It's definitely underdiagnosed. Death toll is probably underreported too. But since they've been prioritizing (according to reports anyway) people who are severely sick to give tests to, I'd guess "diagnosed", proportionally speaking, is much more underreported.
Deaths/diagnosed(reported/confirmed) in headlines:
January 27: 82/2835=2.89%
https://www.livemint.com/news/world/...124072028.html
January 28: 106/4545=2.35%
https://www.nbcnews.com/news/world/c...break-n1123561
January 29: 132/6078=2.17%
https://www.washingtonpost.com/world...617_story.html
I don't think the virus is less deadly. I think the tests are starting to filter down to people less sick. It's also worth noting it's very unlikely they are only testing like 1500 kits per day. That diagnosed/confirmed/reported is going up much slowly than number of kits tested suggests most of people tested had negative results.
For some perspective, since they are prioritizing testing people who are severely sick, the number of cases "diagnosed/confirmed/reported" is probably something akin to (should be higher) to "hospitalized" in CDC numbers. 8-10% of people hospitalized for the flu end up dead in the US.
Another way to look at this is comparing the stats known so far to deaths caused by influenza and pneumonia:
https://www.kff.org/other/state-indi...2:%22asc%22%7D
At the US average of 14.3 per 100,000, just under 1600 people out of Wuhan's 11 million per year, on average, die to the flu and pneumonia. The same stats applied to China's 1.4 billion population: right about 200,000 a year to the flu+penumonia. About 47k for US's 327 million people in case you're wondering.
There is a Bayesian analysis final exam question for a stats class in here somewhere. Something like blah blah blah, given these stats, if a patient dies of pneumonia, what's the probability his/her death is attributable to coronavirus?
Last edited by grizy; 01-29-2020 at 09:54 AM.