A Garden-Based Intervention to Improve Dietary Diversity in Kenyan School Children: Results from a Natural Experiment

Abstract

Abstract Objectives School gardens may improve child diet, but little is known about their effectiveness in rural areas in low-income countries. We evaluated the ability of school gardens to improve child diet in rural Kenya. We hypothesized that children in intervention schools would improve their dietary diversity and specifically their produce intake. Methods An non-government organization installed gardens in 2 primary schools. We selected 2 geographically proximal additional schools as comparisons. We conducted baseline assessments in 2013, prior to garden installation, and follow-up assessments a year later in 2014 in all 4 schools. We measured child dietary intake via a single 24-hour recall. We calculated dietary diversity using the women's dietary diversity score (WDDS) (continuous) and also examined each of the 10 food group components defined as adequate ≥15 g (binary). We conducted marginal linear or logistic regression models using a generalized estimating equation and included an exposure x time interaction to assess differences in outcomes between intervention and comparison schools from baseline to follow up. We controlled for child age, gender, and orphan status. Results We assessed 855 children (n = 438 intervention; n = 417 comparison) at baseline and 688 children (n = 383 intervention; n = 305 comparison) at follow up. Children in intervention schools were 51.8% male, compared to 56.5% in comparison schools. Mean (standard deviation) age was 11.6 (2.1) years in intervention and 11.8 (2.3) years in comparison schools. All children's WDDS worsened post-intervention. In adjusted difference in difference analyses, WDDS did not differ in intervention vs. comparison schools pre- to post-intervention (β 0.04, CI −0.19, 0.27). However, we observed less of a decrease in meeting adequate intake for pulses (OR 2.18, CI 1.18, 4.01) and other fruits (OR 1.55, CI 1.00, 2.40) in intervention versus comparison schools. Conversely, children in comparison schools had less of a decrease in meat, poultry, and fish compared to children in intervention schools (OR 0.67, CI 0.45, 0.99). Conclusions Children's WDDS worsened in all 4 schools, likely due to a severe drought that affected the region in 2014. We observed some differences in intervention vs. comparison children, but cannot attribute these improvements to school gardens. Funding Sources Duke Global Health Institute.