BACKGROUND: This study examines the feasibility of developing correction factors to adjust self-reported measures of body mass index (BMI) to more closely approximate measured values. DATA AND METHODS: Data are from the 2005 Canadian Community Health Survey (subsample 2), in which respondents were asked to report their height and weight, and were subsequently measured. Regression analyses were used to determine which socio-demographic and health characteristics were associated with the discrepancies between self-reported and measured values. The sample was then split into two groups. In the first, self-reported BMI and the predictors of the discrepancies were regressed on measured BMI. Correction equations were generated using all predictor variables that were significant at the p < 0.05 level. These correction equations were then tested in the second group to derive estimates of sensitivity, specificity and obesity prevalence. Logistic regression was used to examine relationships between self-reported, measured and corrected BMI and obesity-related health conditions. RESULTS: Corrected estimates provide more accurate measures of obesity prevalence, mean BMI and sensitivity levels (percentage correctly classified). In almost all cases, associations between BMI and health conditions are more accurate when based on corrected versus self-reported values.
BACKGROUND: This study examines the feasibility of developing correction factors to adjust self-reported measures of body mass index (BMI) to more closely approximate measured values. DATA AND METHODS: Data are from the 2005 Canadian Community Health Survey (subsample 2), in which respondents were asked to report their height and weight, and were subsequently measured. Regression analyses were used to determine which socio-demographic and health characteristics were associated with the discrepancies between self-reported and measured values. The sample was then split into two groups. In the first, self-reported BMI and the predictors of the discrepancies were regressed on measured BMI. Correction equations were generated using all predictor variables that were significant at the p < 0.05 level. These correction equations were then tested in the second group to derive estimates of sensitivity, specificity and obesity prevalence. Logistic regression was used to examine relationships between self-reported, measured and corrected BMI and obesity-related health conditions. RESULTS: Corrected estimates provide more accurate measures of obesity prevalence, mean BMI and sensitivity levels (percentage correctly classified). In almost all cases, associations between BMI and health conditions are more accurate when based on corrected versus self-reported values.
Authors: Fatima Al Sayah; Nick Bansback; Stirling Bryan; Arto Ohinmaa; Lise Poissant; Eleanor Pullenayegum; Feng Xie; Jeffrey A Johnson Journal: Qual Life Res Date: 2015-12-10 Impact factor: 4.147
Authors: Deepa P Rao; Erin Kropac; Minh T Do; Karen C Roberts; Gayatri C Jayaraman Journal: Health Promot Chronic Dis Prev Can Date: 2017-03 Impact factor: 3.240
Authors: G A Handrigan; N Maltais; M Gagné; P Lamontagne; D Hamel; N Teasdale; O Hue; P Corbeil; J P Brown; S Jean Journal: Osteoporos Int Date: 2016-08-25 Impact factor: 4.507