The Importance of Interpolation in Computerized Growth Charting

Abstract

Computer growth charting is increasingly available for clinical and research applications. The LMS method is used to define the growth curves on the charts most commonly used in practice today. The data points for any given chart are at discrete points, and computer programs may simply round to the closest LMS data point when calculating growth centiles. We sought to determine whether applying an interpolation algorithm to the LMS data for commonly used growth charts may reduce the inherent errors which occur with rounding to the nearest data point. We developed a simple, easily implemented interpolation algorithm to use with LMS data. Using published growth charts, we compared predicted growth centiles using our interpolation algorithm versus a standard rounding approach. Using a test scenario of a patient at the 50th centile in weight, compared to using our interpolation algorithm, the method of simply rounding to the nearest data point resulted in maximal z-score errors in weight of the following: 2.02 standard deviations for the World Health Organization 0-to-23 month growth chart, 1.07 standard deviations for the Fenton preterm growth chart, 0.71 standard deviations for the Olsen preterm growth chart, and 0.11 standard deviations for the CDC 2-to-18 year growth chart. Failure to include an interpolation algorithm when designing computerizing growth charts can lead to large errors in centile and z-score calculations.