301117
Redrawing the US Obesity Landscape: Bias-corrected estimates of state-specific adult obesity prevalence
Methods: Using non-parametric statistical matching, we synthesized information from the US Census 2010, American Community Survey (ACS) 2010 (5-year), Behavioral Risk Factor Surveillance System (BRFSS) 2011 (n=407,170), and National Health and Nutrition Examination Survey (NHANES) 2005-2010 (n=15,729 adults) to estimate national and state-specific obesity prevalence. We compare this approach, which validates estimates against measured NHANES data, with uncorrected BRFSS data and with previously published regression-based approaches to correct self-report bias.
Results: The CDC self-reported data from BRFSS underestimate the national prevalence of obesity by 18% [27.88% (27.60%-28.17%) vs. NHANES: 33.89% (32.97%-34.81%)]. Regression-based correction methods also underestimate the prevalence of obesity by 6% [31.96% (31.67%-32.26%)] and grade III obesity (Body Mass Index (BMI) ≥40) by 11% [5.14% (5.0%-5.28%) vs. NHANES: 5.80% (5.35%-6.24%)]. Our method provides a closer estimate of both overall obesity prevalence [34.90% (34.89%-34.92%)] and grade III obesity [5.61% (5.61%-5.63%)]. Using this approach, no state had a corrected adult obesity prevalence below 30%, resulting in strikingly different obesity maps.
Conclusion: In contrast to previous estimates, this study accurately reproduces measured nationally-representative BMI distributions while capturing the wide variation in state-level obesity. For the first time we provide accurate state-specific estimates and maps including the 16.5 million obese adults that self-reported data misclassify.
Learning Areas:
Biostatistics, economicsEpidemiology
Other professions or practice related to public health
Learning Objectives:
Describe the importance of non-parametric statistical matching methods for synthesizing new datasets from existing publicly available administrative and public health sources
Explain how applied public health statistics can shape the public health agenda through the lens of bias-corrected state-specific obesity prevalence estimates
Keyword(s): Obesity, Statistics
Qualified on the content I am responsible for because: I have been the lead programmer for several microsimulation models and have developed non-parametric methods of data synthesis and analysis, especially for publicly available datasets. Among my scientific interests has been the development of a US-focused obesity model to evaluate the cost-effectiveness of various strategies aimed at reducing the burden of obesity.
Any relevant financial relationships? No
I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.