Population-Level Measures to Predict Obesity Burden in Public Schools: Looking Upstream for Midstream Actions.

Abstract

To estimate school-level obesity burden, as reflected in prevalence of obesity, based on the characteristics of students' socioeconomic and geographic environments. Secondary analysis of cross-sectional data. Public schools (N = 504) from 43 of 67 counties in Pennsylvania. Kindergarten through grade 12 students (N = 255 949). School-level obesity prevalence for the year 2014 was calculated from state-mandated student body mass index (BMI) measurements. Eighteen aggregate variables, characterizing schools and counties, were retrieved from federal data sources. Three classification variables-excess weight (BMI ≥ 85th percentile), obesity (BMI ≥ 95th percentile), and severe obesity (BMI > 35% or 120% of 95th percentile)-each with 3 groups of schools (low-, average-, and high-prevalence) were created for discriminant function analysis, based on state mean and standard deviation of school distribution. Analysis tested each classification model to reveal school- and county-level dimensions on which school groups differed from each other. Discriminant functions for obesity, which contained school enrollment, percentage of students receiving free/reduced-price lunch, percentage of black/Hispanic students, school location (suburban/other), percentage of county adults with postsecondary education, and percentage of county adults with obesity, yielded 67.86% correct classification (highest accuracy), compared to 34.23% schools classified by chance alone. In the absence of mandated student BMI screenings, the model developed in this study can be used to identify schools most likely to have high obesity burden and, thereafter, determine dissemination of enhanced resources for the implementation of proven prevention policies and programs.