[ericb]

Keep in mind, this is a self-selected group of people. There could be 10x more asymptomatic, or low impact cases who thought they had the flu, but didn't feel bad enough to go to the hospital. Imagine if I tried to capture the survival rate of the flu, but chose my population of patients from the ICU. Would the numbers be right?

I think this is born out by the really small number of children showing up in the mortality numbers. Proportionally, kids usually over-represent transmission, I think, from kid-level hygiene and being crowded into small classrooms. They seem under-represented in the study population, which makes me think there could be a lot of selection bias being captured.

Additionally, the standard for recovery is really high and is built in a way that it will severely lag deaths (which have no waiting period). I think it was something like total clearance of the virus, not "feels good, now surfs reddit all day."

[cstejerean]

There could also be many X more cases that haven’t been confirmed because they can’t get access to transportation to a hospital, or because the hospitals are overwhelmed, or because they are running low on test kits or behind the schedule on testing. I wouldn’t assume that the unreported numbers are strictly positive.