>>114296742/2
>6.5% fatality rateLet be the ratio between recorded infections and actual infections
Let be the ratio between recorded deaths and actual deaths
Let be the recorded mortality rate and the actual mortality rate, then
where . We want to estimate , thus we must try to determine and . Let us assume that we have an time-adjusted* estimate for and let us try to estimate . Assume . I think this is reasonable as it is easier to record deaths than merely infections. This implies that
thus the actual mortality rate has to be LOWER than the recorded mortality rate .
Notice that this result is completely independent on the numbers we are working with. No matter how much fault there are in the numbers. The actual number of dead could be 50 times higher than the recorded and this result would hold as long as the actual number of infected is 51 times higher than the recorded. I'd be very eager to hear arguments for why we cannot make this assumption.
* Median time from emergence of symptoms to death for those that die is 14 days (with range 6-41 days) [
https://doi.org/10.1002/jmv.25689]. There will also be a delay from symptoms emerge to people get a positive test result for that minority that get tested. Say 2-3 days. Putting all of this together yields an estimate of median time from positive test result to death at about 11 days.
