Abstract:

Conducting targeted mass screening and allocating limited pharmaceutical resources accurately and timely to in different regions is of great importance for epidemic control， especially in a sudden large-scale epidemic outbreak involving asymptomatic patients. Based on the practice of emergency management in COVID-19， a new problem， namely， joint multi-period mass screening and pharmaceutical resource allocation in the epidemic outbreak， is proposed. It is a sensing and control problem of spatial-temporal dynamic network with uncertain transmission parameters， non-stationary transmission dynamics， and partially observable system state. To solve this problem， a data-driven optimization approach using the data of periodically diagnosed cases for the decision-making process is proposed， which combines stochastic programming and rolling-horizon optimization methods. By learning the new arrival data， the uncertain parameter space is continuously reduced and can be expressed as an empirical probability. Then， the epidemic detection and pharmaceutical resource allocation problems are formulated as stochastic dynamic programming problems， which are solved by the rolling horizon optimization method. Finally， an environment interaction simulator is established to verify the superiority of the proposed data-driven decision-making approach over the existing practical decision-making method， and the effectiveness of resource allocation strategies under different resource-level constraints as well as decision-making preferences are discussed. The main findings of our study that can offer policy insights for public sectors include the following four folds. First， implementing large-scale mass screening at the early stage of the epidemic outbreak plays a great role in improving the efficiency of epidemic control and reducing the total number of infections. Second， given an intervention budget， increasing the mass screening scale has a larger marginal effect on improving the efficiency of epidemic control than increasing the amount of pharmaceutical resources. Third， when the intervention resources are sufficient， a more exclusive decision-making preference can help reduce the total infection， while a more inclusive decision-making preference can improve the economic efficiency of resource utilization. Fourth， when intervention resources are in shortage， making decisions moderately with a slightly tilting to some severely infected areas can achieve a good performance in both humanitarian and economic efficiency. An online decision-making framework and optimization approach are provided for resource allocation of public health management departments in epidemic outbreaks. By introducing the corresponding mechanism model and data， the proposed approach can provide epidemic mass screening and pharmaceutical resource allocation decision support for other large-scale epidemic outbreaks （such as H1N1， and Ebola）.

Key words: emergency management in the epidemic, data-driven decision making, mass screening in the epidemic, pharmaceutical resource allocation