Vanadium 50 said:The problem with that explanation is that if you make the fatality rate lower, you make the NYC incidence higher and the mismatch between antibody tests and inferred incidence rates gets even more discrepant.
Vanadium 50 said:Let me propose two other possibilities. I don't think either is right, but I don't think we have evidence against them.
- There are two (or more) strains, both equally contagious and equally dangerous, but only one shows up on the antibody test.
- Antibodies only persist for 6-8 weeks post infection. People can get reinfected (and we do have some examples of that). The decrease in positivity we are seeing in NYC is not driven by testing a healthier population, but is a real change over time.
CDC's number being too low is more plausible than New York having more sick people than people. And yes, as discussed before, we were studying how the numbers work out if the CDC number is too low. All the US numbers fit very nicely if we assume a higher IFR.AstraZeneca, Under Fire for Vaccine Safety, Releases Trial BlueprintsVanadium 50 said:That's an argument that the CDC 0.26% number is wrong. That's a position that's defensible, but should be attacked on it's merits.
Vanadium 50 said:I can't do arithmetic. Fair enough. Show me where.
I don't think the evidence supports that. The media played-up busy hospitals and harried staff, but NYC added substantial emergency capacity, which went almost completely unused.atyy said:...but given that the NYC health system was overwhelmed in the early stages...
Your links don't support that hospitals were not overwhelmed. The extra personnel were used, and some extra capacity was used.russ_watters said:I don't think the evidence supports that. The media played-up busy hospitals and harried staff, but NYC added substantial emergency capacity, which went almost completely unused.
https://www.navytimes.com/news/your...s-nyc-having-treated-fewer-than-200-patients/
https://www.google.com/amp/s/abc7ny.com/amp/coronavirus-nyc-update-corona-virus-cases/6142109/
https://www.militarytimes.com/news/...orkers-are-going-straight-into-nyc-hospitals/
Well not for nothing, but your first link explicitly states that the hospitals were not overwhelmed.atyy said:Your links don't support that hospitals were not overwhelmed. The extra personnel were used, and some extra capacity was used.
It's fair to say resources (in particular, personnel) were stretched, a small number of hospitals filled to capacity and that there were specific shortages of materials and equipment, but "overwhelmed" has a much worse connotation than that -- a connotation that is necessary to support the argument that a substantially higher death rate occurred because of hospital capacity issues. E.G., you cited a 5x higher IFR in Hubei as an example. You can't get a 5x higher IFR due to "overwhelmed" hospitals without at least a system-wide 80% shortfall in critical needs. Sharing ventilators isn't enough: you'd need to be rationing them to every 5th person who needs them. Or have a smaller overage that causes a systemic collapse. But NYC was nowhere close to full capacity overall.“We have 53,000 hospital beds available,” Cuomo, a Democrat, said at a briefing on March 22. “Right now, the curve suggests we could need 110,000 hospital beds...”
...But when New York hit its peak in early April, fewer than 19,000 people were hospitalized with COVID-19.
russ_watters said:It's fair to say resources (in particular, personnel) were stretched, a small number of hospitals filled to capacity and that there were specific shortages of materials and equipment, but "overwhelmed" has a much worse connotation than that -- a connotation that is necessary to support the argument that a substantially higher death rate occurred because of hospital capacity issues. E.G., you cited a 5x higher IFR in Hubei as an example. You can't get a 5x higher IFR due to "overwhelmed" hospitals without at least a system-wide 80% shortfall in critical needs. Sharing ventilators isn't enough: you'd need to be rationing them to every 5th person who needs them. Or have a smaller overage that causes a systemic collapse. But NYC was nowhere close to full capacity overall.
And the article provides a specific reason that the projections were wrong -- it's the issue we're discussing: the disease is/was nowhere near as severe as the early projections. From the article, it says the CDC initially estimated 11 hospitalizations would be needed per death, then later dropped it to 7, then later to 4. This is almost certainly the same as the CFR issue; the disease is less severe than initially thought because the testing shortage meant we were missing most of the people infected. This almost certainly affected Hubei as well. There may also have been political approach factors at play (e.g., if they quarantined every infected person at field hospitals instead of letting them quarantine at home), but I'm not sure of the details of that.
The CDC initially estimated a 9% fatality rate for hospitalized cases then updated that to 14%, then to 25%, and that's less severe than the 9%?russ_watters said:the disease is/was nowhere near as severe as the early projections. From the article, it says the CDC initially estimated 11 hospitalizations would be needed per death, then later dropped it to 7, then later to 4.
I suppose if you ignore the dropping per case death rate that could be confusing. The entire line has stretched, and the death count is the anchor: we were missing most of the less intense cases. As the pandemic has progressed and the testing rate has increased we have seen fewer hospitalizations and many fewer deaths per case than was initially expected. I really don't understand why people seem to be pretending that the case rates in March/April were accurate and misconstruing the resulting shift in the statistics. Are you expecting an explosion of deaths in Germany in the next few weeks due to the increasing case rate there? Germany peaked around 5,000 cases per day and 200 deaths per day (average over about a week - guestimated) in the spring. Now in the second peak it is seeing around 2,000 cases and 10 deaths per day. Are you expecting the death rate to increase by a factor of 4 in the near future? I don't; I expect that like most other countries, the early severity of the virus (in terms of hospitalizations and deaths) was overestimated due to the low testing rates for people with milder symptoms.mfb said:The CDC initially estimated a 9% fatality rate for hospitalized cases then updated that to 14%, then to 25%, and that's less severe than the 9%?
IMO, it shouldn't be controversial. When the testing rate is known to be extremely low, it shouldn't be controversial that it plays a bigger role in the statistics than a "stretched" medical system.atyy said:Yes, "stretched" is a better word. And definitely it is a matter of conjecture, whether stretched hospital care contributes to explaining why the IFR in NYC could have been higher than 0.3% in the early phases of the outbreak.
That does not fit well with the result of some studies from spring where whole societies (!) were tested. Those studies were the ones which established the whole 'half the infected are without symptoms' thing.russ_watters said:...the disease is less severe than initially thought because the testing shortage meant we were missing most of the people infected.
russ_watters said:IMO, it shouldn't be controversial. When the testing rate is known to be extremely low, it shouldn't be controversial that it plays a bigger role in the statistics than a "stretched" medical system.
I really find this truly bizarre that you (not the specific "you", but the general) are trying to stretch a few percent here or there into hundreds of percent. We don't know if single digit, dozens or hundreds of people overall died because they shared ventilators, but we do know that millions of people who should have been tested in New York, alone, were not. Over the past month, the US media anyway has been breathlessly reporting the record daily case rates, without ever reporting (that I have seen) that those case rates are offset by a factor of 5 or 10 from those in March/April due to the higher testing rate.
Right, that's my point: we didn't have a 100% ventilator shortage. Ventilator sharing happened, but it wasn't common, much less universal.atyy said:Let's say 4 people in 1000 die if everyone that dies needs a ventilator and has to share (ie. sharing ventilators is ineffective), and 2 people in 1000 die if everyone who needs a ventilator gets their own (ventilators save 50% of those who need one). Then one would get a 100% increase in fatality rate (from 2/1000 to 4/1000), even taking into account that many do not get tested.
And the CDC initially estimated the fatality rate for hospital cases was 9% then 14 % then 25%. So there are two tracks here. The hospitalizations / fatality has dropped with a corresponding increase in fatality rate/hospitalizations. And mutually inclusive the disease is less severe because the testing shortage was missing large numbers of mild/asymptomatic people.russ_watters said:From the article, it says the CDC initially estimated 11 hospitalizations would be needed per death, then later dropped it to 7, then later to 4. This is almost certainly the same as the CFR issue; the disease is less severe than initially thought because the testing shortage meant we were missing most of the people infected.
Ok, yeah, I see how the data/model can be interpreted from either direction if we don't have the context/method. FYI, I didn't make this interpretation up myself, I read it in the news/commentary article. But now I checked the original source to make sure that the direction the article specified is correct, and yes, it is correct: fewer hospitalizations per death means fewer hospitalizations, not more deaths. Deaths is the anchor, hospitalizations the variable, so when the fraction decreases its because hospitalizations decreased, not because deaths increased:mfb said:@Russ: No one questions that many cases were missed in March and April. That's not what the discussion was about. Cases that go to a hospital are severe cases, they are not missed. If these people die more often than estimated before, how is that change in particular making the disease less dangerous? You cited that change as evidence that the danger was overestimated.
http://www.healthdata.org/sites/default/files/files/Projects/COVID/Estimation_update_040520_3.pdfCDC said:[page 5]
Why did these changes occur?
These changes in predicted hospital resource use and related gaps in states where demand might exceed supply are fairly large. In this section, we explore the changes in our analytical framework that resulted in these revised estimates of overall lower hospital resource use due to COVID-19.
...our overall ratio was 11.1 hospital admissions per COVID-19 death.
...Our estimates released today... 7.1 hospitalizations per death (95% CI 4.0 to 12.7). These lower ratios of admissions to deaths result in predicted peak hospital resource use – total beds, ICU beds, and invasive ventilators – that is lower than previously estimated.
[emphasis added]
Since you already know it that it is in fact false, I don't understand why you are even offering it as a possibility. Please stop trying to use this inaccurate claim as a basis for your interpretation of the data, and instead follow the data where it actually leads. Again, I really don't understand why this should be controversial.If you argue people got less likely to be admitted to a hospital than before - while the disease stays unchanged - that would mean hospitals had to turn down increasingly severe cases. That would mean they are overwhelmed. I don't say that's true, but that's one of the few ways to interpret these numbers without saying it kills more people than expected before.
[emphasis added]
russ_watters said:
Yes.atyy said:For New York, by deaths per day (p10), the CDCs later estimates (April 2 & 5) seem to have been more severe than initially estimated (March 26).
russ_watters said:Yes.
Yes, apologies, you are correct: the first number (11.1) was direct from CDC data whereas the second (7.1) is from the expanded study/data set from IHME. The news article lists a third ("about 4") that is also presumably from IHME, but it doesn't have a link to it. It also mentions a Harvard model that also estimated too high.atyy said:Also, I guess these are IHME's estimates, not the CDC's?
russ_watters said:Yes, apologies, you are correct: the first number (11.1) was direct from CDC data whereas the second (7.1) is from the expanded study/data set from IHME. The news article lists a third ("about 4") that is also presumably from IHME, but it doesn't have a link to it. It also mentions a Harvard model that also estimated too high.
mfb said:0.10% of the population died in Manaus. Some parts of NYC had more deaths. It's possible that the official number underestimates the total deaths of course.
"I don't say that's true" doesn't mean its false. I listed it as one possible interpretation of the data.russ_watters said:Since you already know it that it is in fact false, I don't understand why you are even offering it as a possibility. Please stop trying to use this inaccurate claim as a basis for your interpretation of the data, and instead follow the data where it actually leads. Again, I really don't understand why this should be controversial.
I don't see how you can rule this out, because hospitalized -> death did get steeper based on CDC/IHME estimates. Your own source demonstrates that #1 is the case for at least half of that chain.russ_watters said:#1 was covered in the first section; not the case.
No, but I'm not sure what the conclusion of your post is.Vanadium 50 said:Ever been to Manaus?
A New York Times survey of more than 1,600 American colleges and universities — including every four-year public institution and every private college that competes in N.C.A.A. sports — has revealed at least 130,000 cases and at least 70 deaths since the pandemic began.
*I* am saying it's false and I'm saying that you should understand/accept by now that it is false because we just discussed it in some detail. It's fine to follow a chain of logic to find "possible" interpretations, but then you have to start analyzing them and checking against other facts to verify if they are in fact true, false or actually unknown/possible. This one has been determined to be false, so it should be taken off the list, not continued to be listed as "possible".mfb said:"I don't say that's true" doesn't mean its false. I listed it as one possible interpretation of the data.
I don't see how you can rule this out, because hospitalized -> death did get steeper based on CDC/IHME estimates. Your own source demonstrates that #1 is the case for at least half of that chain.
I agree that starting from deaths does feel weird, but as you pointed out previously, deaths are a solidly known number. So that make them a good, perhaps the best anchor. It's best to rely on the facts we have for an anchor, agreed? That's why, somewhat separately, we are starting from death #s and projecting infected #s. The death numbers are known to be more reliable.I don't see how deaths could be any reasonable anchor. Infections would be the best, [snip] But starting from deaths is weird.
Hospitalizations are the output of the model, so they can't be an input.in the absence of reliable infection numbers we can use hospitalizations (limiting the analysis to cases that are not mild).
A disease that puts 1% into the hospital and kills 0.001% is worse than a disease that puts 0.1% into the hospital and kills 0.001%. Same number dead, more hospitalized is "worse". Or even worse, if having 1% in the hospital increases the death rate to 0.002% due to hospitals being "overwhelmed". That's the claim/prediction, I entered to counter. Here it is again:Going by that definition a disease that puts 1% into a hospital but only kills 0.001% must be the worst disease ever? 1000 hospitalizations per death! In addition hospitalizations per death go up if treatment in a hospital gets better (i.e. the hospital gets better in preventing deaths) - which certainly means the disease gets less dangerous, not more dangerous.
We should all agree/accept by now that the "given" premise is false, not "possible" and therefore it could not have driven the conclusion to be true.atyy said:Overall an IFR of 1% may be a bit high, but given that the NYC health system was overwhelmed in the early stages, it seems plausible that IFR in the early stages of the NYC outbreak was higher
Hospitalizations are known as well. Maybe even better, because they don't include deaths for unknown reasons outside hospitals. Being solidly known alone doesn't make something a good anchor.russ_watters said:I agree that starting from deaths does feel weird, but as you pointed out previously, deaths are a solidly known number.
Past hospitalizations and deaths are known and can be inputs (if it's an output, then you better check it's correct), future hospitalizations and deaths (the reason you make a model) can be an output but not an input.russ_watters said:Hospitalizations are the output of the model, so they can't be an input.
I was commenting on a very specific claim - that more deaths per hospitalization would mean the disease is less severe. I said this is not the case. I made a small side remark how you could get more deaths per hospitalization without a more severe disease (overcrowded hospitals), and I said this was largely a theoretical option without much relevance. I think you missed that part, because you keep going back to that side remark as if it would have been something important. It wasn't. Forget it.russ_watters said:I feel like you may have lost track of what we were discussing.
But how realistic is this comparison? hospital->death ratio depends on how severe the disease is in cases that we can count easily. Why would cases that we can't study easily behave in exactly the opposite way? A disease that puts more people into a hospital than a comparable disease will almost certainly kill more people, too.----russ_watters said:A disease that puts 1% into the hospital and kills 0.001% is worse than a disease that puts 0.1% into the hospital and kills 0.001%. Same number dead, more hospitalized is "worse".
russ_watters said:We should all agree/accept by now that the "given" premise is false, not "possible" and therefore it could not have driven the conclusion to be true.
[edit] ...which doesn't mean the conclusion about the IFR couldn't be true for other reasons. It may well be true due to demographics (old people were practically targeted for infection).
Astronuc said:Abbott rapid test
US President Donald Trump is on his way to military hospital Walter Reed Medical Centre, White House Press Secretary reports.
"The President will be working from the presidential offices at Walter Reed for the next few days," White House Press Secretary Kayleigh McEnany said in a statement on Saturday.
Trump has had a fever since Saturday morning, according to CNN.
McEnany said this was "out of an abundance of caution, and at the recommendation of his physician and medical experts."