Matthias Ritter, Derek V.M. Ott, Friedemann Paul, John-Dylan Haynes, Kerstin Ritter
Abstract: One major bottleneck in the ongoing COVID-19 pandemic is the limited number of critical care beds. Due to the dynamic development of infections and the time lag between when patients are infected and when a proportion of them enters an intensive care unit (ICU), the need for future intensive care can easily be underestimated. To infer future ICU load from reported infections, we suggest a simple statistical model that (1) accounts for time lags and (2) allows for making predictions depending on different future growth of infections. We have evaluated our model for three heavily affected regions in Europe, namely Berlin (Germany), Lombardy (Italy), and Madrid (Spain). Before extensive containment measures made an impact, we first estimate the region-specific model parameters, namely ICU rate, time lag between infection, and ICU admission as well as length of stay in ICU. Whereas for Berlin, an ICU rate of 6%, a time lag of 6 days, and a stay of 12 days in ICU provide the best fit of the data, for Lombardy and Madrid the ICU rate was higher (18% and 15%) and the time lag (0 and 3 days) and the stay in ICU (3 and 8 days) shorter. The region-specific models are then used to predict future ICU load assuming either a continued exponential phase with varying growth rates (0–15%) or linear growth. By keeping the growth rates flexible, this model allows for taking into account the potential effect of diverse containment measures. Thus, the model can help to predict a potential exceedance of ICU capacity depending on future growth. A sensitivity analysis for an extended time period shows that the proposed model is particularly useful for exponential phases of the disease.
The full manuscript is published in Scientific Reports 11:5018 (2021) and is available at https://www.nature.com/articles/s41598-021-83853-2.
The main file is Covid19.m The script can be adapted to data from different cities, regions or countries. The respective data has to be saved in an additional xlsx sheet. Ranges for the parameters ICU rate, average time in ICU and time lag between reported infection and ICU admission have to be defined by the user. The model first estimates those parameters for the available data (by reducing the root mean squared error) and then predicts the future ICU load for a fixed time horizon (e.g., 2 months; can also be determined by the user) assuming different exponential growth rates or linear growth with different slopes. Optionally, capacity limits can be given.
All data we used are public and can be downloaded from the Berlin Senate Department for Health, Nursing and Equal Opportunities (https://www.berlin.de/sen/gpg/service/presse/2020/} for Berlin, the Presidenza del Consiglio dei Ministri -- Dipartimento della Protezione Civile (https://github.com/pcm-dpc/covid-19/blob/master/schede-riepilogative/regioni/dpc-covid19-ita-scheda-regioni-20200421.pdf} for Lombardy, and Datadista (https://github.com/datadista/datasets/tree/master/COVID\%2019) for Madrid. We additionally provide here the xlsx sheets for Berlin, Lombardy and Madrid with data until April 20, 2020.