Abstract

The outbreak of Covid-19 in recent years has once again brought the critical issue of medical resource allocation during a pandemic to the forefront of research and public attention. The dynamic and rapid nature of the pandemic has posed significant challenges in accurately predicting the demands for medical resources and developing effective strategies for their distribution. In this study, we aim to address these challenges by studying the medical resource allocation problem during a pandemic and proposing a data-driven optimization methodology that combines mathematical programming and machine learning techniques.
To tackle the problem of demand prediction, we utilize a Long Short-Term Memory(LSTM) model to predict medical resource demand using historical pandemic time series data. Building upon the demand predictions, we develop a linear programming model to optimize the allocation of medical resources. The objective is to maximize the total accessibility of hospitals within each region while also ensuring a balanced distribution of accessibility across all regions. We also conducted a case study on the application of this framework to the Quebec, Canada, pandemic hospitalization case scenarios. The dataset we utilized consisted of hospitalization case numbers from 16 regions in Quebec, along with the geographical locations of 15 regions and their corresponding healthcare facilities. The prediction performance is evaluated by mean absolute error(MAE) and root mean square error(RMSE), which yielded average values of 3.079 and 5.491, respectively. And after optimizing, the total accessibility of all regions is 4.503. The results indicate the effectiveness of our proposed method in accurately predicting future hospitalization numbers and determining the necessary increase in bed capacity for each region, showcasing its potential to assist in resource planning and allocation during a pandemic.