Perpustakaan Digital ITB

Objective: To conduct a state-wide examination of public schools and the school neighborhood as potential targets for environmental public health tracking to address childhood obesity. Methods: We examined the relationship of social and physical environmental attributes of the school environment (within school and neighborhood) and childhood obesity in California with machine learning (Random Forest) and multilevel methods. We used data compiled from the California Department of Education, the U.S. Geological Survey, ESRI's Business Analyst, the U.S. Census, and other public sources for ecologic level variables for various years and assessed their relative importance to obesity as determined from the statewide Physical Fitness Test 2003 through 2007 for grades 5, 7, and 9 (n = 5,265,265). Results: In addition to individual-level race and gender, the following within and school neighborhood variables ranked as the most important model contributors based on the Random Forest analysis and were included in multilevel regressions clustered on the county. Violent crime, English learners, socioeconomic disadvantage, fewer physical education (PE) and fully credentialed teachers, and diversity index were positively associated with obesity while academic performance index, PE participation, mean educational attainment and per capita income were negatively associated with obesity. The most highly ranked built or physical environment variables were distance to the nearest highway and greenness, which were 10th and 11th most important, respectively. Conclusions: Many states in the U.S. do not have school-based surveillance programs that collect body mass index data. System-level determinants of obesity can be important for tracking and intervention. The results of these analyses suggest that the school social environment factors may be especially important. Disadvantaged and low academic performing schools have a higher risk for obesity. Supporting such schools in a targeted way may be an efficient way to intervene and could impact both health and academic outcomes. Some of the more important variables, such as having credentialed teachers and participating in PE, are modifiable risk factors.