Hacker News Reader

Comment by mrb | Hacker News Reader

If this study is accurate, then 1.8 million NYC resident are infected. So 15 411 deaths (as of today, including 5 121 probable deaths) suggests COVID-19 has a fatality rate of MINIMUM 0.9% which is in line with previous estimates (between 0.5 and 1.6%) and this is MINIMUM 9 times more fatal than the flu (0.1%).

Edited to add "MINIMUM", as there is a lag between case detection and death (I should have known better to mention it, as I blogged about it: https://blog.zorinaq.com/case-fatality-ratio-ncov/)

View on HN

> If this study is accurate, then 1.8 million NYC resident are infected. So 10 290 deaths (as of today) suggests COVID-19 has a fatality rate of 0.6%

No, it doesn't, because the usual time to death, if it causes death, isn't zero, so you need to compare the number of deaths a couple weeks from now with the number of infections today to get a closer-to-accurate number.

Also, you are using the 10,290 confirmed deaths and i ignoring the 5,121+ additional identified probable COVID-19 deaths. So, between those two things, the actual fatality rate if the current infection estimate from this study is right is probably 2-3 times the figure you cite.

Interesting thought experiment: if NY daily deaths over the next two weeks average 400, that's 5,600 upcoming deaths from people presumably infected now. If about one-third of deaths are not accounted for (5,121 unidentified as a portion of identified+unidentified), then we'd have about 8,400 new deaths projected over two weeks including the likely unidentified deaths.

As a percentage of those who were infected as of this study, that's a 1.3% infection fatality rate.

Obviously some conjecture and estimates there, but it seems to be on the high range of early estimates (though much lower than the 3%+ rates derived from CFR with undercounted denominator).

You can derive that from better data. From https://covidtracking.com/ we can see that 57k people have been hospitalized in New York State so far, with 15k still in the hospital. Of the 42k who aren't, 16k died[1], so we'd expect that about 5k of the people currently in the hospital won't be coming home (some of these statistics are pretty depressing).

So two methods give us similar counts, matching up very closely with the separately measured CFR from way back in the original outbreak. That's pretty good back-of-the-envelope work.

[1] this isn't quite right as not every fatality was a hospitalized case.

You are correct, edited.

The time to go positive on an antibody test is not zero either, probably 7 to 14 days. Also the infection rate is lower now than it was in the past. I doubt this adds up to more than 50% higher mortality.

Cuomo made the caveat during the press conference that these samples were collected in grocery stores etc, so might be higher than people at home etc

edit: and also potentially lower than in the essential worker populations. So some caveats to keep in mind

omg yes - tons of people switched to grocery delivery esp since NYC has so many terrific options and virtually every grocery, market & specialty store already delivered, even before COVID.

The city itself actually estimates a total of 15,411 deaths so far the the figure is already closer to 0.9%, and there are still people dying in the city every day so will likely end up a bit higher (of course, all of this assuming the study actually represents reality).

https://www1.nyc.gov/site/doh/covid/covid-19-data.page

So .. drawing a line between two points, if NY had 20% immunity for 15k dead, that implies there will be 60k dead by the end of the process. Roughly one 9/11 every week or so?

You also have to adjust for the people who are already sick but haven't died yet. That would likely get you somewhere between 1 and 1.5%. So for herd immunity in the US at 200,000,000 recovered, that would be 2 to 3 million. That's consistent with earlier projections.

one problem with all these projections is that they all assume everyone is the same and that it is just a matter of chance. Not true in the least bit.

those most susceptible get the disease first, those most sensitive die first -

I believe the ratios will not stay the same, but continuously tick downwards

On the other part, those who are more vulnerable are more aware, and are implementing more measures in their daily living, hence they are not getting infected as much, currently.

Hope it stays that way.

Not sure that makes any difference.

Not only are death rates sharply higher in older people, but their infection rates relative to the population are also higher as well.

Without better testing, you have no way of knowing whether older people have higher infection rates. Infection rates depend on far more variables than age. Witness what's happening on the Teddy Roosevelt.

Making it the worse pandemic in decades but no where near the Spanish Flu.

Hard to compare directly against the Spanish Flu as medicine has had 100 years of progress, but assuming you are talking about overall impact, agreed.

not to mention the Spanish Flu hit children just as hard as the old.

The age distribution was also significantly different then.

The Spanish flu had a smaller outbreak in the Spring and came back with a vengeance the following Fall, so hold that thought.

https://www.history.com/news/spanish-flu-second-wave-resurge...

Or as bad as the 1957-8 "Asian Flu" that most people haven't even heard of.

Well, sure, but the “Spanish” flu (only called that because Spain is the first country severely hit that didn't lie and cover it up because of WWI) was accelerated by both the physical conditions created by WWI and the fact that covering it up to avoid revealing weakness to wartime opponents hampered response in many early-hit countries and obstructed information flow.

It's also nowhere near the HIV/AIDS global epidemic in likely total impact, but it's a lot more acute in it's impact than HIV/AIDS.

> only called that because...

Also worth noting that a huge number of names in biology are quite arbitrary and/or accidents of history.

Or tech, for that matter. Fewer and fewer "hard disks" involve any actual disk, and more non-floppy floppy disks were made than actual floppy ones. Etc.

I disagree. It's simply too early to make that determination.

The Spanish Flu had a estimated infection fatality ratio of >2.5% (https://wwwnc.cdc.gov/eid/article/12/1/05-0979_article). As another commenter pointed out, the COVID-19 IFR is 0.9% in NYC. The difference between 2.5% and 0.9% is only 2.8x. However the IFR is not only underestimated due to the lag between case detection and death, but it's also probably a lot worse in many other parts of the world (third world countries, etc). So it's possible the difference is only 2-fold, which would make both pandemics relatively comparable.

Another factor to consider is that healthcare today is much better than when the Spanish flu occurred and that could explain the 2x difference.

I hope that the numbers stay at .6% and that the difference from previous estimates might even be improvements we've made in figuring out how to care for people suffering from Covid-19. But I remember being happy at how low South Korea's numbers were when their epidemic peaked and receded then watching the case fatality rate slowly climb up to the 2.2% it's at today. So I'm not counting any chickens just yet.

Be careful not to conflate CFR with IFR. CFR is the number of deaths divided by the number of confirmed cases. IFR is deaths divided by total infections. It's hard to determine IFR, but CFR is almost always an exaggeration of the risk.

Yes. Hiwever, it is valid to compare the CFR of one pandemic to the CFR of another pandemic if you find the CFR from the same relative point in time of that other pandemic.

Yeah, but with the success of contact tracing in South Korea I'm assuming IFR is within a factor of 2 of the CFR or so.

The study is biased in that it pulled samples from a grocery store based on people opting in. If I walk in to a grocery store and someone offers to give me this test, I am going to say no for 2 reasons:

1) I have not been sick since last fall

2) I don't want to take additional risk getting bloodwork done by someone who is testing people for covid19 antibodies. I want to take a little risk as possible and return home to my family.

So basically this test only possibly grabs people who are willing to risk going to the grocery store rather than having food delivered, and who are willing to take on additional risk of getting tested. It is even less likely that a person not taking risk, who never had any symptoms would get tested than a person who had had symptoms.

The entire premise that this represents the population as a whole is completely flawed.

these comparisons to the flu usually don't seem to factor in transmissibility. seems you cant leave that out of the equation and only look at death rate

It’s important to know how risky it is to get this thing. That way we can make informed decisions about opening up.

And also the risks of non-death injury from it. Long-term reduced lung function isn't good.

I feel a blanket IFR figure isn't all that helpful.

Won't there be a high level of age stratification in there. Over 60, death rate will be way higher than if you are younger and probably close to zero under 20.

It seems better if we could have a risk per age group to see who can go back to a somewhat normal life and protect the most vulnerable.

Also,

I don't the flu's fatality is calculated against every exposed to the flu but against everyone who gets sick with the flu. If the flu also 50% asymptomatic exposed individuals, that would know it's fatality rate down further for comparison's sake.

We calculate both. Case fatality rate (CFR) is based on patients who test positive, whereas infection fatality rate (IFR) is based on all infections. We don't yet have adequate random sampling to have confidence in IFR estimates.

Please don't use uppercase for emphasis. If you want to emphasize a word or phrase, put asterisks around it and it will get italicized.

Note that deaths lag positive cases by some time (2 weeks maybe?).

It's also lower than the 3% that was predicted that people used to justify the shutdown and closer inline with the swine flu or rotovirus.

Nobody ever seriously suggested it was 3%. The estimates were always around 1%

https://www.cnbc.com/2020/03/03/who-says-coronavirus-death-r...

I mean there were articles like this all of march. I'm not saying the WHO or CDC might of said something different but lets acknowledge there were articles proclaiming a 3% death rate.

Lots of people were seriously suggesting >3%.

Edit: Removed the term revisionism since I wasn't aware of the historical connotation.

An IFR of 0.6%, with 2/3 of the US infected (about what you'd need for herd immunity at an R0 of 3), means ~1.3 million deaths. That's well in line with the models. The distinction between CFR and IFR was not news to the experts.

Regardless of initial estimates, how is 0.6% - 1% "closer inline" to 0.026%[1]? Even based upon the low range that's 30X more fatal.

[1] http://news.bbc.co.uk/2/hi/health/8406723.stm

zso your saying its closer inline with the seasonal flu. With a 0.1% death rate.

https://www.healthline.com/health-news/how-deadly-is-the-cor...

The point still stands is that enough to warrant the current strategy compared with countries like Korean or TW.

The shutdown addresses more qualitative aspects of the situation. Even if flattening the curve doesn't reduce the area under the curve, it still avoids a potential large numbers of otherwise avoidable deaths because fewer people have it at once, keeping hospitals under capacity, and more people will get it later when we know better how to treat it.

And the earliest predictions have to be the least accurate, because they're based on the least amount of data. But early on is when you need to act.

> It's also lower than the 3% that was predicted that people used to justify the shutdown and closer inline with the swine flu or rotovirus.

The shutdown wasn't, and isn't, based on the fatality rate under current conditions as much as the shape of the curve of various medical resources being required (total hospital beds, ICU beds, ventilators, PPE for healthcare workers, etc.), the anticipated speed of being able to bring more online, and the likely impact of exceeding available supply on death rates and other health outcomes, both for COVID-19 victims and patients with other health conditions. (A potential problem which was significantly realized in Italy, where things like ventilators were for a while allocated based on other factors besides medical need, since the number with need exceeded the supply.)

It's important to remember that the fatality rate of a disease isn't a constant, and changes to conditions (like, we can no longer provide ventilators to everyone who needs one) can change that rate.

You're conflating CFR with IFR.

Reply on news.ycombinator.com