Another unpopular reminder: Evolution of the lethality of #Covid_19. There should be no evolution, these numbers should be ~constant in space & time for countries with comparable health systems. They are not because WE HAVE NO IDEA how many contagions and hence about lethality. 
2. What this 'mortality' is measuring is how many tests have been performed in each country. And even then, lethality is very biased by whether those tests were done randomly or only on symptomatic patients.
Not to talk about what each country defines as "death by covid-19".
Not to talk about what each country defines as "death by covid-19".
3. It seems, for an outsider amateur, that the @WHO should have a much tighter control on statistics next time.
4. Since 3 weeks ago we do know that the best estimation of mortality rate is Korea's, and it was around 0.6%, in the lowest range of the previous plot. scmp.com/week-asia/heal…
5. But as more randomised tests are performed, that number decreases, the last estimate is this:
"Our current best assumption, as of the 22nd March, is the IFR is approximate 0.20% (95% CI, 0.17 to 0.25)."
cebm.net/global-covid-1…
of course, higher in elders with preconditions
"Our current best assumption, as of the 22nd March, is the IFR is approximate 0.20% (95% CI, 0.17 to 0.25)."
cebm.net/global-covid-1…
of course, higher in elders with preconditions
6. This is slightly higher but within the range of normal influenza (flu -> IFR=0.16%)
Even "in those without pre-existing health conditions, and over 70, the data is reassuring that the IFR will likely not be above 1%".
Even "in those without pre-existing health conditions, and over 70, the data is reassuring that the IFR will likely not be above 1%".
7. Then there is the problem about the lack of immunity against covid. This is not included in IFR and it increases the number of contagions. If IFR is truly ~0.2% and 2000 people have died in Spain, then this means that we had one million infected about 10 days ago
8. (question: anyone knows the exact average time between infection and death? any better source?)
9. Because 10 days is the approximate period in which cases spread by a factor 10, then my bet is that we have right now about 10 million people carrying the virus in Spain, about 20% of the population. Unless the strong measures undertaken 10 days ago took effect.
10. To avoid bias ni the tests, what is really needed is random tests of the population. Fortunately, this has been done already, in Iceland: government.is/news/article/2…
They find 0.86% present infection rate, whereas before this they 473 cases in Iceland (0.13%).
They find 0.86% present infection rate, whereas before this they 473 cases in Iceland (0.13%).
11. So, they had underestimated the amount of infected people by a factor 6 when using the tests only on people with symptoms!
And this is how all other countries are estimating the number of infections...
And this is how all other countries are estimating the number of infections...
