Franz, B., Wasserburger, A., Hametner, C., & Jakubek, S. (2022, June 27). Forecasting of COVID-19 Hospital Occupancy Using Differential Flatness [Conference Presentation]. 8th International conference on Time Series and Forecasting, Gran Canaria, Spain.
The COVID-19 pandemic has vividly demonstrated the great importance of reliable forecasts in order to enable policy makers to take educated decisions. In this work an easy, fast and accurate method of predicting the number of occupied beds in the hospitals' normal and intensive care units is presented. These metrics have become the most important decision-making factor in many countries. The basic idea in the presented approach is to extend an age-structured compartmental model with artificial inputs which can be interpreted as an aggregation of all exogenous drivers, such as people's behavior, governmental interventions as well as the onset of a new virus variant. With this methodology it is possible to model multiple epidemic waves and also detect a new resurgence of infectious activity early on. Simulating the model with the possible future courses of the exogenous inputs yields the predicted number of infected individuals. Furthermore, by estimating the case-specific hospital and ICU admission rates and extrapolating them into the future, forecasts of the normal and intensive care occupancies are readily obtained. The validation on Austrian data shows the high accuracy of the obtained predictions.
en
Research Areas:
Mathematical and Algorithmic Foundations: 50% Modeling and Simulation: 50%