CISSM Global Forum | Predicting Acute Malnutrition for Early Warning and Anticipatory Action on Food Security Crises

Food insecurity resulting in acute malnutrition poses a major, persistent challenge to the international community. An estimated 45 million children (6.7%) under the age of 5 were affected around the world as of 2020. Humanitarian stakeholders have lacked timely, evidence-based, objective projections of acute malnutrition that are vital to facilitate anticipation and action ahead of crises. Doing so offers a prospect to mitigate negative impacts of acute malnutrition on morbidity and mortality. In the process, the targeting of humanitarian interventions could be improved, increasing the efficacy and cost-efficiency of assistance.

The line of research being presented aims to bolster these capabilities relative to traditional approaches. On-the-ground surveillance of the prevalence of malnutrition is difficult and costly to implement regularly at scale, especially in dangerous contexts. This monitoring is typically suited to capture the progression of crises only once they are underway. Existing early warning systems do not tend to spotlight acute malnutrition. Many of these systems involve judgments of experts, who follow a structured process of consulting data measuring the nutritional status of populations and relevant risk factors, though not in a manner that is fully transparent, replicable, and validated. An emergent alternative is statistical modeling that leverages available historical data to ascertain patterns of relationships, which are combined with the latest information on factors to enable reliable forecasting.

The talk will feature a project undertaken in partnership with Action Against Hunger, a leading international NGO. The project was organized with multiple workstreams of modeling, to investigate potential options and gauge the robustness of findings. Each workstream concentrated on a different level of analysis and employed a distinctive statistical methodology. A central consideration is to evaluate metrics of the fit and accuracy of models, which exhibited strong performance that remained consistent across lead-time horizons. The project prioritized attention to settings vulnerable to climate and conflict shocks in Sub-Saharan Africa, substantiating relevant factors as leading indicators.

In addition, the talk will reflect on the agenda of translating modeling results to practical applications. This agenda was pursued by designing analytical frameworks responsive to the needs of addressing the problem. A capstone activity entailed conducting pilot tests of use cases, which involved engagement with key actors who were supplied model-based forecasts for multiple scenarios that could be explored through dashboard tools with interactive features. The talk will conclude by looking ahead to future ambitions of exploring the integration of model-based forecasts into humanitarian decision-making and operational processes from the international to local levels.