COVID-19 fatality rate may be much lower than previous estimates

[GenXguy]

I hadn't checked this in a while, but the death rate among closed cases in the U.S. is much higher than I expected.

https://www.worldometers.info/coronavirus/country/us/

 

[RonBoyd]

How does (would) knowing the exact, verifiable, number of deaths make one feel safer?

 

[DenverGuy42]

How does (would) knowing the exact, verifiable, number of deaths make one feel safer?

Having accurate numbers allows decision makers to make sound decisions.

garbage in, garbage out.

One of the big discussions people are having is "the death percentage is no where near what they told us, it is very very low. This is a nothing burger. Open everything up, go back to normal now.". (look at all of the protests going on)

Well, it we don't know

* infection rate (complete lack of ability to test sufficiently)

* hospitalization rate (we actually do know this)

* recovery rate (we don't know this because of the testing)

* death numbers (we only partially know this and will have to guesstimate the number of deaths outside of hospitals due to covid-19, this is due to inability to test properly for covid-19)

* percentage of those that get infected that actually die from this (we don't know infection rate and death numbers - we cannot accurately compute this)

We cannot make truly sound decisions about how to proceed in the future.

Good data in = Better decisions out.

 

[Sunset]

How does (would) knowing the exact, verifiable, number of deaths make one feel safer?

If the numbers are low, who cares, example 15,003 vs 16,100 for the USA, is tiny and less than annual auto deaths. That would make me feel safer.

We are on track to exceed annual auto deaths by a wide margin, so I feel less safe.

Estimates are the 1918 Spanish flu USA death total was 675,000 (that is more than any war including the Civil War for US deaths). But remember that was a much smaller population back then, so as a percentage of the population it was higher and was about 0.5%

Same percentage today would be approximately 1.7 million
I'm hoping we stay below that record, and that size of deaths would not make me feel safe, as there is no actual rule saying this has to be the top limit.

 

[Gumby]

I hadn't checked this in a while, but the death rate among closed cases in the U.S. is much higher than I expected.

https://www.worldometers.info/coronavirus/country/us/

We have discussed this before. You cannot use that metric to determine the death rate. Among other problems, the reporting on case resolution is grossly inaccurate. For example, in my state of Connecticut, the worldometer site says there are 17,550 total cases, 1086 deaths and 16,399 active cases - implying only 65 recovered. Using your metric (dead/dead +recovered) would generate a 94% fatality rate.

And, yet, the figures released by the state department of public health yesterday (as they do every day) list only 1938 people currently in the hospital with COVID-19. Ergo, there cannot be only 65 recovered people.

If you looked at actual recorded deaths versus total positive tests (which to date have only been done on people exhibiting symptoms sufficient to get a doctors order for the test) the rate is about 6%, and that is almost certainly overstated due to asymptomatic cases and a lack of testing.

 

[ejman]

We have discussed this before. You cannot use that metric to determine the death rate. Among other problems, the reporting on case resolution is grossly inaccurate. For example, in my state of Connecticut, the worldometer site says there are 17,550 total cases, 1086 deaths and 16,399 active cases - implying only 65 recovered. Using your metric (dead/dead +recovered) would generate a 94% fatality rate.

And, yet, the figures released by the state department of public health yesterday (as they do every day) list only 1398 people currently in the hospital with COVID-19. Ergo, there cannot be only 65 recovered people.

If you looked at actual recorded deaths versus total positive tests (which to date have only been done on people exhibiting symptoms sufficient to get a doctors order for the test) the rate is about 6%, and that is almost certainly overstated due to asymptomatic cases and a lack of testing.

I generally agree with your statements but I would like to point out two areas of concern:
1) the ratio of deaths to total cases seems to be going up over time not down as one would expect given that
2) the number of tests is increasing substantially - almost 4 million in the US, of which 3/4 of million were confirmed cases (this data from today's worldmeter results.


So, as more and more tests are performed the ratio should start going down hopefully soon but I wish it would do so already!

 

[cbo111]

I read a stat the other day that historically in NYC, the average number of people dying at home is typically 20-30 a day. In recent weeks they are seeing 200+ home deaths a day. The vast majority of these home deaths are not tested for COVID. That alone tells me nationwide deaths due to COVID are being probably underreported.

 

[Sojourner]

If confirmed, that means that the fatality rate is actually something like 0.1 - 0.2 percent, not the 1.0 - 3.0 percent that some epidemiologists estimated. A fatality rate of 0.1 - 0.2 percent would be similar to regular influenza.

While I'm very hopeful this turns out to be the case, I'm skeptical that the CFR (case fatality rate) is really that low. Primarily, this is because COVID-19 is a coronavirus, very closely related to SARS and MERS, and not an influenza virus. The CFR for the 2003 SARS pandemic turned out to be 9%, and the CFR for MERS is roughly 35%. It seems unlikely that COVID-19 would have a CFR as low as the seasonal influenza virus (0.1%), as opposed to a rate more inline with its coronavirus cousins.

Also, how would one explain the roughly 2% CFR in South Korea, a country with extremely widespread, readily available testing?

 

[brett]

One reason for the decrease in the death rate may be the increase in the number of tests.

In our jurisdiction tests are available to anyone presenting any symptoms. My spouse had her drive in test done on Thursday PM. She got the results this morning (Sunday) via a automated voice response.

We expect the covid numbers to rise as more people are tested but the mortality rate as a percentage of those testing positive to drop. The next step hopefully will be some sort of program to focus on contacts for those that have tested positive.

 

[Sojourner]

According to the Washington Post:

Covid-19 killed more people [in the United States] from April 6 to April 12 than any other cause of death (except heart disease) typically does in a normal April week.

And in New York City, there were five times more Covid-19 deaths from Apr 6-12 than from cancer, heart disease, seasonal flu (pneumonia), accidents, and diabetes combined. Just 44 people died from seasonal flu, compared with 3,850 from Covid-19. That alone seems to argue for a much higher CFR than 0.1%.

 

[Winemaker]

The death rate in Sweden (minor mitigation) per million population density is greater than that of Norway (great mitigation) but less than NYC, (big mitigation).

 

[ivinsfan]

So the preliminary estimates are out for the latest flu season.

CDC number of deaths estimated as of April 4 20...24,000 to 62,000 somewhere in between that number I guess.

Hospital admittance between 410-740,000..

I don't know what to make of this huge range that seems to be the best they can do.

 

[donheff]

The Santa Clara County study (video posted above) imputes a high prevalance of virus in the population. If the data is good then we could have a low death rate as we have been discussing. I have be optimistic about this and other reports like it. BUT. if the test is even 98% accurate and the actual prevalence in the population is very low, the false positives could vastly outnumber the real positives making the results less than useful. I'm waiting to hear some level of science consensus around these guestimates before I start thinking this is all a nothing burger. If you want to take a deep dive into the difficulty at getting good estimates from antibody tests of random samples of the public watch this video from Dr. Peter Atia that was posted in another thread.

.

 

[Sunset]

Problem with numbers:

As I see it, a large problem with the number of Confirmed cases, is it really is meaningless until it's in the hundreds of millions, and even then perhaps meaningless.
By this I mean it does not tell us the severity of the actual infections. It could simply mean nearly all these confirmed cases had a sore throat.

Now if they would tell us, number of hospital admissions due to confirmed Covid-19 for the USA (it's hard to find for each state), that would be a useful number to me as I figure anyone admitted to hospital for it, is a pretty serious case.

If they confirm you have Covid-19 by testing, but not admitted to hospital, do they actually follow up with a phone call 2 weeks later to see if you recovered/died/still have it ?

So far, the only number I've found "useful" is the total deaths, which may be too high/low, but given it's hard to fake/hide/inflate it may be a rough estimate of the severity. So far this seems worse than most flu seasons.

 

[aiken]

While I'm very hopeful this turns out to be the case, I'm skeptical that the CFR (case fatality rate) is really that low. Primarily, this is because COVID-19 is a coronavirus, very closely related to SARS and MERS, and not an influenza virus. The CFR for the 2003 SARS pandemic turned out to be 9%, and the CFR for MERS is roughly 35%. It seems unlikely that COVID-19 would have a CFR as low as the seasonal influenza virus (0.1%), as opposed to a rate more inline with its coronavirus cousins.

Also, how would one explain the roughly 2% CFR in South Korea, a country with extremely widespread, readily available testing?

South Korea has only tested 1% of their population. Also, you can't generate a true CFR unless you know the number of people who were infected but had mild/non existent symptoms that didn't cause them to seek a test. Widespread antibody testing would help with that order of magnitude. The current CFR numbers include a severity bias where people who are sick are disproportionately more likely to seek out a test.

 

[urn2bfree]

A couple of data points as food for thought. I mean to draw no conclusions or suggest anything other than we need more information-

Interesting data point number 1- total deaths ALL CAUSES in the USA are 92% of expected for this time compared to deaths in 2017-2019. Mitigation may be affecting death from all causes. -In California auto accidents are less than half the usual.

Interesting data point number 2- on the USS Roosevelt they have tested 94% of the sailors, 600 are positive, 350 have no symptoms, 7 ended up in hospital of these 3 to ICU one of whom has died. That is in young healthy males.

 

[Bigdawg]

Interesting data point number 2- on the USS Roosevelt they have tested 94% of the sailors, 600 are positive, 350 have no symptoms, 7 ended up in hospital of these 3 to ICU one of whom has died. That is in young healthy males.

600 of 4800 living in ridiculously close quarters. 8% infection rate. 7/4800 hospitallized. 1/4800 died. 80% 25 yrs old or younger. Just more numbers to think about.

 

[Kwirk]

The Diamond Princess cruise ship with (complete?) testing of passengers and crew had 713 infections with 13 deaths (so far). That's a 1.8% CFR. I suspect that the population age skewed to the older folk and that a significantly lower mortality rate will apply to our population as a whole. I'll be surprised if the final determination is <0.5%.

 

[Sojourner]

South Korea has only tested 1% of their population. Also, you can't generate a true CFR unless you know the number of people who were infected but had mild/non existent symptoms that didn't cause them to seek a test.

True, except that you should have written IFR instead of CFR there.

Actually, this raises an interesting point regarding the difference between CFR (case fatality rate) and IFR (infection fatality rate). CFR, the one more widely reported, will always be higher than IFR, because in order to be a "case" the symptoms have to be severe enough to warrant being testing. I think what most people actually want to know is: If I get infected with Covid-19, what are my odds of dying? CFR doesn't tell you that, but IFR does. CFR only tells you what percentage of people die after they feel sick enough to go get tested and end up being diagnosed with the disease. I think, based on lots of stuff I've read and heard, that the IFR for Covid-19 might be extremely low (far less than 1%), but the CFR is probably much higher (in the 1-2% ballpark).

 

[urn2bfree]

600 of 4800 living in ridiculously close quarters. 8% infection rate. 7/4800 hospitallized. 1/4800 died. 80% 25 yrs old or younger. Just more numbers to think about.

600/4800 is 12.5%.

 

[ivinsfan]

600/4800 is 12.5%.

Still not that big of a number all things considered, those sailors live in areas that would make most people claustrophobic not to mention communal bathrooms and cafeteria style eating.

Of course it's a big number to the sailors infected and I don't mean to imply otherwise.

 

[aiken]

True, except that you should have written IFR instead of CFR there.

Actually, this raises an interesting point regarding the difference between CFR (case fatality rate) and IFR (infection fatality rate). CFR, the one more widely reported, will always be higher than IFR, because in order to be a "case" the symptoms have to be severe enough to warrant being testing. I think what most people actually want to know is: If I get infected with Covid-19, what are my odds of dying? CFR doesn't tell you that, but IFR does. CFR only tells you what percentage of people die after they feel sick enough to go get tested and end up being diagnosed with the disease. I think, based on lots of stuff I've read and heard, that the IFR for Covid-19 might be extremely low (far less than 1%), but the CFR is probably much higher (in the 1-2% ballpark).

Correct...I meant IFR.

 

[athena53]

I worked with numbers for a living and have seen some cringe-worthy misinterpretations of the data. Ironically, I try not to dwell on numbers knowing all the pitfalls, apples-to-oranges comparisons by country, etc.

But- one thing I'm finding interesting is the percentage of "unresolved" cases. The most current numbers I have for the US show 764,265 reported cases, 40,565 deaths and 71,012 recoveries. So- 5% have died and 9% have recovered. That leaves 86% unresolved. I'm sure there's some lag in reporting, especially deaths since the data would come from death certificates- but what's happening to the rest? Are they dying at home or recovering at home and escaping notice?

The answers could have a huge effect on our assessment of the ultimate death rate. Right now the numerator and denominator are both moving.

 

[Badger]

It's important to consider the long term effects of CV19 and the way it attacks the body. From Medscape article. IMHO, the fatality rate is secondary to long term serious effects of this virus. I think I'd rather pass than live with what is discussed here.



"In particular, Yale New Haven (Conn.) Hospital is dealing with unexpected complications up close. Almost half of the beds there are occupied by COVID-19 patients. Over 100 people are in the ICU, and almost 70 intubated. Of the more than 750 COVID admissions so far, only about 350 have been discharged. "Even in a bad flu season, you never see something like this; it's just unheard of," said Harlan Krumholz, MD, a Yale cardiologist and professor of medicine helping lead the efforts there."


"When they get to the ICU, we are seeing lots of people with acute kidney injuries; lots of people developing endocrine problems; people having blood sugar control issues, coagulation issues, blood clots. We are just waking up to the wide range of ways this virus can affect people. Our ignorance is profound," Dr. Krumholz said, but physicians "recognize that this thing has the capability of attacking almost every single organ system, and it may or may not present with respiratory symptoms."

Because of issues like these I don't really trust numbers for being anything other than a "ballpark" guess. Even the methods of testing for COVID-19 infection is not accurate. And not that many people are being tested other than those who appear to have symptoms or who are in a sensitive category.


Cheers!

 

[Gumby]

I worked with numbers for a living and have seen some cringe-worthy misinterpretations of the data. Ironically, I try not to dwell on numbers knowing all the pitfalls, apples-to-oranges comparisons by country, etc.

But- one thing I'm finding interesting is the percentage of "unresolved" cases. The most current numbers I have for the US show 764,265 reported cases, 40,565 deaths and 71,012 recoveries. So- 5% have died and 9% have recovered. That leaves 86% unresolved. I'm sure there's some lag in reporting, especially deaths since the data would come from death certificates- but what's happening to the rest? Are they dying at home or recovering at home and escaping notice?

The answers could have a huge effect on our assessment of the ultimate death rate. Right now the numerator and denominator are both moving.

Both NY and CT, and probably other states that I don't follow, have recently "batch added" some deaths that were subsequently determined to be COVID-19 related, so that number will change occasionally, but not in drastic fashion. I think the likely disparity in the recovered number is due to the people who are tested positive but not admitted to the hospital. There are simply not enough resources to follow up on them. I would think the vast majority will be recovered, because if they were likely to die, they would probably have been admitted to the hospital in the first instance.