There has been a lot of debate about how deadly #covid19 really is, with various estimates being bandied about, some as low as 0,04%. I was asked about this on German radio yesterday and I want to briefly share what I said: thread
Caveats first: Deadliness is not some fixed number for a pathogen. It depends on the age structure of the population for instance, on how many people have pre-existing conditions and on the quality of care for sick patients, how many ICU beds are available etc.
CFR is often calculated by simply dividing number of deaths by number of cases, but that is a crude estimate, because we know that not every infection leads to a disease and not every disease is diagnosed and counted. We really want to know how likely infection will lead to death
As with everything related to #SARSCoV2, a virus that we have barely known about for three months, there is a lot of uncertainty here. So let’s try and look very roughly at an upper and lower limit first.
China has reported 82,240 cases and 3309 deaths. That gives you a case fatality rate of roughly 4%. No-one believes that China found all the cases given that we didn't even know the virus existed in the beginning and had no tests. So this can work as a rough upper limit.
In Italy we have had 11,591 deaths. Let’s assume for a minute that every single person in the country has already been infected and no more deaths will come (obviously not true). Given there are 60 million people in Italy that would be CFR of 0,02%.
So that’s something we can say with decent certainty: The real case fatality rate is somewhere between 0,02% and 4%. That’s a wide spread of course, a factor of 200. So where in that space are we likely to end up. We have some datasets to go on.
Take the Princess Diamond: The ship had 3711 people on board and because of its special situation, quite a lot of testing was done. They found 619 infections with #covid19. 7 died. That’s about 1 percent. (again: very crude, older population etc.) Paper: eurosurveillance.org/content/10.280…
Several other models and papers arrived at similar values. And many countries that have done a lot of testing also find a CFR of 1-2 percent. Currently crude estimate in Germany is 1%, South Korea is 1,7%.
Given the time lag between infections and deaths those rates are likely to go up a bit. Many infections already counted in Germany and S Korea will still lead to death. Passengers from the Diamond Princess may yet die of #covid19.
On the other hand we are not finding all infections. Is it realistic that we have only diagnosed about half of all #covid19 cases in Germany? Absolutely. Maybe in a more extreme scenario only every fifth. Even that would still mean 0,2-0,4% infected die (instead of 1-2%).
Only serology (looking for antibodies to identify people who have had infection and weren’t diagnosed) will give us a good answer. For now a range of 0,5-1% seems realistic IMHO with all the uncertainty that comes with it and the caveat that it will differ from place to place.
Two more points: 1. I’m a journalist not an epidemiologist. I’m sharing a crude way of how I talk/think about what we know re deadliness of #covid19 here without getting deep into the weeds. For actual experts, check out @AdamJKucharski, @C_Althaus, @DFisman and others.
@AdamJKucharski @C_Althaus @DFisman 2. We should not focus on deaths alone. Main point for the health care system is serious disease overwhelming hospitals and ICUs. And we haven’t even talked about possibility of long-term sequelae the virus might cause. We just don’t know.
