1)I've deleted a thread on German testing for #coronavirus. It should have read: I find it unlikely that Germany undercounts death by #COVID2019 in large numbers. Reason: There's our institute for disease control and prevention (@rki_de) that has reasonable criteria for counting,
2) like counting everyone who tested positive for corona and dies as death by corona. These criteria are similar to other countries (source: br.de/nachrichten/am…). This raises the question of who gets tested, in particular post mortem:
3) RKI says post-mortem tests "can" be done but are not mandatory, as they're confident to have ante mortem test results from the people who died. I've collected some thoughts on why I think post-mortem testing may not be so relevant here:
4) However, it is important to understand that RKI can only collect what is reported to them. Testing and decision-making is done by local health offices - which I would assume is the case in most other countries, too. This can result in inconsistencies:
5) Ex.: There is a corona death reported from Hamburg today (ndr.de/nachrichten/ha…) where the person had died in home quarantine. Corona was found post-mortem and reported! But another corona-positive death wasn't reported because pre-existing conditions made cause ambiguous!
6) Not reporting the second death goes against RKI criteria, but these are local idiosyncrasies that RKI can't do much about (which again, I think is similar to other countries). E.g., in hotspots, Italy doesn't test every death post-mortem anymore (ilpost.it/2020/03/23/mor…).
7) If we assume that these idiosyncrasies are symmetrically distributed within countries, they won't matter much for the national death counts. At the moment, I believe we can't test this assumption.
8) A worrying addition regarding France: German news @BILD says that the French government only reports deaths from COVID-19 who died in hospitals. Deaths in retirement and care facilities are just about to be added to the statistics. Expect a surge in French death nunmbers then.
