Gupta on CNN now using IMHE model projections...

[blitzen]

which were just LOWERED from 68,000 to 60,000 total U.S. deaths by Aug. 4. Yet, for several days now, the actual death count in most states (especially severely affected states) has been far higher than the IMHE projections for each particular day. Something doesn't add up. I saw an earlier post about several experts saying that IMHE methodology is flawed.

[The Velveteen Ocelot]

1. The IMHE model seems pretty optimistic when you compare it

to what's really going on.

[greenjar_01]

2. New York state deaths on Worldmeters just jumped from 620 to 1,025

For yesterday.

[Igel]

16. The deaths are for when they're reported.

Not when they happened.

The jump may be that big, smaller, or bigger. Depending on whether they reported all of yesteday's deaths yesterday (unlikely), failed to report some yesterday (likely), and how many of today's deaths they won't report until tomorrow.

[Botany]

3. We had 4,600 just today

And by what reasoning do they get only 60,000 by August 1st? Especially if they
open things back up.

How many total deaths do we have now in America now? 36,000?

https://www.democraticunderground.com/10142472929

[uponit7771]

4. K&R, the model still say go out to end of may. Now Birx is going against the model she said the base

.. based their models off of?

Yeah, something must be really wrong

[coti]

6. Yeah, there are 36.5K total dead and 2K dead just today (so far), so unless

this thing goes away completely within three weeks, after dropping down to less than 1K dead per day for an extended time, I don't think we're staying under 60K dead.

[Celerity]

20. At the rate just based off today (and no increase or decrease)

The US will pass 60,000 deaths in around 10 days. The models are fucked IMHO.

[BigmanPigman]

7. The CDC says to report PROBABLE deaths by Covid now.

Some cities and states are complying but it is very inconsistent. China just did the same thing, counted more probable deaths. All the death numbers will increase, probably 33%.

[kurtcagle]

8. Models

A model is a projection, attempting to determine the future by trying to take a complex set of variables, quantifying them, then fitting them into a dynamic, constantly shifting pattern. We know that there are at least three strains of coronavirus, an early strain that came out of China likely by mid-Fall and settled into the West Coast, a second strain that primarily impacted China itself, and the third (likely a variant of the second) which spread first to Iran than to Europe and eventually the East Coast of the US and Canada, which seems to be the most lethal. This also makes the IMHE model questionable, as these likely have very different profiles (and consequently different variables and constants). Finally, WHO guidelines have now started incorporating probable coronavirus cases rather than confirmed ones, which are not accounted for in the IMHE models. I still think we'll see total deaths in the US closer to 100K rather than 60K, as this settles into the Southeast and Midwest.

[blitzen]

12. Yes, but one should add that the IMHE projections have been proven wrong when actual official...

death numbers are reported by the states. The IMHE graph lets you pick a state, pick a specific date, and see their projection for that date. I've been doing that for the past few days, and the projections are generally way too low.

Example: Michigan, for April 16 IMHE Projection: 75 // Actual reported deaths: 172

This has been going on for days and is similar for the worst hit states. So, since reality is showing their projections to have been wrong, why do they now lower their overall U.S. death projection? There may be a good explanation, but I don't get it.

[Igel]

15. They were high until the last update on 4/13.

Not by much, to be sure.

They lowered them and piddled with the projections a bit.

Having New York change the *basis* of their reports throws any model into chaos. The before and after numbers can't be merged without normalizing one of the two sets, and NY isn't giving the necessary information to do that for current numbers. And as far as I know, it also didn't say how to apportion that large lump sum historically, so you can't adjust the past numbers.

They did this and dragged the CDC along just as some states were peaking. That's the real test of a model--does it predict inflection points. Couldn't have timed screwing over any basis for predictions better if they'd planned it. (But being politicians, it's unlikely the decision makes planned it.)

The last update I can see online is 4/12. They were supposed to recalculate on 4/15, but I haven't seen those numbers.

[VMA131Marine]

9. Given that COVID19 deaths are being undercounted

due to the lack of testing, and particularly post-mortem testing, the US death toll is probably already beyond 60,000. It will take detailed analysis of so-called “excess deaths” to figure out the true death toll of this virus.

[patphil]

11. I agree. It's a whole lot worse than we have been led to believe.

[TreasonousBastard]

10. It probably is, but when did they start the calculations? I didn't take that many statistics...

courses in college, but I do seem to remember that you choose a model, largely by guesswork, and start poking numbers into it. After a while, you see how well it matches reality and you make changes.

Statistical modelling is a black art and eventually works, but should not be taken seriously at first, if at all. FWIW, my education in economics started with the department head in the lecture hall telling us that we economics majors should try to pass the tests, but not believe anything we were taught. Modeling and forecasting, he told us, were complete bullshit.

[Shermann]

13. I didn't see any graphs showing the predictions for the fall yesterday

Fauci just waved his arms and said he thought things would be OK

No models, no graphs, no foundation for this claim.

Did I miss it?

[uponit7771]

22. +1,

[msongs]

14. the model projects probable cases vs hospital beds and ICu NOT total deatht

[SoonerPride]

17. What are you talking about? The model clearly projects death totals

https://covid19.healthdata.org/united-states-of-america

It shows that August 4 the total will be 60,308 dead.

But it is bizarre because it shows the country reaching 59,184 by May 10th and then for the rest of May, all of June, and all of July only have 800 more people die?

In what universe are they talking about?

[blitzen]

18. Scroll down, my friend, scroll down.

https://covid19.healthdata.org/united-states-of-america/michigan

[Roland99]

19. And they moved the FL peak from May 3 to 3 days ago!

Many states had peaks moved way up.

[Squinch]

21. Here are experts talking about why that model is pure crap.

https://www.democraticunderground.com/?com=view_post&forum=1002&pid=13300791

There's a reason Donnie Bodybags chose this model.

Down the road, we will find that a right wing group is a big donor to this model.

[bullwinkle428]

24. Look, Bill Gates is the major funding force behind IHME modeling. We can certainly debate

the predictive or analytical power or lack thereof, using this particular model, but I don't think Gates has any of kind of agenda to make Needy Amin look any better.

SEATTLE, WA, USA (January 25, 2017) - The Bill & Melinda Gates Foundation and the University of Washington’s Institute for Health Metrics and Evaluation (IHME) announced today the foundation’s commitment to invest $279 million in IHME to expand its work over the next decade.

The investment will allow IHME to build on its work providing independent health evidence to improve population health. The award complements other investments from the Gates Foundation to further the work of the University of Washington’s Population Health Initiative, which was launched in May 2016 and is establishing a university-wide, 25-year vision to advance the health and well-being of people around the world.

“IHME provides critical data about global health trends that can empower policymakers worldwide to identify better solutions in the fight against disease,” said Bill Gates, co-chair of the Bill & Melinda Gates Foundation.

+

http://www.healthdata.org/news-release/bill-melinda-gates-foundation-boosts-vital-work-university-washingtons-institute

[Squinch]

25. Look. The model is crap. And I'm guessing Bill Gates isn't its only funder.

And there is a reason why Donnie Bodybags chose it early and quotes it often.

[BGBD]

23. IMHE....

was always a best case scenerio model. I dont know why any responsible official would ever use it to make decisions. Use the median projection or even the worse case to be prepared.

Trump getting on TV and telling people the models were showing 60K and that it was going to be much lower was irresponsible to the level of criminality. That gave a false sense of confidence and likely caused behavior that led to more infections and deaths.

We will be close to 60k by the end of April, if not before.

[ThoughtCriminal]

26. A year from now, we will be looking at overall mortality

And the spike will be obvious and that these counts and estimates were far below actual deaths caused directly and indirectly by the pandemic.

In some ways, it is similar to the number of fatalities during the Iraq war. Counts put the number of deaths at around 109,000, but population-based studies suggested the actual fatalities were actually around 650,000 or more.

[ThoughtCriminal]

27. Today's XKCD

Without widespread testing, guess what the projections are.

[standingtall]

28. At this rate we will be past 60,000 deaths by the middle of next month

if not sooner.