Four Models For Predicting Pediatric Body Composition For Anglo And Hispanic Youth

Abstract

Body composition is becoming recognized as a critical component in the reporting of health-related measures in children due to the increase in metabolic disorders. However, measuring body composition is often difficult due to lack of training, privacy issues and cost of accurate measuring devices. Therefore, it may be advisable to develop a prediction equation that can be used to accurately determine body composition in youth. PURPOSE: To develop a body composition prediction equation for elementary age youth. METHODS: A stratified sampling method was used to select volunteer children for this study. Height, weight, BMI, age, and skinfold were collected on 886 Anglo (n=445) and Hispanic (n=441) children between the ages of 7 and 12 (9.1+2.1). Body composition was determined by using a standard two site procedure that is common in most schools. Group body fat means were examined through two-way ANOVA. Correlation matrices were run with each group on the variables of height, weight, BMI, age, and body fat percent. Regression prediction models were developed using stepwise regression techniques. Cook‘s D and Mahalanobis D distances were examined to determine outlier affects. R2 and standard error of the estimate (SEE) were examined for each resulting model. RESULTS: The data was split by gender and ethnicity after two-way ANOVA revealed a significant difference between males and females and between Anglo and Hispanic groups. Prediction models were established with the following variables for the following groups: Anglo Females: BMI, Weight, Age (R2 =.555, SEE = 4.34); Hispanic Females: Weight, Height (R2 =.478, SEE = 4.52); Anglo Males: BMI, Weight, Age (R2 =.561, SEE = 5.83); Hispanic Males: Weight, Age (R2 =.403, SEE = 5.32). All models indicated acceptable means and ranges for Cook‘s and Mahalanobis distances. CONCLUSIONS: The results indicate that reasonable field-based prediction models can be used as a general assessment tool when clinical methods are not available or practical within Anglo and Hispanic youth populations. The prediction variables also indicated consistency within ethnicities. Weight, age and BMI are consistent body fat prediction variables for Anglo youth while weight is a consistent body fat prediction variable for Hispanic youth. Future studies should examine these variables more closely.