The impact of menu composition on school lunch participation and entrée selection in elementary children : a multi-level model

Nearly one-third of school-aged children are overweight or obese, putting them at an increased risk for negative health consequences during childhood and as adults. Modification of the food decision environment in the school setting is one potential mechanism to impact healthy eating in youth. This dissertation uses a multi-level model to determine if changes in menu composition will maintain participation in the school lunch program and nudge children towards selecting healthier choices at lunch. Individual lunch purchases for one school year (N=147 days) were collected from 10,134 students (grades 1-5) during August 2009-June 2010. The schools offered three entrées per day, and days were categorized according to the number of low-fat (LF) entrées offered (0, 1, 2) on a given day. Primary outcomes of interest included 1) student participation in the lunch program, and 2) selection of a LF entrée if he/she participated. Data were analyzed using separate two-level logistic hierarchical models that partitioned the variance in each outcome into one day-level predictor (number of LF entrées offered) and four child-level predictors (gender, age, ethnicity, and SES). The final model for participation demonstrated significant main effects for student SES (p<0.001) and ethnicity (p<0.001). The predicted probability of purchase was reduced by 0.17 when a student was classified as high SES, and this effect was constant across days offering 0, 1, or 2 LF entrées. White students were less likely to purchase school lunch compared to other, black, and Hispanic children across all types of days. The predicted probability of purchase was 0.10 lower for white students and this effect was irrespective of day score. In contrast, the final model for LF selection indicated that the difference in the probability of selection was entirely attributed to the number of LF options available to the student. The predicted probability of selection was 0.15 and 0.59 on days with one and two LF entrées, respectively, and these values were similar for all types of students. Strong support exists for the modification of the lunch menu to “nudge” children towards healthy food choice. Implications may be particularly potent for low-income and minority students.