Machine Learning for Estimating Catastrophic Health Spending in Disaster-Affected, Data-Scarce Settings

Disasters can increase health costs and push vulnerable households into poverty. Mitigation measures require understanding changes in household health spending patterns but related data are unobserved in many disaster-affected countries. We develop a hybrid machine learning approach to estimate household health spending using longitudinal survey data from Indonesia around the 2006 Yogyakarta earthquake. The model learns spending patterns across income, hazard intensity, and other characteristics, achieving >70% accuracy in a noisy and complex domain. Applied to post-2004 Indian Ocean tsunami survey data, it predicts plausible baseline health spending, uncovering that without targeted aid, catastrophic health spending incidence would have increased from 4.5% to 29.4%. Moreover, moderately damaged households experienced more cost increases than heavily damaged ones. By combining artificial intelligence with household survey data, our framework enables new insights into disaster impacts in data-scarce settings, offering a tool to inform disaster response and poverty reduction strategies worldwide.