Impact Evaluation of a Comprehensive Nutrition Program for Reducing Stunting in Children Aged 6-23 Months in Rural Malawi.

BACKGROUND The prevalence of stunting in central rural Malawi is ∼50%, which prompted a multipronged nutrition program in 1 district from 2014 to 2016. The program distributed a daily, fortified, small-quantity lipid-based nutritional supplement, providing 110 kcal and 2.6 g of protein to children aged 6-23 mo, and behavior change messages around optimal infant and young child feeding (IYCF) and water, sanitation, and hygiene. OBJECTIVES Our objective was to perform an impact evaluation of the program using a neighboring district as comparison. METHODS Using a quasi-experimental study design, with cross-sectional baseline (January-March, 2014; n = 2404) and endline (January-March, 2017; n = 2453) surveys, we evaluated the program's impact using a neighboring district as comparison. Impact on stunting was estimated using propensity score weighted difference-in-differences regression analyses to account for baseline differences between districts. RESULTS No differences in mean length-for-age z-score or prevalence of stunting were found at endline. However, mean weight, weight-for-length z-score, and mid-upper arm circumference were higher at endline by 150 g, 0.22, and 0.19 cm, respectively, in the program compared with the comparison district (all P < 0.05). Weekly reports of high fever and malaria were also lower by 6.4 and 4.7 percentage points, respectively, in the program compared with the comparison district (both P < 0.05). There was no impact on anemia. Children's dietary diversity score improved by 0.17, and caregivers' infant and young child feeding and hand-washing practices improved by 8-11% in the program compared with the comparison district (all P < 0.05). CONCLUSIONS An impact evaluation of a comprehensive nutrition program in rural Malawi demonstrated benefit for child ponderal growth and health, improved maternal IYCF and hand-washing practices, but a reduction in stunting prevalence was not observed.