• Kumarasan Roystonn, Edimansyah Abdin, Rajeswari Sambasivam, Yunjue Zhang, Sherilyn Chang, Saleha Shafie, Boon Yiang Chua, Janhavi Ajit Vaingankar, Siow Ann Chong, and Mythily Subramaniam.
• Research Division, Institute of Mental Health, Singapore.
• Ann Acad Med Singap. 2021 Apr 1; 50 (4): 306-314.

IntroductionThe study assessed whether self-reported height, weight and derived body mass index (BMI) can provide an accurate measure of anthropometric data in a multiethnic adult population in Singapore.MethodsStandardised anthropometric measurements were compared against the self-reported values from 5,132 adult residents in a cross-sectional, epidemiological survey. Discrepancies in self-reports from measurements were examined by comparing overall mean differences. Intraclass correlations, Cohen's kappa and Bland-Altman plots with limits of agreement, and sub-analysis by sex and ethnicity were also explored.ResultsData were obtained from 5,132 respondents. The mean age of respondents was 43.9 years. Overall, the height was overestimated (0.2cm), while there was an underestimation of weight (0.8kg) and derived BMI (0.4kg/m2). Women had a larger discrepancy in height (0.35cm, 95% confidence interval [CI] 0.22 to 0.49), weight (-0.95kg, 95% CI -1.11 to -0.79) and BMI (-0.49kg/m2, 95% CI -0.57 to -0.41) compared with men. Height reporting bias was highest among Indians (0.28cm, 95% CI 0.12 to 0.44) compared with Chinese and Malays, while weight (-1.32kg, 95% CI -1.53 to -1.11) and derived BMI (-0.57kg/m2, 95% CI -0.67 to -0.47) showed higher degrees of underreporting among Malays compared with Chinese and Indians. Substantially high self-reported versus measured values were obtained for intraclass correlations (0.96-0.99, P<0.001) and kappa (0.74). For BMI categories, good to excellent kappa agreement was observed (0.68-0.81, P<0.0001).ConclusionSelf-reported anthropometric estimates can be used, particularly in large epidemiological studies. However, sufficient care is needed when evaluating data from Indians, Malays and women as there is likely an underestimation of obesity prevalence.

26 keywords...

hide…