R B Jain1. 1. National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA 30341, USA. rij0@cdc.gov
Abstract
BACKGROUND: Surveys such as the Behavioral Risk Factor Surveillance System (BRFSS) collect only self-reported data on height and weight to estimate obesity prevalence rates. Because of biased self-reporting of height and weight, obesity prevalence rates reported by these surveys are too low. OBJECTIVE: To develop regression models that can predict corrected height, weight and obesity prevalence from self-reported data, as well as to compare obesity prevalence rates based on self-reported and modeled data and test for trends in obesity prevalence by gender, age and race/ethnicity. DESIGN: Data from the National Health and Nutrition Examination Survey (NHANES) for the period 1999-2006 were used to develop regression models to predict corrected height, weight and obesity prevalence. Regression coefficients estimated from these models were used to predict corrected height, weight and obesity prevalence for BRFSS data for 1999-2007. RESULTS: Self-reported weights for males were higher by 0.1-0.2 kg and lower by about 1.25 kg than corrected weights for females. Underreporting of weights was lowest for Hispanics when compared with other race/ethnicities. In addition, underreporting of weight increased with an increase in body mass index. Self-reported heights for males were higher than corrected heights by about 2 cm, and for females, by about 1 cm. Overreporting of height increased with an increase in age. Self-reported obesity prevalence was 4.5-5.8% lower than corrected rates for males and by 4.4-5.1% for females. Underreporting of obesity prevalence increased with an increase in age. Obesity prevalence rates increased over time for each gender, race/ethnicity and age group for BRFSS data. CONCLUSION: Obesity prevalence calculated from self-reported data is too low and should be used with caution for health-care planning purposes. When it is not possible to have measured data, corrected heights and weights may be predicted by using models such as those presented by us from a relatively large data set that has both measured and self-reported data.
BACKGROUND: Surveys such as the Behavioral Risk Factor Surveillance System (BRFSS) collect only self-reported data on height and weight to estimate obesity prevalence rates. Because of biased self-reporting of height and weight, obesity prevalence rates reported by these surveys are too low. OBJECTIVE: To develop regression models that can predict corrected height, weight and obesity prevalence from self-reported data, as well as to compare obesity prevalence rates based on self-reported and modeled data and test for trends in obesity prevalence by gender, age and race/ethnicity. DESIGN: Data from the National Health and Nutrition Examination Survey (NHANES) for the period 1999-2006 were used to develop regression models to predict corrected height, weight and obesity prevalence. Regression coefficients estimated from these models were used to predict corrected height, weight and obesity prevalence for BRFSS data for 1999-2007. RESULTS: Self-reported weights for males were higher by 0.1-0.2 kg and lower by about 1.25 kg than corrected weights for females. Underreporting of weights was lowest for Hispanics when compared with other race/ethnicities. In addition, underreporting of weight increased with an increase in body mass index. Self-reported heights for males were higher than corrected heights by about 2 cm, and for females, by about 1 cm. Overreporting of height increased with an increase in age. Self-reported obesity prevalence was 4.5-5.8% lower than corrected rates for males and by 4.4-5.1% for females. Underreporting of obesity prevalence increased with an increase in age. Obesity prevalence rates increased over time for each gender, race/ethnicity and age group for BRFSS data. CONCLUSION:Obesity prevalence calculated from self-reported data is too low and should be used with caution for health-care planning purposes. When it is not possible to have measured data, corrected heights and weights may be predicted by using models such as those presented by us from a relatively large data set that has both measured and self-reported data.
Authors: Mariaelisa Graff; Julius S Ngwa; Tsegaselassie Workalemahu; Georg Homuth; Sabine Schipf; Alexander Teumer; Henry Völzke; Henri Wallaschofski; Goncalo R Abecasis; Lakatta Edward; Cucca Francesco; Serena Sanna; Paul Scheet; David Schlessinger; Carlo Sidore; Xiangjun Xiao; Zhaoming Wang; Stephen J Chanock; Kevin B Jacobs; Richard B Hayes; Frank Hu; Rob M Van Dam; Richard J Crout; Mary L Marazita; John R Shaffer; Larry D Atwood; Caroline S Fox; Nancy L Heard-Costa; Charles White; Audrey C Choh; Stefan A Czerwinski; Ellen W Demerath; Thomas D Dyer; Bradford Towne; Najaf Amin; Ben A Oostra; Cornelia M Van Duijn; M Carola Zillikens; Tõnu Esko; Mari Nelis; Tit Nikopensius; Andres Metspalu; David P Strachan; Keri Monda; Lu Qi; Kari E North; L Adrienne Cupples; Penny Gordon-Larsen; Sonja I Berndt Journal: Hum Mol Genet Date: 2013-05-12 Impact factor: 6.150
Authors: Davene R Wright; Paula Lozano; Elizabeth Dawson-Hahn; Dimitri A Christakis; Wren L Haaland; Anirban Basu Journal: Acad Pediatr Date: 2016-02-12 Impact factor: 3.107
Authors: Carol S Weisman; Marianne M Hillemeier; Danielle Symons Downs; Mark E Feinberg; Cynthia H Chuang; John J Botti; Anne-Marie Dyer Journal: Womens Health Issues Date: 2011-05-04
Authors: Catherine Schairer; Barbara J Fuhrman; Jennifer Boyd-Morin; Jeanine M Genkinger; Mitchell H Gail; Robert N Hoover; Regina G Ziegler Journal: Cancer Epidemiol Biomarkers Prev Date: 2015-12-04 Impact factor: 4.254
Authors: R G Radin; S L Mumford; L A Sjaarda; R M Silver; J Wactawski-Wende; A M Lynch; N J Perkins; L L Lesher; B D Wilcox; S N Hinkle; T C Plowden; K Kim; E F Schisterman Journal: BJOG Date: 2017-10-25 Impact factor: 6.531