Application of Probabilistic Sampling Algorithm in COVID-19 Medical Resources Allocation

Abstract

During the COVID-19 pandemic, the virus responsible for this global pandemic is highly contagious. This caused shortage in critical medical resources including vaccines, ventilators, and viral test kits. Thus, the health-care system around the world encountered heavy pressure. In the case of handling the surge in infection population and frequent changes in medical conditions, deterministic approaches cannot perform as well as probabilistic sampling algorithms that incorporate dynamic mechanisms to adapt to environments with high uncertainty. This paper reviewed several studies exploring the application of probabilistic sampling approaches – involving Thompson Sampling, Bayesian inference, and multi-armed bandits – in pandemic medical resource triage. By comparing with the data for static allocation policies, these studies revealed that such methods consistently made improvement in beneficial use of the medical resources among diverse types of resources. Though their developed approaches were constrained by common limitations for probabilistic frameworks, such as high reliance on the accuracy and computational efficiency of environmental estimations, these works suggested one promising direction for formulating allocation in the future pandemics: combining existing medical scoring systems with the probabilistic frameworks while exploring auxiliary tools, such as ones particularly enhancing interpretability or pre-training the data pipeline.