A novel method for estimating distributions of body mass index. Issue 1 (December 2016)

• Record Type: Journal Article
• Authors: Ng, Marie; Liu, Patrick; Thomson, Blake; Murray, Christopher
• Is Part Of: Population health metrics. Volume 14:Issue 1(2016)
• Page Start: 1
• Page End: 7
• Publication Date: 2016-12
• Subjects: BMI distribution -- Overweight -- Obesity -- L-BFGS-B optimization -- Beta distribution; Health status indicators -- Periodicals; Population -- Statistics -- Periodicals; Health status indicators -- Measurement; Health status indicators -- Statistical methods; 614.420727
• DOI: 10.1186/s12963-016-0076-2
• Languages: English
• ISSNs: 1478-7954

Abstract

Abstract Background Understanding trends in the distribution of body mass index (BMI) is a critical aspect of monitoring the global overweight and obesity epidemic. Conventional population health metrics often only focus on estimating and reporting the mean BMI and the prevalence of overweight and obesity, which do not fully characterize the distribution of BMI. In this study, we propose a novel method which allows for the estimation of the entire distribution. Methods The proposed method utilizes the optimization algorithm, L-BFGS-B, to derive the distribution of BMI from three commonly available population health statistics: mean BMI, prevalence of overweight, and prevalence of obesity. We conducted a series of simulations to examine the properties, accuracy, and robustness of the method. We then illustrated the practical application of the method by applying it to the 2011–2012 US National Health and Nutrition Examination Survey (NHANES). Results Our method performed satisfactorily across various simulation scenarios yielding empirical (estimated) distributions which aligned closely with the true distributions. Application of the method to the NHANES data also showed a high level of consistency between the empirical and true distributions. In situations where there were considerable outliers, the method was less satisfactory at capturing the extreme values. Nevertheless, it remained accurate at estimating the central tendency and quintiles. Conclusion The proposed method offers a tool that can efficiently estimate the entire distribution of BMI. The ability to track the distributions of BMI will improve our capacity to capture changes in the severity of overweight and obesity and enable us to better monitor the epidemic.