Hacker News Reader

Comment by cm2187 | Hacker News Reader

Another datapoint suggesting the death rate of this virus is within 0.5%.

I had the opposite conclusion. There are between 10k-15k COVID-19 associated deaths in NYC, depending on how you count (the higher rate is consistent with the excess-mortality data, I believe). That by itself produces a range of ~0.56%-0.86% if you assume 21.2 % of NYC has been infected (larger range if you allow for some error on the infection rate, obviously). Moreover, we don't know what fraction of the currently-infected who will die have already died. A reasonable guess might be 50%, which would mean that the fatality rate for NYC is somewhere between 1.1 and 1.7%.

Either way, based on NYC, 0.5% might be a hard lower bound on the all-population fatality rate (of course, maybe NYC is not a representative sample for some reason)

“ Moreover, we don't know what fraction of the currently-infected who will die have already died. A reasonable guess might be 50%” I lost you at reasonable guess. Why not ~1%?

If the mean time between infection and death is 2 weeks (Mostly true), and the virus has doubled every two weeks (True in NYC), then assuming that not a single new infection takes place starting today (Obviously not true, but this is a hypothetical), then in two weeks, you would expect to have twice the deaths that you do right now.

1% would imply an unrealistically high-fatality rate, so that can't be right.

The virus takes time (2-3 weeks) to kill people, and the prevalence was presumably measured within the last week.

Therefore, we have to take out the cases that occurred in the last week or two. I would suspect that 50% is on the upper end of how much the death lag could skew the results, but 20-50% doesn't seem unreasonable.

0.5% comes from Cuomo’s presentation [1].

Also your 50% seems very high to me. My understanding is that 50% is about the death rate of covid patients requiring ventilation, so the overall death rate of people hospitalised for covid must be well below that.

Also if you factor this timing impact, you have the same timing impact on the denominator, i.e. people who are infected but haven't developped antibodies yet.

[1] https://youtu.be/TisDYYWJgBA?t=967

[edit] in fact for the 50%, we have an idea from Cuomo's presentation charts. The gross daily hostpitalisation rate as of 10-15 days ago was about 2000-2500 per day, and the number of death now is about 500 per day, so that suggests roughly a 20-25% death rate for hospitalised people (I am sure there must be some more precise figures somewhere).

50% is not the death rate, but the fraction of those who are an infected and have died to those who are infected and who will have died (since there is some lag between infection and death).

The death rate of those requiring ventilation in NYC is closer to 88% https://jamanetwork.com/journals/jama/fullarticle/2765184

Outside the context of this virus, for adult men, the survival rate one year after mechanical ventilation might be 30 percent: https://www.ncbi.nlm.nih.gov/pubmed/8404197/

no, the death rate for ventilated covid patients is something like 80-90%. this is based on a NYC study published yesterday.

It’s not clear what kind of antibodies they’re testing for, but it can take up to 4 weeks after infection to develop detectable levels of antibodies. It takes on average ~3 weeks after infection to die. There needs to be some time lag adjustment in your calculation, which would likely result in a lower IFR.

IMO the two effects probably about cancel each other out and we’re looking at <1% IFR for the population tested (AFAIK children weren’t included).

I guess I would assume they would have corrected for any time-lag in the result, but I have not read the study in detail

Right now it's just a press release so there's not much to read. Cuomo's statement reads to me that is just demographically adjusted though.

Yes, but the age is a 10x factor, os CFR would be much different in Nigeria (average age is about 19) and Italy (average age 45). We have two good petri dishes:

USS Theodore Roosevelt - 600 cases out of 4800. 5 sick, one dead (41 years old). which means hospitalisation is at 0.83% and CFR is 0.016% (1% dead or in hospital). Not idea what is the average age, probably around 30-35.

Diamond Princess: 700 of 3711. CFR 1.1%, average age around 50.

What worries me the most though is potential mutations and death rates when you are infected with Covid-19 and flu at the same time (or other pathogens).

https://www.reuters.com/article/us-health-coronavirus-usa-mi...

https://www.nature.com/articles/d41586-020-00885-w?utm_sourc...

Average age on an aircraft carrier appears to be 19 [0], 20 [1], and/or 24 [2] years old.

0: https://www.nytimes.com/2008/04/20/arts/television/20jens.ht...

1: https://archive.defense.gov/news/newsarticle.aspx?id=15334

2: https://mashable.com/2016/06/23/what-its-really-like-to-live...

> which means hospitalisation is at 0.83% and CFR is 0.016% (1% dead or in hospital).

0.16%, not 0.016%.

You can't measure CFR or IFR this quickly in an isolated environment like the Roosevelt due to mortality lag.

Also, the CFR for Diamond Princess is now up to 1.83%.

Mortality rate, case fatality rate and infection fatality rates are all different things and have different values.

Infection fatality rate of the virus could be within 0.5% (because its the one that tries to take into account everyone who has been infected), mortality rate is very likely still the one WHO has published and the third, case fatality rate is ~38% in the US.

What? 38 out of 100 infected in US die?

Infection fatality rate, the big number everyone is trying to figure out, best guess is somewhere around 1% right now, with a confidence interval of something like 0.3%-2.5%

Case fatality rate, calculated as "number of deaths" / "number of known cases", usually something like 10% in most countries, although occasionally much lower or higher for the odd country that has really good or really bad testing.

Resolved case fatality rate, calculated as "number of deaths" / "number of recoveries + number of deaths", which has been around 40% in the countries that have been hit hardest for the duration of their outbreaks. Tends to go down after the curve is bent, because it takes people longer to recover than die in most of these cases.

Also another data point suggesting that the virus is much more contagious than the early estimates said.

I personally don't see how the test, track and trace efforts that have been touted as the only way to open things up can work with a virus with these characteristics. You would need to test a huge percentage of the population each day with quick results tests and something like that is many months away from being possible assuming it ever would be possible.

I agree. But it seems to have worked in Korea. The math doesn't seem to reconcile somehow. Actually probably the prevalence of masks explains the difference.

The math seems to point more to there being another factor to what makes it severe, in my eye.

Going by the news we have seen, elementary schools could have 80%+ infection rates and fewer deaths that NY has seen. Similarly, are there other factors besides age? (Age feels way more like a proxy.)

More contagious OR began community spread and became endemic much earlier.

r0 is part of the formula for total infected at X/Y/Z date but so is date of initial infection and size of initial infected population.

The key is testing, contract tracing, and quarantining when the number of cases is small. That's what South Korea, New Zealand, and a few other countries did and they've all suffered relatively few fatalities. You still need a massive amount of tests to contain an outbreak, but the upside is that if you contain it early you don't need toe enact clumsy and broad measures like lockdowns.

I wonder if we did these preliminary tests with the flu if we would see a large death rate drop for influenza as well

The US's influenza mortality figures are a combined influenza and pneumonia figure because they usually don't even bother actually screening for influenza.

Research attempting to determine how much of that figure is infected with influenza have found figures like 1/3rd (though with large differences year to year).

The death rate would go up if more people had the virus at once, which is the entire reasoning behind isolating.

I am wondering that number, if the limiting factor is ventilators, but the CFR of people on a ventilator is close to 90% either way.

Other cases may just need oxygen supplementation, which is more manageable.

It is pretty clear now that Germany is not that far from from Italy in terms of CFR when controlled for age.

15,000 deaths (So far) in NYC. 8.4 million people. If 20% of the population has had it, the death rate is 1%... Assuming that none of the currently infected die. Which is an unreasonable assumption (As the mean time to death from the point of infection seems to be ~2 weeks, and infections have doubled over the past 2 weeks.)

But those currently infected are not included in the 20% of the population that has had it. Or maybe some are, but most of those infected now who may still die, do not have antibodies yet.

…when people are able to get treatment.

Reply on news.ycombinator.com