Building a Stronger System for Tracking Nutrition-Sensitive Spending: A Methodology and Estimate of Global Spending for Nutrition-Sensitive Foreign Aid

Global aid for nutrition-specific and nutrition-sensitive interventions and proportion of stunted children across low- and middle-income countries: does aid matter?

Development assistance for health (DAH) has increased dramatically over the past two decades, and this increase has led to a debate on the benefits and perverse effects of scaling-up vs scaling back DAH, and the type of interventions DAH should support. Nutrition remains a contested category viewed as essential to achieving primary healthcare objectives but as falling outside of the direct ambit of the health system. Thus, despite the increase in DAH, it continues to remain an underfunded area and little is known about the relationship between aid for nutrition-specific and nutrition-sensitive interventions and the proportion of stunted children across low- and middle-income countries. We hypothesize that as nutrition-specific aid targets local needs of countries and is less fungible than nutrition-sensitive aid, it will contribute more to a reduction in the proportion of stunted children, with the steepest gains among countries that have the highest burden of malnutrition. We use fixed-effects regressions to examine the relationship between the proportion of stunted children and aid for nutrition interventions (specific and sensitive) to 116 low- and middle-income countries (2002–16). We construct our panel using the Creditor Reporting System, Institute of Health Metrics and Evaluation, Food and Agriculture Organization, World Health Organization and World Development indicators databases. We find a one-dollar increase in per capita nutrition-specific aid is associated with a reduction in the proportion of stunted children by 0.004 (P < 0.05). When stratified by burden of malnutrition, a one-dollar increase in per capita nutrition-specific aid to countries with the highest burden of malnutrition is associated with sharper reductions in the proportion of stunted children (0.013, P < 0.01). We also find a significant association for per capita nutrition-sensitive aid and proportion of stunted children when per capita aid for nutrition is lagged by 3 and 4 years (0.0002, P < 0.05), suggesting a long-run association between nutrition-sensitive aid and proportion of stunted children. Our findings suggest that in spite of criticisms that development assistance fails to adequately reach its intended beneficiaries, aid for nutrition has been successful at reducing the proportion of stunted children. Our findings imply a need to scale-up nutrition funding and improve targeting of aid.

Improving Estimates of Numbers of Children With Severe Acute Malnutrition Using Cohort and Survey Data

Severe acute malnutrition (SAM) is reported to affect 19 million children worldwide. However, this estimate is based on prevalence data from cross-sectional surveys and can be expected to miss some children affected by an acute condition such as SAM. The burden of acute conditions is more appropriately represented by cumulative incidence data. In the absence of incidence data, a method for burden estimation has been proposed that corrects available prevalence estimates to account for incident cases using an "incidence correction factor." We used data from 3 West African countries (Mali, Niger, and Burkina Faso, 2009-2012) to test the hypothesis that a single incidence correction factor may be used for estimation of SAM burden. We estimated the incidence correction factor and performed meta-analysis to calculate summary estimates for each country and for all 3 countries. Heterogeneity between countries and years was assessed using the I² statistic. We estimated a pooled incidence correction factor of 4.82 (95% confidence interval: 3.15, 7.38), although there was substantial between-country heterogeneity (I² = 69%). Knowing how many children in a particular area will be malnourished is fundamental to planning an effective operational response. Our results show that the incidence correction factor varies widely and suggest that estimating the burden of SAM with a common incidence correction factor is unlikely to be adequate.