Longitudinal School-Based BMI Surveillance: Informing Obesity Prevention Strategies

Objectives. Provide best analysis practices evidence for school-based body mass index (BMI) surveillance by evaluating measurement reliability, assessing differences of relative odds of obesity among students measured by cross-sectional compared to longitudinal analysis, and assessing student BMI percentile (PCT) growth trajectories over an eight-year intervention. Methods. Bland-Altman plots, mean absolute differences, and intraclass correlation coefficients (ICC) were used to estimate reliability in controlled and natural settings. Using eight years of data from the school-based BMI surveillance, cross-sectional and longitudinal relative odds ratios of obesity were compared at baseline and two intervention time points. Cross-sectional relative odds were determined by logistic regression and longitudinal relative odds by hierarchical linear modeling (HLM). Differences were assessed by overlap of confidence intervals. HLM was used to estimate CPS students’ BMI PCT growth trajectories as a function of cumulative exposure to intervention, controlling for sociodemographic variables. Results. In the controlled setting, 2.0% height and 1.33% weight measurements and in the natural setting, 0.9% weight measurements were determined outliers. The inter-rater mean absolute height difference in the natural setting, 0.22 inches (SD 0.21), was the only clinically tolerable difference. ICC values were all ≥0.96 for height and weight. Equipment deficiencies, data recording issues, and lack of students’ preparation were challenges to reliable measurements. Over time, cross-sectional analysis overestimated relative odds of obesity compared to longitudinal analysis. Controlling for sociodemographic variables, the mean BMI PCT trajectory started at 57.64 BMI PCT at the first study time point, sloped upward at a rate of 0.81 BMI PCT units for each 12-month unit of time, but decreased this growth rate by -0.19 BMI PCT units with each additional year of exposure to intervention. Conclusion. Evidence for best analysis practices for school-based BMI surveillance consisted of: 1). Assuring measurement reliability through staff training and data cleaning; 2). Analyzing obesity time trends longitudinally—to track obesity trends or measure intervention effects over time, cross-sectional analysis may not be a reliable method since it may overestimate the longitudinal rate. The correlation of cumulative exposure to school-based intervention and decreased BMI PCT growth reinforces the importance of longitudinal obesity prevention methodology and analysis.