COVID-19 Nowcasting


The time series of reported COVID-19 ICU admissions and deaths published by the Swedish Public Health Agency (FoHM) are downward biased for the most recent days. The reason is that the time series show the number of events by date of ICU admission and day of death, respectively, but reporting of admission and death events takes a number of days to reach FoHM. These delays in reporting lead to systematic under-reporting for the most recent days. As a result, the currently reported figures may give the impression of a declining trend, while such a trend no longer exists in the complete data some time later. Here, we present so called “nowcasts” for the number of ICU admissions and COVID-19 associated fatalities in Sweden, i.e. predictions for how the actual numbers look right now if there had not been such reporting delay. The aim of these real-time estimates is to provide situational awareness.

We do not longer update this website as the incidence is currently low.

Nowcasts as of 2023-03-23

ICU admissions


The grey bars show the currently reported ICU admission and fatality numbers; the solid line present the nowcast estimates of the complete number of ICU admissions/fatalities. The color shaded intervals show 95% prediction intervals.

Statistical methodology

The nowcasts are based on a Bayesian hierarchical model implemented in R based on Günther et al. (2020), which is investigated for Swedish mortality data in Lindroos (2021). Our model includes information from additional data streams: For the ICU admission nowcast, we use registered case numbers from previous days and number of new ICU admissions as additional data stream for the fatality nowcast respectively. Detailed documentation of the method can be found in our paper published in PLOS Computational Biology.

The presented work is joint work by Fanny Bergström, Felix Günther, Michael Höhle and Tom Britton at the Department of Mathematics, Stockholm University. This work is partly funded by the Nordic Research Agency (NordForsk, grant 105572).