 41 235 for the fully adjusted covariate models in 2007.
Estimates of the percentage of children and adolescents identified as obese and overweight in 2003 and 2007 were computed for all 50 states and DC. The χ2 statistic was used to test the overall association between each covariate including state of residence and obesity or overweight prevalence. The 2-sample t test was used to test the difference in prevalence between any 2 states at 1 point in time or to test for change in prevalence between 2003 and 2007 for a specific state. Multivariate logistic regression models were used to derive the relative odds and adjusted prevalence of childhood obesity and overweight for each state in 2007 after controlling for selected demographic, socioeconomic, and behavioral factors. We selected Oregon as the reference state because it had the lowest obesity prevalence in 2007. The rationale for selecting Oregon as the reference was that it represents, at least in theory, the lowest level to which another state's obesity rate is potentially reducible. Comparing state prevalence estimates to the US average or to a state with prevalence close to the US rate may represent another useful analytic strategy.
To account for the complex sample design of the NSCH, SUDAAN software18 was used to conduct all multivariate logistic analyses and to determine crude and adjusted prevalence.18 Sample weights, used to produce population-based estimates for each state and the United States, were adjusted for nonresponse and were poststratified to match the July 2003 and July 2007 Census Bureau population estimates, respectively.10,11
Table 1 and Table 2 show observed state-specific obesity and overweight prevalences for children aged 10 to 17 years, respectively. Overall, in 2007, 16.4% of US children were obese and 31.6% were overweight. The obesity prevalence in 2007 varied from a low of 9.6% for children in Oregon to a high of 21.9% for those in Mississippi. Overweight prevalence varied from a low of 23.1% for children in Utah to a high of 44.5% for children in Mississippi. Among male children, the observed obesity prevalence in 2007 was lowest in Oregon (11.0%) and highest in Arkansas (27.2%), while female children in Wyoming and Texas had the lowest and highest obesity prevalences, 5.5% and 20.2%, respectively.
The Figure shows a relatively higher prevalence of obesity and overweight in the Southeastern region of the United States, and a larger number of states showed a shift toward higher prevalence in 2007. Although obesity prevalence increased between 2003 and 2007 by 10% for all US children, it increased by 46% for children in Arizona and by 32% for children in Illinois. Between 2003 and 2007, obesity prevalence increased by 46% and 49% for male children in Illinois and Nebraska but declined by 33% and 35% for male children in Oregon and Delaware, respectively. While obesity prevalence increased between 2003 and 2007 by 18% for all female children in the United States, it nearly doubled among female children in Arizona and Kansas. Overweight prevalence did not increase significantly between 2003 and 2007 for the total population; however, it did increase by 9% for female children. Among the states, overweight prevalence increased by 21% and 29% for children in Mississippi and Nevada, respectively.
Obesity and overweight prevalence among US children and adolescents aged 10 to 17 years, 2003 and 2007.
After adjusting for sex- and age-compositional differences, children in many states had significantly higher odds of being obese and overweight in 2007 than children in Oregon (data not shown). In particular, children in Mississippi, Georgia, and Tennessee had 170%, 162%, and 152% higher odds of obesity and 153%, 89%, and 83% higher odds of overweight, respectively, than those in Oregon. Because states vary substantially in their racial/ethnic and socioeconomic composition, adjusted odds ratios and prevalence in Table 3 are more appropriate for state comparisons. The adjustment for socioeconomic, behavioral, and neighborhood factors reduced geographic disparities and decreased the excess odds of obesity among children in several states, particularly those in the South. Specifically, children in Illinois, Tennessee, Kentucky, West Virginia, Georgia, and Kansas (adjusted prevalence, >18.5%) had 105% to 141% higher adjusted odds of being obese than children in Oregon (adjusted prevalence, 10.4%). The other states that showed at least 75% higher adjusted odds of obesity than Oregon include Arkansas, Louisiana, North Carolina, Nebraska, New Jersey, New York, Ohio, and Texas (Table 3). Furthermore, children in Mississippi, Tennessee, West Virginia, Georgia, Illinois, New Jersey, Arkansas, Kentucky, and Kansas had at least 50% higher adjusted odds of being overweight than children in Oregon.
Broad geographical patterns in obesity and overweight prevalence were generally similar for male and female children, with the Southeastern states having relatively higher prevalence. However, the magnitude of state disparities in obesity and overweight prevalence was substantially greater for girls than for boys. Compared with their counterparts in Oregon, male children in Illinois, Arkansas, Nebraska, Tennessee, Ohio, West Virginia, Alabama, and Kentucky, respectively, had 216%, 160%, 157%, 149%, 148%, 139%, 133%, and 131% higher adjusted odds of obesity. Female children in Kansas, Tennessee, North Carolina, Georgia, Kentucky, West Virginia, New Jersey, and Texas, respectively, had 294%, 217%, 208%, 203%, 187%, 175%, and 168% higher odds of obesity than female children in Wyoming (data not shown).
Besides geographical disparities, the substantial effects of several sociodemographic and behavioral factors are worth noting (Table 4). The adjusted odds of obesity or overweight increased significantly in relation to decreased levels of household income, lower neighborhood access to parks or sidewalks, lower levels of physical activity, and increased television viewing time and recreational computer use. After adjusting for socioeconomic and behavioral factors and state of residence, black and Hispanic children, respectively, had 71% and 76% higher odds of obesity and 55% and 78% higher odds of overweight than their white counterparts who were not Hispanic. Compared with children with family incomes exceeding 400% of the poverty threshold, those below the poverty threshold had 121% higher odds of being obese and 117% higher odds of being overweight. Children who watched television more than 2 h/d had 48% higher odds of obesity and 56% higher odds of overweight than those who watched television less than 1 h/d. Children who did not engage in any vigorous physical activity had at least 32% higher odds of obesity or overweight than those who exercised 5 or more d/wk. Children in neighborhoods with no access to sidewalks or walking paths had 24% higher odds of obesity, and children who did not have access to parks or playgrounds had 20% higher odds of overweight than their counterparts with access to such amenities.
To determine the extent to which geographic disparities in childhood obesity could be attributed to state differences in racial/ethnic, socioeconomic, behavioral, and neighborhood characteristics, we compared variances of observed and adjusted state-specific prevalence estimates. The variance of state-specific obesity and overweight prevalence estimates decreased by 45% and 42% after adjustment for the individual- and neighborhood-level covariates.
Using extensive individual, household, and neighborhood data from 2 large, nationally representative cross-sectional surveys—the 2003 and 2007 NSCH—we analyzed changes in state-specific prevalence in childhood and adolescent obesity over time and examined the extent to which selected sociodemographic, behavioral, and neighborhood characteristics explain the observed geographic disparities. No such analyses, to our knowledge, have been carried out previously. Although individual and neighborhood characteristics were strongly associated with obesity and overweight risks and accounted for 42% to 45% of the state variance in childhood obesity and overweight, substantial geographic disparities remained. Children in a number of states, particularly those in such Southern states as Tennessee, Kentucky, West Virginia, and Georgia, remained at more than twice the odds of being obese than children in Oregon. In 2007, all fell considerably short of the Healthy People 2010 goal of child obesity prevalence of 5%.19 Wyoming, with a female childhood obesity prevalence of 5.5%, came closest to meeting the target.
Between 2003 and 2007, obesity prevalence increased by 10% for all US children aged 10 to 17 and by 18% for female children. Female children also experienced a 9% increase in overweight prevalence. Much of the recent increase in the overall prevalence was, therefore, owing to a marked increase in the obesity and overweight prevalence of female children. Moreover, the magnitude of geographic disparities in obesity and overweight prevalence was not only higher for girls than for boys, it increased from 2003 to 2007, primarily for girls. Oregon was the only state for which childhood obesity declined significantly (by 32%). There were a number of states (such as Arkansas, Colorado, Georgia, Florida, Ohio, and Utah) that had large percentage increases in obesity and/or overweight prevalence but that were not statistically significant. However, either male or female children in states such as Alaska, Arizona, Illinois, Kansas, Nebraska, Nevada, and Mississippi experienced a significant increase in obesity or overweight prevalence. Overall, when geographic patterns for 2003 and 2007 are compared, an apparent shift toward higher obesity and overweight prevalence could be seen in 2007 for several states.
Some of the recent rise in the prevalence of childhood obesity at the national level and in several states can be attributed to increases in the proportion of socially disadvantaged populations as the percentage of households with Hispanic children and children from low-income, high-unemployment, and non–English-speaking households grew between 2003 and 2007.8,9 It is not clear to what extent changes in the social, built, or obesogenic environments might have contributed to recent increases and trends in childhood obesity, as no such data were available for the 2003 survey. While state-specific analyses are needed to distinguish the effects of physical activity, dietary factors, and social conditions on changes in prevalence for specific states, the 2003 and 2007 NSCH data did not show any marked changes in levels of physical inactivity or other sedentary activities at the national level.4,8,9 Dietary factors such as mean calorie and fat intake have increased consistently over time among both youth and adults in the United States.2,6 It is conceivable that recent trends in dietary factors may have contributed to the increase in childhood obesity at the national level as well as in specific states.
The geographic patterns in US childhood obesity shown here are similar to those observed for obesity among the US adult population aged 18 years or older.5,20 A number of Southern states such as Mississippi, Georgia, Kentucky, Louisiana, and Tennessee that were in the top quintile of childhood obesity also belonged in the top quintile of the adult obesity rates in 2007.20 For both adult and childhood obesity rates, the Southern region had the highest prevalence and the Western region the lowest.20 Indeed, there was a high correlation between the observed adult and childhood obesity and overweight rates in the 50 states and DC for 2007 (correlation coefficient, 0.50 for obesity and 0.53 for overweight).
Although neighborhood safety and other neighborhood characteristics considered here are important measures of the social environment at the individual or household level, they are perceived or parent-reported measures and may not capture the contextual effect of the more direct or area-based measures of the social environment that might operate at the neighborhood or state level.1,5,15- 17 Geographic differences in the area-based social environmental variables as well as state-level policy measures may explain further geographic disparities in childhood obesity.1,5 The area-based measures of the neighborhood environment may include socioeconomic deprivation, crime rates, the number of (or the proportion of the population having access to) recreation facilities, outdoor parks, playgrounds, and other amenities for physical activity, modes of transportation, vehicular traffic congestion, fast food outlets and restaurants, and media advertising promoting unhealthful food choices.1,5,16,17 Examples of state-based policies that target obesity prevention may include (1) providing increased opportunities for physical activity by improving the existing trail/path system, sidewalks, and creating bike trails, playgrounds, and recreational facilities; (2) increasing access to healthy foods in socioeconomically disadvantaged neighborhoods by encouraging the development of grocery stores and farmers' markets; (3) launching educational or media campaigns that encourage parents to limit adolescents' television viewing and other recreation screen time; and (4) enhancing programmatic resources for surveillance, monitoring, and prevention intervention research on obesity.1,5
A few limitations of our study should be noted. First, the childhood obesity measure in our study was based on parental reports of children's height and weight, which may not accurately reflect the true prevalence. However, limited research has suggested parental or self-reported height and weight data to be a reliable indicator of childhood and adolescent obesity, particularly for studying its correlates.4,21,22A recent comparison of the 2003 NSCH and 2001-2006 NHANES data showed a fairly close correspondence between the obesity and BMI estimates for children aged 10 to 17 years.4 We found that the 2007 NSCH estimates were remarkably close to the 2005-2006 NHANES estimates.6 In the 2005-2006 NHANES, the obesity and overweight prevalences for children aged 10 to 17 years were 18.0% (95% confidence interval, 14.9-21.1) and 33.1% (95% confidence interval, 29.5-36.8), respectively, not significantly different from the corresponding prevalence estimates of 16.4% (95% confidence interval, 15.4-17.3) and 31.6% (95% confidence interval, 30.5-32.8) in the 2007 NSCH. The possible error introduced by parent-reporting of height and weight in the NSCH is likely to be similar in 2003 and 2007.
Second, as with most sample surveys, the potential for nonresponse bias exists for the NSCH, implying that the sample interviewed differed from the targeted child population in a systematic fashion.10 The overall interview completion rate, a measure of the response rate, was 66.0% for the 2007 NSCH, ranging from 60.5% for New Jersey to 76.6% for North Dakota. Because the response rates varied by state, differential nonresponse bias might affect state comparisons in our study. Third, because of a lack of data in the NSCH, we were unable to consider additional individual-level covariates such as children's dietary patterns and parental obesity status, which could potentially affect children's obesity risks and may partly account for the observed geographic differences in obesity reported here.1,5 Fourth, as noted above, our analysis did not include direct area-based measures of the neighborhood environment.
Documenting geographic and social group disparities in childhood obesity should be considered an essential component of a national surveillance system that routinely tracks progress toward meeting the overall health objective of reducing and/or eliminating health inequities across population subgroups.4,19 Marked geographic disparities shown here indicate the potential for considerable reduction in childhood obesity. However, such disparities, if they continue to rise, would have profoundly adverse implications for any future efforts to reduce health inequalities among children and adults. Prevention programs for reducing disparities in childhood obesity should not only include behavioral interventions aimed at reducing children's physical inactivity levels and limiting their television viewing and recreational screen time but should also include social policy measures aimed at improving the broader social and physical environments that create obesogenic conditions that put children at risk for poor diet, physical inactivity, and other sedentary activities.4,5
Correspondence: Gopal K. Singh, PhD, US Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau, 5600 Fishers Ln, Room 18-41, Rockville, MD 20857 (firstname.lastname@example.org).
Accepted for Publication: December 5, 2009.
Published Online: May 3, 2010 (doi:10.1001/archpediatrics.2010.84).
Author Contributions: Dr Singh has full access to all of the data and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Singh. Acquisition of data: Singh, Kogan, and van Dyck. Analysis and interpretation of data: Singh and Kogan. Drafting of the manuscript: Singh. Critical revision of the manuscript for important intellectual content: Singh, Kogan, and van Dyck. Statistical analysis: Singh. Study supervision: Singh, Kogan, and van Dyck.
Financial Disclosure: None reported.
Disclaimer: The views expressed are the authors' and not necessarily those of the Health Resources and Services Administration or the US Department of Health and Human Services.
Additional Information: No institutional review board approval was required for this study, which is based on the secondary analysis of 2 public-use federal databases.
Singh GK, Kogan MD, van Dyck PC. Changes in State-Specific Childhood Obesity and Overweight Prevalence in the United States From 2003 to 2007. Arch Pediatr Adolesc Med. 2010;164(7):598-607. doi:10.1001/archpediatrics.2010.84