Prediction models for planning health care resources. During the first wave of the Covid-19 pandemic 2020

Abstract

Due to the Covid-19 pandemic, has emphasized a need for planning health care resources based on only a few aggregated data points and little knowledge of the data-generating process.

In the first part of the report, we present the process of our work in spring 2020 during the first wave and especially during the first part with the high demands on health care resources. In the second part of the report, we discuss the logistic growth model (LGM), one of our models used to predict the peak height and the peak timing. We present some different approaches to use the LGM, and compare these to a different data set, Belgium data.

For the Swedish regional data, the LGM on raw observations gave a good estimate on the peak height. The adjusted LGM, using cumulative new inpatient beds, fitted the Swedish regional data to a satisfying degree. For the Belgium data, the LGM on raw observations gave a good estimate on peak height and timing. The adjusted LGM, using cumulative new inpatient beds, did not work for the Belgium data as it gave a too early peak time and a too low peak height.

The experience from our work, in combination with now existing literature, the process in a similar future situation would include better knowledge on how to find and combine data to get as reliable forecasts as possible and to use creativity in combination with theoretical competence.