Incomparable COVID death data

[Nettsentrisk]

I'm not sure if this only pertains to Norway, but their way of counting and reporting COVID deaths has since 2021 differed from many other countries, for instance Sweden.

The Norway COVID deaths number that is used here is based on the country's death registry, and not the "death within 30 days of positive COVID test" rule that many countries have been using since early on in the pandemic.

Yet other countries also have this type of more accurate number available, such as Sweden.

Your database now reports over 23,000 COVID deaths for Sweden, yet if their death registry data is used instead, just like the Norwegians, that number falls to around 18,000. In September 2021 the difference between these two counting methods in Sweden was only around 4%, but now the "30 days rule" gives almost a 30% larger number.

There needs to be some sort of variation in how these numbers are reported so that this difference is made clear, so that incomparable numbers are not being compared.

[CSSEGISandData]

Hello,

Thank you very much for your question. This is a complicated issue and I'm hoping I can provide some clarity.

Our approach is and has been to use the total COVID-19 cases and deaths that appear to either be the metric of choice for the national authority to use when reporting their own data. In rare instances, we have made changes based on one metric better representing the US CDC's definition of COVID-19 case or death (as in #6313), but I would stress this evaluation has not been completed for all locations.

In the case of Sweden, the national COVID-19 dashboard reports 23,279 deaths. As this is the metric Sweden chooses to publish in their national reporting, it's the number we report in our repository.

In the case of Denmark, we don't have independent sourcing (as noted in the README), and our total comes from the WHO COVID-19 dashboard. The WHO takes a similar approach to our team in adjudicating source metrics to use (see their Sweden data), so we are comfortable relying on their interpretation.

I would stress that all locations evaluate and report COVID-19 cases and deaths with their own protocols and procedures and those should be taken into account when making comparisons between two locations.

[Nettsentrisk]

I appreciate that each country is responsible for which numbers they choose to publish, yet I think this underappreciates the fact of how these numbers and services based on your data are used by the public and the media. When there are multiple sources of this type of data from the same country, it should be possible to publish and make available the different datasets, so that it's easier to use these in the appropriate context, or at least make it possible to provide a notice that there are differences between the countries' different ways of counting.

For instance, Our World In Data uses the deaths data under the label "confirmed deaths", when that's not really the case. They are only confirmed as far as happening within ~30 days of a positive COVID test, not confirmed to have been a death attributable to COVID. The public will get the wrong picture when it's not being made clear what is actually meant by this number, and that some countries mean something else, such as Norway.