Childhood obesity, physical fitness and educational outcomes: the use of a large, complex administrative data system to inform public health and education policy.
Keywords: administrative data, obesity, physical fitness. health monitoring and evaluation
Childhood obesity is a major focus of public health policy due to its increasing prevalence and potential for adverse health outcomes throughout the life course. While lack of physical activity is a well known risk factor for obesity, physical education in schools has declined for a number of reasons including fiscal constraints and overarching academic priorities. In New York City, the Department of Health and the Department of Education have cooperated in analyzing a large administrative data system that combines obesity and physical fitness measures from the NYC FITNESSGRAM program with educational outcomes and demographic information. The analyses address three policy goals: 1) monitoring childhood obesity; 2) establishing the relationship among obesity, physical fitness, and academic test performance, and 3) supporting the evaluation of tailored interventions. We outline the statistical challenges arising in these analyses and the methods developed to meet them. The NYC FITNESSGRAM dataset contains information on 600,000 New York City school children ages 5 to 17. The data includes BMI measures for K-12th grade, physical fitness measures for 4th through 12th grade, educational test scores for 3rd to 8th grade, and demographics from enrollment data for approximately 1.1 million students. The analytic data set contains information from 2005-2006 to 2008-2009. As such, data from this project allow for a variety of trend analyses to better understand the development of childhood obesity on an individual-level in a large multi-ethnic population as well as the effect of obesity and physical fitness on academic standardized test performance. Statistical challenges include: data quality and measurement issues, the generalization of obesity measures to all public school students and all New York City children, adjusting for the collinearity of obesity and physical fitness measures when modeling, and identification of control schools for evaluations of interventions. Despite data quality issues, early results suggest that the NYC Fitnessgram data can be used for health monitoring, that the trend and dynamics of childhood obesity are more complicated than previously reported, and that physical fitness is a better predictor of academic performance than obesity.