Let’s start by assuming that we believe the official numbers. I know it is a big if, but that is a whole different can of worms. We’ll get to that later. What I want to do here is to point out a fundamental, basic error that is widely made in estimating the Case Fatality Rate (CFR).
On 7 Feb 2020, 22:17 Singapore time, the latest figures are:
Confirmed: 31586 Deaths: 639 Recovered: 1777
A large number of commentators, including mainstream media, report this as a 2% CFR, calculated as 639/31586 = 2%. While this is worse than the seasonal flu, it is much less than SARS (10%) and MERS (30%), so no need to worry, they say. We, the smart informed people who watch CNN and the like, can feel smug, having made our part in avoiding a panic by relaying this reassuring information to our friends.
HAVE THESE PEOPLE NEVER HEARD OF COHORT ANALYSIS?
Of the 31586 confirmed cases, about 85% have been diagnosed in the last 10 days. Many are in worsening condition and we do not know if they will die or recover. Therefore, to estimate the CFR, we can do one of two things:
- We can consider only the cases for which an outcome is known: dead or recovered. This would lead us to estimate the CFR as 639/(639+1777) = 26%.
- We can assume that after a certain time period, e.g. 1 week, the patient has either died or recovered. A week ago, there were 9803 confirmed cases. This would lead us to estimate the CFR as 639/9803 = 6%. Obviously, we do not know the right lag, but something between 1 and 2 weeks seems plausible given what we know about the disease. And it is certainly not zero.
Conclusion: Although there is a lot of uncertainty in estimating the CFR, it is certainly not 2%. Something between 5% and 30% seems more likely. Interestingly, this is in line with SARS and MERS. What makes 2019-nCoV potentially more dangerous is that transmission by asymptomatic patients seems possible, and even easy.
Now, all that is if we believe the official numbers, but that is another story.