This study analyzes data from the Behavioral Risk Factor Surveillance System (BRFSS), a cross-sectional telephone survey of noninstitutionalized adults, between 1986 and 2010.11
Individuals are classified into weight categories based on the body mass index (BMI) (weight in kilograms divided by the square of height in meters) calculated from self-reported weight and height. In addition to the standard “obese” category, defined as a BMI of greater than or equal to 30 kg/m2, the main groups of interest are more extreme categories: BMI>=40 (type III obesity); and BMI>=50.
Self-report on average understates weight (more so for women) and overstates height (more so for men).12
The underreporting of weight increases with weight and prevalence at the same cutpoints are therefore lower than if BMI were calculated by independent measurement. A cutpoint of 40 for direct measurement would be more comparable with a cutpoint for self-report of 37.3 for men and 37.0 for women based on our calculations using NHANES 2007–2008 (we also show this “adjusted” trend). The effect on relative change trends is relatively minor unless reporting biases change. A diminishing reporting bias over time, possibly due to increased awareness about obesity and increased self-monitoring, would lead to overestimates. We used the NHANES to assess this possibility, but found no evidence for changes in reporting biases between 1999 and 2008. Similarly, Gorber and Tremblay13
report no change over time in self-report bias in the US, although there was an increasing (rather than diminishing) bias in Canada.
The descriptive statistics in use sampling weights and adjust standard errors for the BRFSS design. The main statistical analysis underlying uses individual-level logistic regression with an indicator of weight category as the dependent variable. Time trend is measured as a linear spline with knots at 1991, 1996, 2001, and 2006 (i.e., linear trends within each 5-year period, but those slopes can differ between 1986–1990, 1991–1995, 1996–2000, 2001–2005, and 2006–2010). The results are adjusted for sociodemographic changes to isolate the unique trend in obesity rates. Regressors include: age (in 5-year intervals), educational achievement (less than high school, high school, some college, college degree), racial group (white, black, Hispanic, other), and sex. State indicators are included to control for the changing survey participation by states over time. Tests are based on the regression model and all results are statistically significant at p<0.01 unless indicated otherwise. The adjusted results are based on the sociodemographic characteristics in the year 2010 survey. We also estimated separate models stratifying by sex and by race/ethnicity.
The prevalence of different levels of obesity among US adults: 2000–2010
Prevalence growth by severity of obesity (in percent over 1986 baseline)