As of November 30, 2020, due to our switch to Johns Hopkins University as our main data source for COVID-19 cases and deaths, the contents of this post no longer apply to the data presented on Our World in Data. Unlike our previous source the European CDC, Johns Hopkins University collects and publishes COVID-19 deaths in Sweden by date of report.
The analysis below, however, remains relevant for people who look at the official Swedish government dashboard, where deaths are presented by date of death.
There are two ways that COVID-19 deaths can be presented over time: by the date of death, or the date on which the death is reported. Neither of these methods is necessarily better than the other—but it can affect comparisons across countries and over time if these methods are not consistent.
In the data that we present on Our World in Data, which comes directly from the European CDC, deaths in Sweden are shown by date of death, while deaths in other countries are shown by date of report.
This matters because it takes a number of days until all deaths for a particular day are reported. In practice this means that Sweden might today only report 10 deaths for yesterday, but once reporting is complete the death count for that day might increase to 40.
The death counts for the last 10 days in Sweden should therefore always be interpreted as an incomplete count of the deaths that occurred in this period.
COVID-19 deaths are reported by the Swedish government by the date on which the death occurred.
Since there is a lag between the time a person dies and the time the death is reported, the death counts for the most recent period are always incomplete. They are often most incomplete for the latest 2 to 5 days, but can be incomplete for 10 days or more. This undercount in recent days means that they often appear to be falling; but when this is later completed, data shows that more deaths were occurring during that period.
This means that for the last 10 days of data, death counts in Sweden must only be interpreted as incomplete measures of mortality.
As an example, this chart shows what confirmed deaths looked like for the period from October 20 to October 29, when the data was first published on October 30 (red series), and once many more death certificates had been added on November 12 (blue series).
One day after October 29, it looked as if deaths had peaked on October 27 and then started to fall, but in reality that’s not what happened over this period. What actually happened is shown by the blue series: deaths increased steadily.
This also means that each day, the Swedish government will add new deaths for multiple days in the past—mostly on recent days, but on average up to 10 days in the past, and sometimes even more, if deaths have been reported with a long delay.
Our source for COVID-19 deaths, the European Centre for Disease Prevention and Control, updates its figures for Sweden directly from the Swedish government’s data. This means that these daily changes affecting the historical data will be visible on our charts.
Sweden is the only country for which the European CDC currently applies this method for the reporting of deaths. It is also important to note that this does not apply to confirmed cases, but only to confirmed deaths.
For other countries, the European CDC does not build its dataset based on the date of death, but rather based on the date of report. More precisely, every morning the European CDC collects each country’s cumulative total number of deaths since the start of the pandemic, and subtracts the previous day’s total from it.
This results in a daily figure that corresponds to the number of deaths reported in the last 24 hours—regardless of when those deaths actually happened. This means that if the death toll for a country was 20 for a given day, it will remain 20 indefinitely.1
There is nothing wrong with how Sweden or other countries are reporting deaths. But it is important to know these differences when studying the data from Sweden, and even more when comparing it with other countries.