Classification and Regression Tree (CART) analysis to predict successful delivery of a YMCA childhood obesity intervention program

Abstract

Background: Obesity and its co-morbidities account for a large proportion of chronic disease in the U.S. In an effort to reduce rates of obesity among children, YMCA of the USA (Y-USA) has developed a program called Healthy Weight and Your Child (HWYC), a multi-session intervention program for children carrying excess weight aged 7-13 years old delivered through local YMCAs. To date, 19 local YMCAs have successfully delivered pilot versions of the program. As Y-USA moves from the validation stage of HWYC to translating and scaling the program, Y-USA hopes to improve the model by identifying factors in program delivery that best lead to successful delivery of the program. This analysis used classification and regression tree (CART) modeling to identify such explanatory factors that predict successful delivery of HWYC in local YMCAs, as measured by Change in child BMI and % Attendance.Methods: Participants in HWYC pilot programs included children with obesity (BMI > 95th percentile for age and gender) (n = 782). Data collection included baseline and session 20 weight, height, and waist circumference measures, as well as demographic information. Attendance was measured at every program session. Bivariate and CART analysis was performed in R using previously collected and de-identified data.Results: CART analysis found that a participants attendance is most predicted by the YMCA site delivering HWYC, as well as demographic factors such as race and eligibility for free/reduced lunch. The baseline BMI of the accompanying adult also influenced attendance. CART also found that class size was a major predictor of change in childs BMI, as were program model (yearlong vs. condensed) and sex.Conclusions: The CART analysis successfully identified several factors 40 that predict both attendance and childs BMI change among participants in the HWYC program. Results suggest that Y-USA should scale HWYC to more YMCAs using the condensed 4-month program model to achieve greater successful outcomes. Additionally, individual YMCAs may need to consider the demographics of their unique communities in order to improve both attendance and BMI reduction. As the sample size of the condensed program was small (n = 120), future studies are required to better understand unique explanatory factors that predict successful delivery of HWYC under this program model.

Files

Additional details