Seismic resilience enhancement of EMS system under hospital functional uncertainty: a multi-stage location-assignment-treatment stochastic programming approach

This study proposes a resilience-driven optimization framework for earthquake emergency medical service (EMS) systems. The framework integrates a three-stage stochastic programming model (Location–Assignment–Treatment, LAT) to address temporary hospital location, dynamic patient assignment, and treatment scheduling. To quantify interdependent damage among structural systems, architectural non-structural components, utilities, and medical equipment within permanent hospitals, Bayesian networks are employed for probabilistic modeling. Monte Carlo simulation is used to generate stochastic damage scenarios, and K-means++ clustering is applied to reduce dimensionality. Together, these techniques yield near-optimal EMS solutions with sub-1.0% gaps and minute-level runtimes, demonstrating robustness and scalability. Compared with a two-stage model, the proposed three-stage LAT framework further improves seismic resilience by 4.2%, highlighting the benefit of integrating temporary hospital location decisions. By explicitly accounting for structural damage, equipment failures, and dynamic patient assignment and treatment, this work advances stochastic programming methodologies for EMS resilience enhancement and provides actionable insights for post-disaster medical resource planning.