[vhfpmr (OP)]

https://ourworldindata.org/grapher/covid-19-tests-deaths-scatter-with-comparisons
This is an interesting plot, isn't it showing that countries with a lot of deaths do a lot of tests, and not that countries that do a lot of tests have fewer deaths?

[alancalverd]

Beware of mortality statistics!

"Cause of death" is a matter of opinion, even when you have a positive test for one causative factor*, and that opinion may itself be colored by local custom and practice (e.g. heart disease is far less often cited in France than in the rest of Europe, although diet and smoking habits wold appear to predispose towards it). If you don't test for COVID you can only report "respiratory failure" as anything else would be a guess.

Death is a matter of fact.

So the only universally valid estimate of the impact of COVID would be an unseasonal spike in the overall death rate. But I've also pointed out in another thread that short spikes in recorded deaths may be due to national holidays, civil service strikes, etc.

Best advice: always poke statistics with a sterilised barge pole.

*a detective told me that basic training includes "Unless the head is actually separated from the body, do not assume cause of death".

[vhfpmr (OP)]

If there's a consistent trend across multiple countries that provides some padding against random noise from one country to another.

[alancalverd]

That's half the point. An unseasonal blip in overall deaths that occurs in several countries at once, is indicative of a pandemic factor. So the best estimate of true pandemic deaths in any country is not the reported number, but the difference between that blip and last year's figure for the same period.