Beer Virus V

[SheepDog4Life]

Since the worldometers-derived CFR was calculable over a continuous period of months, it seems that comparing one data point to another calculated the same way would still be relevant unless somehow the length of time between diagnosis and death was lengthening quite markedly. Recall that the steady, downward trend of the calculation was uninterrupted by the 'second spike' which at the time we were told was due to the 'death latency'. Are you suggesting that those increased deaths postulated from the increased cases in August and September are just now showing up in the numbers? That's quite a large latency

Here is a snippet from Worldometer's take on CFR which is what I was talking about:

Coronavirus (COVID-19) Mortality Rate

Once an epidemic has ended, it is calculated with the formula: deaths / cases.

But while an epidemic is still ongoing, as it is the case with the current novel coronavirus outbreak, this formula is, at the very least, "naïve" and can be "misleading if, at the time of analysis, the outcome is unknown for a non negligible proportion of patients." [8]

(Methods for Estimating the Case Fatality Ratio for a Novel, Emerging Infectious Disease - Ghani et al, American Journal of Epidemiology).

In other words, current deaths belong to a total case figure of the past, not to the current case figure in which the outcome (recovery or death) of a proportion (the most recent cases) hasn't yet been determined.

Also, if you're trying to match peak cases to peak fatality, that might be problematic because it's not only a delay in time but also a spreading function (1-6 weeks post symptom) and no idea what the spreading function looks like.

Since you're looking at the summer numbers, where do US cases start rising from flat or shallow down trend? I see that around June 14-15. Do the same for fatalities... I see July 6-7... about 3 weeks.

 

[nonobaddog]

This is what fudge factors are for - or just make up numbers, equally valid.

 

[BugI02]

Here is a snippet from Worldometer's take on CFR which is what I was talking about:

Coronavirus (COVID-19) Mortality Rate

You go to calculation with the numbers you have, not the numbers you might want or wish to have at a later time

Also, if you're trying to match peak cases to peak fatality, that might be problematic because it's not only a delay in time but also a spreading function (1-6 weeks post symptom) and no idea what the spreading function looks like.

Since you're looking at the summer numbers, where do US cases start rising from flat or shallow down trend? I see that around June 14-15. Do the same for fatalities... I see July 6-7... about 3 weeks.

The problem is, you were quite vocal about the coming spike in deaths as cases began to spike in early september and have always indicated that lag was 3 to 4 weeks. The spike in cases, and the subsequently expected spike in deaths expected in early October never affected the declining CFR numbers. You never adequately explained that

Nor is latency of death an adequate explaination for why the trend only now seems broken. What changed?

 

[T.Lex]

Here is a snippet from Worldometer's take on CFR which is what I was talking about:

Coronavirus (COVID-19) Mortality Rate



Also, if you're trying to match peak cases to peak fatality, that might be problematic because it's not only a delay in time but also a spreading function (1-6 weeks post symptom) and no idea what the spreading function looks like.

Importantly, the other necessary number to determine CFR is... "cases." Our testing was ramped up by the summer and into the Fall. The early, high CFR numbers depended greatly on testing, and since we were testing mostly sick people. As testing was more widespread, the CFR declined because we were "narrowing in" so to speak on the "actual" CFR. The more non-life threatening cases we found, the lower the CFR was destined to be.

I still believe the actual CFR will be lower than where it is now, but we'll have to wait and see on that.

One of the things I pointed out with my post that generated the 4-sigma issue was that there isn't an obvious answer to why the CFR stopped declining.

When I look at Indiana, and our relative success, we are clearly an outlier. Or at least on the good end of the spectrum.

Other places, SoCal, NYC, and Dallas come to mind (the last one has major metropolitan centers at 90+% ICU occupancy) as examples where the fatalities may be higher because of overstressed and undersupplied health care locally. That could be one factor that is skewing the results.

Also, with regard to variants, there's been little to no study (that I'm aware of) as to whether they are more deadly.

Lots of factors involved.

 

[SheepDog4Life]

The problem is, you were quite vocal about the coming spike in deaths as cases began to spike in early september and have always indicated that lag was 3 to 4 weeks. The spike in cases, and the subsequently expected spike in deaths expected in early October never affected the declining CFR numbers. You never adequately explained that

Nor is latency of death an adequate explaination for why the trend only now seems broken. What changed?

That was late September into early October, after Indiana went full capacity restaurants/bars phase 5 and I was alarmed by the rising cases... doubling every 3 weeks, IIRC.

YOUR argument was that cases were up, but deaths were not... I said just wait, they are coming... and come they did.

We did the same dance on Sweden's rising cases... and the deaths did indeed follow. Bigly.

If you want to argue exactly by how many days or weeks, I would say you are burying the lede.

And, again, I would say measure based on leading edge, not peak due to the spreading function.

 

[BugI02]

So just back from worldometers. Using the case totals from Nov 1, Dec 1, and Jan 1, as well as using the death totals from one month later to allow for that supposed latency, I get:

Dec 1 Deaths = 276961/9567171 = 2.89%
Nov 1 Cases

Jan 1 Deaths = 356667/14191129 = 2.51%
Dec 1 Cases

Feb 1 Deaths = 454213/20738333 = 2.19%
Jan 1 Cases

All numbers derived from the country specific interactive charts, and again we see a steeply and steadily declining CFR

I would love to hear your rationalization explanation

 

[SheepDog4Life]

So just back from worldometers. Using the case totals from Nov 1, Dec 1, and Jan 1, as well as using the death totals from one month later to allow for that supposed latency, I get:

Dec 1 Deaths = 276961/9567171 = 2.89%
Nov 1 Cases

Jan 1 Deaths = 356667/14191129 = 2.51%
Dec 1 Cases

Feb 1 Deaths = 454213/20738333 = 2.19%
Jan 1 Cases

All numbers derived from the country specific interactive charts, and again we see a steeply and steadily declining CFR

I would love to hear your rationalization explanation

Because it's case fatality rate... which is only as good as the measure of cases... which is only as good as the availability of testing... which for Feb/March/April/May of 2020 was .

For example:
July 1 Deaths = 131,125/1,897,000 = 6.91%
June 1 Cases

and

June 1 Deaths = 109,799/1,154,981 = 9.51%
May 1 Cases

Do those CFRs really mean anything other than there was nowhere near enough testing capability early on to detect any but the most severe cases?

 

[jamil]

Is a change of 0.0003 ever statistically significant?

It depends. 0.0003 what?

 

[nonobaddog]

It depends. 0.0003 what?

It could be very significant in the calculations for a moon shot - you could end up smacking into uranus.

 

[jamil]

It could be very significant in the calculations for a moon shot - you could end up smacking into uranus.

 

[drillsgt]

Because it's case fatality rate... which is only as good as the measure of cases... which is only as good as the availability of testing... which for Feb/March/April/May of 2020 was .

For example:
July 1 Deaths = 131,125/1,897,000 = 6.91%
June 1 Cases

and

June 1 Deaths = 109,799/1,154,981 = 9.51%
May 1 Cases

Do those CFRs really mean anything other than there was nowhere near enough testing capability early on to detect any but the most severe cases?

I'm not sure exactly what you're arguing, you did a good job of showing the calculation of a CFR but i'm not sure what your point is. The CFR is just a good calculation for what is known but it doesn't account for those asymptomatic or those who actually got sick but didn't bother to get tested. Our actual numbers can be expected to be even better than what we can calculate.

 

[JTScribe]

The South African strain eats masks!!

I heard it has diplomatic immunity.

 

[SheepDog4Life]

I'm not sure exactly what you're arguing, you did a good job of showing the calculation of a CFR but i'm not sure what your point is. The CFR is just a good calculation for what is known but it doesn't account for those asymptomatic or those who actually got sick but didn't bother to get tested. Our actual numbers can be expected to be even better than what we can calculate.

I agree except that the CFR calculation is a "good" calculation....

My point is that CFR as calculated isn't very meaningful to begin with... doubly so since tests were so scarce for months at the beginning... and as you indicate, even now with readily available (though not necessarily quick) testing, cases are still an under count of the number infected.

It would take a pretty significant statistical analysis of the case data to even estimate the IFR from all this raw data, which is the only "real" number to indicate how deadly the virus is, and from pretty early on had been estimated to be in the 0.5-1% range.

So, yeah, with more widespread testing, CFR will trend towards IFR even if nothing else changes (which other things have changed).

Plus the whole difference in time of weeks or months between the two terms in the "instant" CFR calculation.

 

[indyblue]

Problem is, just what is a case? So many deaths actually unrelated to covid (.i.e. rooftop lightning strike victims, motorcycle accident victims, elderly folks that would have passed soon anyhow) have been attributed to it.

As they say, follow the money. $3K per case is a lot of incentive to stretch the facts.

 

[printcraft]

It could be very significant in the calculations for a moon shot - you could end up smacking into uranus.

N.I.M.B.Y.

 

[JettaKnight]

N.I.M.B.Y.

More like NIMBD, but that's unpronounceable.

 

[HoughMade]

It depends. 0.0003 what?

I've been trying to convince my wife for years that 0.0003 is significant.

 

[HoughMade]

I heard it has diplomatic immunity.

 

[nonobaddog]

The South African strain eats masks!!

I heard it has diplomatic immunity.

Plus it is gluten free.

 

[BugI02]

Because it's case fatality rate... which is only as good as the measure of cases... which is only as good as the availability of testing... which for Feb/March/April/May of 2020 was .

For example:
June 1 Deaths = 109,799/1,154,981 = 9.51%
May 1 Cases

and

July 1 Deaths = 131,125/1,897,000 = 6.91%
June 1 Cases

Do those CFRs really mean anything other than there was nowhere near enough testing capability early on to detect any but the most severe cases?

After placing your calculations in the propeer temporal order, let's continue:

Aug 1 Deaths = 158426/2817366 = 5.62%
July 1 Cases

Sep 1 Deaths = 189285/4846016 = 3.91%
Aug 1 Cases

Oct 1 Deaths = 212998/6355786 = 3.35%
Sep 1 Cases

Nov 1 Deaths = 236801/7585202 = 3.12%
Oct 1 Cases

At which point we catch up to the figures I've already listed. So, as you see, the best CFR we can calculate at this time has been dropping steadily and rather steeply since early in the 'pandemic'. I am hearing a lot from you about how deaths are delayed so using contemporary figures are wrong, so I have attempted to remove that concern. You have cited a 'death lag' of 3 to 4 weeks, so I spaced data points 1 month apart (30 or 31 days, slightly more than 4 weeks)

I am seeking one explanation for this, but I keep receiving many. 'There isn't enough testing', 'There is too much testing' etc.
What one explaination consistently accounts for the data that we are seeing? It would seem to me, if you think the figures are somehow wrong, then it either has to be an overcount of positive tests or an undercount of deaths (or both). How is that manipulation of the data happening consistently?

Otherwise the numbers must be correct, and the disease likely has never been as deadly as it was hyped to be whether within a spike or not