Answers are grounded in this paper and cite their sources.
The term “flow-sizing” in healthcare systems refers to the strategic operations management of critical care resources in response to emergencies, adjusting resource availability to meet fluctuating demand. Conventional optimization and static queueing formulations struggle in this setting because objectives are non-compensatory and evolve over time, whereas the effects of policy levers are nonlinear, interrelated and lagged. We propose a novel framework that combines a multiplicative utility function with Machine Learning techniques to evaluate and support critical care capacity planning. The proposed framework relies upon: i) a time-dependent utility function based on hard mortality data and key explanatory features using a Long Short-Term Memory model, ii) domain-informed capacity attributes through logistic-based univariate utility functions, and iii) a calibration of their relative weights using a module that combines differential evolution with deep reinforcement learning. Using the Greek healthcare system during COVID-19 as a case study, the framework reproduces mortality-linked dynamics and identifies high-agility and targeted flow-sizing policies that consistently deliver higher aggregate utility than status-quo triggers, initiate earlier capacity pivots as risk escalates and remain robust across multiple what-if scenarios. The framework captures the structural complexity and temporal dynamics of critical care under strain, providing interpretable, adaptable, evidence-based decision support grounded in real data.
No takes yet. Share an insight, caveat, or question.
Rachaniotis et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: