Changemaker

Data analysis

We will evaluate the intervention effects on the primary and secondary nutrition and health, environmental, and economic outcomes. The primary health outcome is MUAC z-score. Secondary health outcomes are Body Mass Index (BMI), anemia (Sanguina App), diet quality, physical activity, and knowledge regarding WASH measures. Secondary environmental outcomes are knowledge and practices in waste management, and knowledge in urban farming practice. Linear regression models will be used for the primary outcome (continuous), and logistic regression models will be employed for binary outcomes. Generalized estimating equations will be used to account for the clustered design by school. We will compare the intervention group with the control group in each study setting and depending on the intra-class correlation coefficients (ICCs), we will pool the data for overall intervention effects.

Sample size calculation

For the sample size calculation, we used the school-based programme of nutrition-sensitive (urban farming) and nutrition-specific approaches (dietary, health, environmental education) as the main intervention and changes in BMI z-score as the outcome in the sample size calculation. We rely on BMI-score as the proxy for MUAC z-score, as this novel tool has not been used in previous intervention studies. The design effect was calculated to account for clustered design and was computed as 1+(n−1)×ICC, where n is the number of students needed to be enrolled in each school. ICC represents the intra-class correlation coefficient due to clustering. Assuming an ICC of 0.025 (80) and an n of 150, the design effect equals 4.725. Therefore, the statistically effective sample size from each school was calculated as 32 (ie, 150/4.725), and the total effective sample size was calculated as 64 in each intervention arm. Then, assuming a two-sided α-level of 0.05, a power of 80% and a within-group SD of 0.5 z-score (81), we will be able to detect a mean difference in BMI z-score of 0.175, similar to the expected effect size ranging between 0.1 and 0.2 z-score. To further allow for a 20% loss to follow-up (e.g., due to school dropout) before the end-line data collection, we will need to enrol 190 students from each school (calculated as 150/0.8). The calculations were conducted using the Power and Sample Size Calculation Programme.