Do new procedures for estimating growth curves introduce outstanding changes in height estimates?

Abstract

Introduction: The use of growth curves to monitor physical development in children is essential for pediatric health care. Cuban height charts were created during the seventies decade following technics recommended at that time. In the last years, statistical methods based computational technics have been obtained a better model pattern for growth process. In particular WHO in the Multicenter Growth Study used the Box-Cox Power Exponential (BCPE) method. It is important to determine whether the application of such procedure in drawing curves introduce changes in the percentiles' estimates commonly used. Objective: To evaluate discrepancies between percentiles' estimates from both methods. Material and methods: supine-decubitus length data and height from Cubans under 20 years old from the first National Growth and Development Study (ENCD) were examined. Perpercentiles were estimated viEl método BCPE a the BCPE method implemented in GAMLSS' package supported in R language. Comparison of smoothed percentiles obtained from both procedures were plotted to evaluate discrepancies. Results: The best fitted models were: for female supine-decubitus lengths the NO (DFµ=10;DFϭ=5;age0.001) and NO(DFµ=14;DFϭ=6;age0.006) for male respectively and NO(DFµ=16;DFϭ=10;age0.544) and NO(DFµ=16;DFϭ=12;age0.117) for female and male height. Height percentiles estimates were closely enough using either method. Some differences were detected for length percentiles, possibly due to computational corrections done while calculating standard deviations of ENCD. Conclusions: Some differences were detected for length percentiles estimations but not for the height. The use of computational procedure is recommended because its considerable reduction of the subjective charge as compare with the method used before. Keywords: Supine-decubitus length, height, growth curves, BCPE method, smoothed percentiles, GAMLSS, children.