Hacker News Reader

Hacker News

Comment by contravariant | Hacker News Reader

Parent Full thread

contravariant

It's only a 3%(-ish) chance of dying going by numbers published so far.

That's much higher than a regular flu though.

View on HN

The numbers published so far are probably biased to the high side because only serious cases seek treatment.

But it spreads much faster... 3% of a billion people infected is quite a lot.

Otherwise most reports cite that the dead were people who were already on their "deathbeds", and would (probably) die from a normal flu + complications (eg. pneumonia) anyways... but there are a few cases of younger people who died, that wouldn't if this was a "normal" flu.

Also depends on which demographic has been infected so there's of course some general bias as we don't have an even distribution across all age groups, sexes and profession, level of activity and and and ...

For example, if predominantly senior humans have been infected, the statistics may be skewed towards that group and you won't be able to make flat statements about younger people getting the virus.

I also still think that we don't really have solid numbers on the virus due to:

- unconfirmed sample characterization - likely false data spread on behalf of the Chinese government in order to save face/contain panic

It's ~20% "severe" plus fatal, per China CDC. Since many people who would end up dead is currently not dead, in-progress fatality rate is not very informative.

How do you get to that number?

I'm just dividing the number of (known) deaths by the number of known infections.

As other's have pointed out this isn't the most reliable way of doing it, but it gives some indication.

No, it does not give any information for a multitude of reasons.

1. Deaths are lagged by known cases

2. Both are skewed distributions

3. These people might have received professional care when it was available. During an outbreak, the available health care will decline.

Those just mean the estimate is inaccurate. When dealing with something entirely unknown almost any estimate is better than nothing.

Sure it could end up being off by a factor 10 either way, but even then it's still informative.

Estimations work when dealing with symmetrical distributions. You might be wrong either way. But these distributions are fundamentally not symmetrical! Also, you should compare deaths with number of cases a couple of days ago (like 7), since that’s the time between diagnosis and death.

The distributions are asymmetrical in exactly the way that argues against your fear-mongering across this thread: only the most serious cases are self-selecting for treatment and testing.

The infection rate is, in all likelihood, much higher than these early numbers suggest. People don’t go to the hospital for sniffles, and therefore, the bulk of infections don’t show up. Deaths almost always get counted.

How do you figure that? Going off this source I see 170 deaths and 133 recovered. So of those 303 that have been infected with it, 170 have died which is 56%. You can't count the other 7783 that are currently infected since we haven't seen their outcome yet.

https://gisanddata.maps.arcgis.com/apps/opsdashboard/index.h...

The estimate requires some extrapolation based on other epidemics. Dying from it isn't going to take as long as recovering so you should expect a particularly large skew this early. But more generally you'd expect more mild cases to go underreported, be a lower priority for testing (which seems to be the bottleneck for reporting cases in some areas), etc.

This is not the right way of counting it. If it takes 40 days to be declared fully recovered, the recovered number is going to stay low at least 40 days (even though the people on their way to full recovery will most likely survive), while the death number will be increasing from the pool of people which became infected later.

To count it properly you need to follow what happened to people infected at the same date.

Interesting, that makes sense!

Reply on news.ycombinator.com