I think for approach 2 you want the ratio of deaths:(deaths+recoveries) rather than just deaths:recoveries. This means that the 2 approaches actually give the same result.
This method may lead to an overestimate as deaths occur earlier in the illness than declarations of recovery. If deaths occur at a constant rate between diagnosis and recovery then it would be more like half this value (depending on how the infection rate is changing). Looking at it like this gives 5.8 − 11.5% mortality.
An alternative method would look at average time from diagnosis to death and look back to how many diagnoses there were that many days ago. For Sars/Mers this looks like 10-18 days which would give death rates for Covid-19 of 4.4% − 12.4% which is pleasingly similar to the above method but might not actually mean anything!
I think for approach 2 you want the ratio of deaths:(deaths+recoveries) rather than just deaths:recoveries. This means that the 2 approaches actually give the same result.
This method may lead to an overestimate as deaths occur earlier in the illness than declarations of recovery. If deaths occur at a constant rate between diagnosis and recovery then it would be more like half this value (depending on how the infection rate is changing). Looking at it like this gives 5.8 − 11.5% mortality.
An alternative method would look at average time from diagnosis to death and look back to how many diagnoses there were that many days ago. For Sars/Mers this looks like 10-18 days which would give death rates for Covid-19 of 4.4% − 12.4% which is pleasingly similar to the above method but might not actually mean anything!
Yes, bad mistake there! Thanks.