COVID-19 Patient Volume Prediction Using Time Series Modeling

Since the widespread of COVID-19 pandemic, there is a higher demand for Personal Protective Equipment (PPE). The shortage in the supply of PPEs has forced hospitals to find ways to manage and utilize available resources to avoid deficiencies in inventory more efficiently. This research uses seasonality forecast and trend projection to support decision-making associated with hospital operations by taking into consideration weekly seasonality. The model was trained using data from April 1, 2020, to April 29, 2020, from a tertiary hospital system in the States of Maryland and Delaware to predict COVID-19 suspected ED patient volumes, with the ultimate goal being to estimate the number of PPE, beds, and staffs needed. The Mean Absolute Percentage Error (MAPE) of the model was 8.41% during the training period. Based on the results of this newly developed tool, the limited resources available were assigned more efficiently and staff were scheduled properly during COVID-19 at the hospital system under study. This research emphasizes the need to apply a data-driven decision-support approach to resource allocation and effective staff scheduling during the pandemic.