Aim: Despite the burgeoning research interest in weight status, in parallel with the increase in obesity worldwide, research describing methods to optimise the validity and accuracy of measured anthropometric data is lacking. Even when 'gold standard' methods are employed, no data are 100% accurate, yet the accuracy of anthropometric data is critical to produce robust and interpretable findings. To date, described methods for identifying data that are likely to be inaccurate seem to be ad hoc or lacking in clear justification.
Methods: This paper reviews approaches to evaluating the accuracy of cross-sectional and longitudinal data on height and weight in children, focusing on recommendations from the World Health Organization (WHO). This review, together with expert consultation, informed the development of a method for processing and verifying longitudinal anthropometric measurements of children. This approach was then applied to data from the Australian Longitudinal Study of Indigenous Children.
Results: The review identified the need to assess the likely plausibility of data by (a) examining deviation from the WHO reference population by calculating age- and sex-adjusted height, weight and body mass index z-scores, and (b) examining changes in height and weight in individuals over time. The method developed identified extreme measurements and implausible intraindividual trajectories. It provides evidence-based criteria for the exclusion of data points that are most likely to be affected by measurement error.
Conclusions: This paper presents a probabilistic approach to identifying anthropometric measurements that are likely to be implausible. This systematic, practical method is intended to be reproducible in other settings, including for validating large databases.