For example, using the simple methodology, SARS' death rate was only about 3%, and that was the number most media used during the outbreak. But years after the epidemic ended, once researchers had greater access to medical records, the death was determined to be 10-15%, or roughly 3-5x higher than originally reported.

In the cases of NCV, there is already significant evidence of under reporting of fatalities by Chinese authorities in Wuhan (some sources suggest that up to 90% of NCV deaths in China may be unreported).

And as the cruise ship stuck outside of Japan demonstrates, the virus may be far more virulent (i.e., contagious) than previously thought.

Followup: https://www.nytimes.com/2020/02/12/world/asia/china-coronavi...

China just changed how it classifies NCV cases to reduce the number of cases by eliminating those who are infected but not symptomatic meaning the numbers are even less reliable than they were before.

The problem with this is that a number of the superspreaders, like the British fellow that infected 11 other people, and the passenger that infected the cruise ship quarantined in Japan, were not symptomatic when they were infecting others.

It does, but that's a known factor which all medical professionals are aware of. The true number is unknowable until the epidemic is over, so they give the best number we've got in full knowledge that the numerical value will be wrong.

This seems dangerous, if it leads to unwarranted complacency among those who take the number seriously. Do you know the rationale behind reporting a number known to be skewed in this way -- is the idea that it is roughly canceled out by other factors (e.g. the number of undiagnosed mild cases) that push toward an overestimate?

That is a factor, but bluntly put, the ones who are reporting this single number aren't medical professionals.

Those might still use it in conversation amongst each other, sure, but any actual analysis will be based on far more data than a single number. Fundamentally this is the press trying to boil things down to a double soundbite. As usual.

The case fatality ratio is literally defined as deaths over cases, and that's not useless, but it isn't something you should spend much time thinking about either. Reality is always more complex than any single number. Probably you want to know something like "What are the odds of dying if I'm infected" -- that's a different parameter entirely.

I'm not so much worried about how the experts interpret this number behind closed doors, but how everyone else reacts to it. This thread began with someone confidently saying that the 2% death rate will decline as we get better data, and using that to argue against the cancellation of the MWC. I've also seen about a million thinkpieces and social media posts telling us that we're crazy to worry about this virus because the flu is so much worse -- and while they're mostly focusing on the fact that the current number of cases is small compared to seasonal flu, they might not be so dismissive if they knew that 2% could be a significant underestimate of the mortality rate. And I suspect that many (most?) countries' quarantine and related policies are affected by public opinion, rather than being a pure function of expert advice.

It's an issue.

2% is almost certainly an underestimate, in this case. I read an article the other day which did statistical analysis on various groups of cases; it came up with a CFR of 18% for Wuhan (95%, 11% to 80-something), and while the expectation for cases in Singapore is still a little over 2%, that's because the medical system in Singapore isn't overwhelmed.

Reporting that would be an equally bad idea, though. The case fatality ratio only counts infections that are severe enough for the patient to go to hospital and become a case, and we know a large fraction of infections don't. Probably over 90%. Which would push that back down to 2%.

Not that a death rate of 2% of all infected is anything short of apocalyptic, but that's why they're trying so hard to quarantine it.