ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Assessing Methods for Correcting Response Bias in Self-Reported BMI Data in Australian National Health Surveys

Tim Ayre, Jason Wong

Building: Law Building
Room: Breakout 7 - Law Building, Room 028
Date: 2012-07-10 03:30 PM – 05:00 PM
Last modified: 2012-06-19


Body Mass Index (BMI) is commonly used to measure the prevalence of obesity in populations for health research. Population surveys often ask respondents to report their height and weight, rather than taking physical measurements. Previous research has shown that the discrepancies between self-reported and measured values of height and weight can lead to inaccurate estimates of the population BMI distribution. The accuracy of estimates derived from self-reported BMI data can potentially be improved by adjusting the self-reported values to account for these response biases.

In this paper we investigate the reporting errors in height, weight and BMI of Australian adults using the National Nutrition Survey (NNS) 1995 and National Health Survey (NHS) 2007–08. Both surveys collected measured and self-reported height and weight. We examine several alternative methods for adjusting self-reported data to get more accurate estimates of BMI. Both linear and semi-parametric regressions are used to adjust self-reported BMI, both by modelling BMI directly and by modelling height and weight separately. The resulting BMI distributions are compared with the distributions of measured and self-reported BMI to assess the improvement in accuracy obtained by the different adjustment methods.

All the examined methods for adjusting self-reported BMI are found to provide significantly more accurate estimates of the distribution of BMI than using self-reported BMI directly. The semi-parametric models are found to provide slightly better estimates than linear models, but the improvement in accuracy is not likely to be sufficient to justify their additional complexity in practical settings.