sity in youth. An analysis of the 1996 Medical Expenditure Panel Survey Household Component found that a combination of lacking health insurance and having public insurance (Medicaid, Medicare, or other public hospital coverage) were directly associated with obesity among adolescents (Haas et al., 2003).
Regional differences in the prevalence of U.S. childhood obesity were already apparent in 1998 based on NLSY data (10.8 percent in western states and 17.1 percent in southern states) (Strauss and Pollack, 2001). However, most data available for regional differences are for adults. In 1998, adult obesity prevalence based on the CDC Behavioral Risk Factor Surveillance System (BRFSS) exceeded 20 percent in several states—Alabama (20.7 percent), Alaska (20.7 percent), Louisiana (21.3 percent), South Carolina (20.2 percent) and West Virginia (22.9 percent)—predominantly in the Southeast (Mokdad et al., 1999). By 2002, BRFSS data revealed that seven states had adult obesity prevalence rates greater than 25 percent: Alabama, Louisiana, Michigan, Mississippi, South Carolina, Texas, and West Virginia (CDC, 2002). Systematic data reflecting regional differences in obesity prevalence for children and youth are currently not available.
Researchers can monitor changes in the nature of the obesity epidemic by comparing the BMI distribution curves derived from population-based surveys and noting shifts in any particular distribution over time. A shift toward higher BMIs over the entire distribution would indicate that virtually everyone is becoming heavier, with lean individuals gradually moving into the overweight range, overweight individuals moving into the obese range, and the number of obese individuals becoming more severely obese. However, a graphical analysis comparing NHANES III (1988-1994) with earlier data found that the distributional patterns of BMIs differed among age groups (Flegal and Troiano, 2000).
For adults, there was a general shift upward in the BMI distribution, with the greatest shift occurring at the upper end of the distribution, reflected by the heaviest subgroups becoming heavier. For younger children aged 6 to 11 years, and to a lesser extent in adolescents, the distributions of BMI values were characterized by little or no difference in the lower part of the distribution, though there was also a greater shift at the upper end, as shown schematically in Figures 2.5a and 2.5b (Flegal and Troiano, 2000). The results of this study indicate that the heaviest children and youth were heavier in NHANES III than in earlier surveys; the authors caution, how-