Started By
Message
The IHME Model is Wrong
Posted on 4/16/20 at 4:20 pm
Posted on 4/16/20 at 4:20 pm
This model projected the US would end up with 82,142 total COVID-19 fatalities by August.... that projection was made back on 3/30.
It has since updated twice.... first reducing down to around 60,000 projected fatalities by August, and then slightly raising that figure up to 68,841 (last updated on 4/13).
I think this latest update is projecting far too low.
On 4/13, it projected that New York State would have 11,909 fatalities by today (4/16). It has 12,192... slightly higher.
On 4/13, it projected that New Jersey would have 3,285 fatalities by today (4/16). It now has 3,518... about 7% higher than expected
On 4/13, it projected that Michigan would have 1,887 fatalities by today (4/16). It now has 2,093... that's 11% higher than what they expected just three days ago.
On 4/13, it projected that Louisiana would have just 978 fatalities by today (4/16). It now has 1,156... that's 18% higher than expected. In fact, it was so far off on Louisiana that it projected that state to end up with only 1,141 total fatalities by August. Just three days later and its already blown past the mark that this vaunted study projected them to have 4 months from now.
----
Take the nation as a whole. This highly regarded study projected that as a country, we would hit peak back on April 10th.... and then we'd have slowly falling fatalities each day following that peak day. Here is how many it projected for the last 3 days and how many we actually had:
4/14: Projected 1,953 deaths; we actually had 2,407
4/15: Projected 1,928 deaths; we actually had 2,482
4/16: Projected 1,888 deaths; we have 2,026 so far
It projected that as a whole, after today we'd have 29,902 total fatalities. We now sit at 34,469.... that's 15.3% higher than expected and this latest updated was likely calculated on Sunday. That's a pretty bad miss for only be four days out, and worse yet they missed on the low side.
I don't know when they'll update their model again, but my guess is that it again bumps up the overall projections.... this time substantially.
I think they expected these hardest hit states that were further along in their cycle (NY, NJ, MI, LA) to all start showing very positive signs and once they hit peak, we'd see a noticeable but slow path back down to zero.... but instead, they are peaking and this continuing to stay at those peak numbers for much longer than expected.
Take NY for example. Here are how many daily deaths they have had in recent days:
4/8: 779
4/9: 799
4/10: 777
4/11: 783
4/12: 758
4/13: 671
4/14: 778
4/15: 752
4/16: 606
The good news is that NY has certainly peaked and is no longer rising. The bad news is we're not really seeing the numbers improve. Today's number showed some improvement but we've yet to see 2 measurably improving days in a row.
If every state follows this track.... hits their peak and then stays up around that peak for a week to 10 days prior to starting to measurably fall, then the projected fatality rates will be much, much higher than currently projected.
Unfortunately that means a much slower time to get back to normal.
It has since updated twice.... first reducing down to around 60,000 projected fatalities by August, and then slightly raising that figure up to 68,841 (last updated on 4/13).
I think this latest update is projecting far too low.
On 4/13, it projected that New York State would have 11,909 fatalities by today (4/16). It has 12,192... slightly higher.
On 4/13, it projected that New Jersey would have 3,285 fatalities by today (4/16). It now has 3,518... about 7% higher than expected
On 4/13, it projected that Michigan would have 1,887 fatalities by today (4/16). It now has 2,093... that's 11% higher than what they expected just three days ago.
On 4/13, it projected that Louisiana would have just 978 fatalities by today (4/16). It now has 1,156... that's 18% higher than expected. In fact, it was so far off on Louisiana that it projected that state to end up with only 1,141 total fatalities by August. Just three days later and its already blown past the mark that this vaunted study projected them to have 4 months from now.
----
Take the nation as a whole. This highly regarded study projected that as a country, we would hit peak back on April 10th.... and then we'd have slowly falling fatalities each day following that peak day. Here is how many it projected for the last 3 days and how many we actually had:
4/14: Projected 1,953 deaths; we actually had 2,407
4/15: Projected 1,928 deaths; we actually had 2,482
4/16: Projected 1,888 deaths; we have 2,026 so far
It projected that as a whole, after today we'd have 29,902 total fatalities. We now sit at 34,469.... that's 15.3% higher than expected and this latest updated was likely calculated on Sunday. That's a pretty bad miss for only be four days out, and worse yet they missed on the low side.
I don't know when they'll update their model again, but my guess is that it again bumps up the overall projections.... this time substantially.
I think they expected these hardest hit states that were further along in their cycle (NY, NJ, MI, LA) to all start showing very positive signs and once they hit peak, we'd see a noticeable but slow path back down to zero.... but instead, they are peaking and this continuing to stay at those peak numbers for much longer than expected.
Take NY for example. Here are how many daily deaths they have had in recent days:
4/8: 779
4/9: 799
4/10: 777
4/11: 783
4/12: 758
4/13: 671
4/14: 778
4/15: 752
4/16: 606
The good news is that NY has certainly peaked and is no longer rising. The bad news is we're not really seeing the numbers improve. Today's number showed some improvement but we've yet to see 2 measurably improving days in a row.
If every state follows this track.... hits their peak and then stays up around that peak for a week to 10 days prior to starting to measurably fall, then the projected fatality rates will be much, much higher than currently projected.
Unfortunately that means a much slower time to get back to normal.
8
Posted on 4/16/20 at 7:16 pm to BHMKyle
( Overhead projector )
Posted on 4/16/20 at 7:48 pm to BHMKyle
quote:
that figure up to 68,841 (last updated on 4/13).
Well some would say influenza A/B kills appx 60k Americans every year. I guess the fallacy is that it took shutting down 98% of non-essential businesses for 2+ months to keep it that low. Otherwise we might have had the 1-2M dead & still have a significant business disruption. Flattening the curve creates as much danger for round 2 preparation or
Covid-2X preparation. Just like sitting on one’s porch during a hurricane just because the last one was just a passing shower.
Posted on 4/16/20 at 10:10 pm to BHMKyle
Tennessee is passed the peak and in recovery despite the model saying we have another week. The model over estimated deaths by 50% and cases beds/ICU beds/ventilators by a similar percentage. If you scale down the model to fit reality we are expected to be free and clear in under 10 days. 0 beds needed. But the governor just canceled the rest of the school year. He does plan a controlled restarting of the state beginning 1 May.
My county is down to 3 active cases out of a population of 125,000. 39 recovered, 3 dead, 3 active (not hospitalized). They started offering free open testing yesterday so I'm sure the number of positive and recovered will jump in 10 day but the more we open schroders box the more we see the cat is alive and well. The next county over is knox, population of 477,000 has about 50 active cases.
Population density and cultural/ethnic conditions were just not favorable to an upper respiratory viral infection spreading here. And are likely why the models were so drastically wrong here.
My county is down to 3 active cases out of a population of 125,000. 39 recovered, 3 dead, 3 active (not hospitalized). They started offering free open testing yesterday so I'm sure the number of positive and recovered will jump in 10 day but the more we open schroders box the more we see the cat is alive and well. The next county over is knox, population of 477,000 has about 50 active cases.
Population density and cultural/ethnic conditions were just not favorable to an upper respiratory viral infection spreading here. And are likely why the models were so drastically wrong here.
Posted on 4/17/20 at 2:22 pm to NoMansLand
quote:
Well some would say influenza A/B kills appx 60k Americans every year.
The flu argument is preposterous. Folks who are constantly arguing about this using the flu as their example are trying to take annualized numbers and compared them to these coronavirus numbers that are only a few weeks old.
There have been 18,000+ Americans die of COVID-19 in just the past week. If that rate kept up for 52 weeks, it would equal around 940,000 annual deaths (that's 15x more than the 60,000 flu deaths per year).
Posted on 4/17/20 at 2:34 pm to BHMKyle
What are the flu deaths during that time?
Posted on 4/17/20 at 2:50 pm to Smokeyone
quote:
Tennessee is passed the peak and in recovery despite the model saying we have another week.
Looking at the model for Tennessee, it currently says that projected daily deaths peaked on 4/13.... In actuality, Tennessee looks like its daily deaths peaked on back on 4/9 (looked at on a 7-day rolling average), so four days earlier.
It projected that on 4/16, Tennessee would have 179 cumulative deaths when in actuality it had just 141 (overestimating by 27%).
There are definitely some states that the model overestimates, but one of the problems I see with this model is that I believe it is projecting far too fast of a recovery.
Using Tennessee as an example, it projects deaths/day to peak on 4/13 and then it has deaths decreasing down to zero on 4/18.... that's exactly 5 weeks to go from peak to zero.
For New York, it projects the state to go from peak (4/10) down to zero deaths per day by 5/9 (just 29 days). I hope these estimates are true, but I just don't see that happening based on other countries such as Italy that are further along in their cycle.
Take the Lombardia region of Italy for example. They peaked back on 3/31.... 7-day fatality rate was averaging 446 per day. Seventeen days later and they are still averaging 223 fatalities per day (7-day average) and actually had 243 just today.
Take that back to New York. It projected NY would peak on 4/10 with 800 daily deaths.... 17 days later it projects that NY will have just 56 daily deaths on 4/27. That's just 10 days from now and NY is still experiencing 600+ deaths every single day. I just see no way that they can go from 630 today to just 56 in ten days. I'd love it if it were true, but it seems highly unlikely.
I believe the IHME Model is underestimating the amount of time it takes to go from peak down to zero. Seeing that there is no region of the world that was hard hit and has since fully recovered, this is a difficult thing to project..... I wonder if they are relying at all on Chinese data out of Hubei province since Hubei has been falsely claiming zero or very few deaths for weeks now.
Posted on 4/17/20 at 3:03 pm to kywildcatfanone
You tell us. I have no idea of weekly deaths from either strain. The only thing I can see is exponential cases where I live. It’s not that way everywhere & very understandable why certain states either did not shelter in place or are easing things up now. The balance of safety against economics is needed but should be decided at the micro level. No heroes need apply to manage this for everyone.
This post was edited on 4/17/20 at 3:06 pm
Posted on 4/17/20 at 5:08 pm to BHMKyle
quote:
I believe the IHME Model is underestimating the amount of time it takes to go from peak down to zero.
I also fear it is not taking into account fresh outbreaks. South Dakota and Iowa were pretty free from cases until recently but now are seeing hot spots appear.
Saying a state or region will hit its peak infection rate and go down from there assumes the best case scenario and doesn't represent the start and stop nature of a modern pandemic.
Posted on 4/17/20 at 5:40 pm to Arksulli
Except that most states have peaked and declined. Dont let mini hot spots confuse you. It was the same fallacy that panicked Italy. 1 region in italy was a hot spot while the rest of the country was coping relatively well.
And the mortality rate for the US is a joke. Their is no standards for what is or is not a corona death. It changes from jurisdiction to jurisdiction.
And the mortality rate for the US is a joke. Their is no standards for what is or is not a corona death. It changes from jurisdiction to jurisdiction.
Posted on 4/17/20 at 6:14 pm to BHMKyle
Just revised their projection on Florida.
Initially thought they’d peak in May. Now say they’ve already peaked. Lowered death numbers
Initially thought they’d peak in May. Now say they’ve already peaked. Lowered death numbers
This post was edited on 4/17/20 at 6:17 pm
Posted on 4/19/20 at 2:23 am to Smokeyone
quote:
Their is no standards for what is or is not a corona death.
Tell us more about your epidemiology studies
Posted on 4/19/20 at 5:04 am to BHMKyle
Time will tell on their power minds,
Making war just for fun
Treating people just like pawns in chess,
Wait 'till their judgement day comes, yeah
Making war just for fun
Treating people just like pawns in chess,
Wait 'till their judgement day comes, yeah
Posted on 4/22/20 at 2:33 pm to BHMKyle
Getting back to this....
The IHME Model updated again this week to show an increase in projected US fatalities.... this time back up 65,976. I knew it would have to revise up, but I still don't think it went up far enough. I still think this is very much under-estimating the death toll.
If you look at average daily fatalities on a 7-day rolling average, the US hit its highest total yet yesterday: 2,213 daily fatalities on average during the previous week.
We're not even positive that this is the peak for the US, but let's just assume the best case scenario that yesterday was indeed the peak.
We can look to other countries who peaked before us to see how they performed one week and two weeks out after their peak.
Italy (-23.4%), Spain (-19.4%), France (-7.9%), the Netherlands (-17.6%), and the UK (-13.0%) all saw their fatalities fall one week out from hitting their 7-day rolling average peak. The amounts are fairly similar and average out to a decrease of 16.3%.
Now if we apply that to the US' current numbers, then we should expect the next 7 days to average 1,851 fatalities per day.
If you look out two weeks past the peak, Italy (-31.7%), Spain (-43.9%), France (-38.9%), and the Netherlands (-14.4%) are now all at least 2 weeks past their peak, and they also saw similar decreases in the second week following their peak..... averaging out to a 32.3% decrease in fatalities during the second week out from peaking.
Apply this to the US, and we can expect that two weeks from today, we should have seen the previous week average 1,500 fatalities per day.
Those numbers are all we really have to go by. If the US follows a similar pattern as those countries did, we should expect to see a total of 23,457 more fatalities during the next two weeks.
If you add that to our current total (45,313 deaths as of yesterday), then that means two weeks from now, total fatalities should sit around 68,770.
.... that's more than 3,000 fatalities than the IHME model is projecting us to have in August. And yet if we follow those other countries' patterns, we'll still be seeing 1,000+ fatalities per day two weeks from now.
Once again I think this model is way off. Unfortunately I think we're looking at at least 80,000 US fatalities by August. I hope they are correct and I'm wrong, but I don't see any reasoning for their low projections.
The IHME Model updated again this week to show an increase in projected US fatalities.... this time back up 65,976. I knew it would have to revise up, but I still don't think it went up far enough. I still think this is very much under-estimating the death toll.
If you look at average daily fatalities on a 7-day rolling average, the US hit its highest total yet yesterday: 2,213 daily fatalities on average during the previous week.
We're not even positive that this is the peak for the US, but let's just assume the best case scenario that yesterday was indeed the peak.
We can look to other countries who peaked before us to see how they performed one week and two weeks out after their peak.
Italy (-23.4%), Spain (-19.4%), France (-7.9%), the Netherlands (-17.6%), and the UK (-13.0%) all saw their fatalities fall one week out from hitting their 7-day rolling average peak. The amounts are fairly similar and average out to a decrease of 16.3%.
Now if we apply that to the US' current numbers, then we should expect the next 7 days to average 1,851 fatalities per day.
If you look out two weeks past the peak, Italy (-31.7%), Spain (-43.9%), France (-38.9%), and the Netherlands (-14.4%) are now all at least 2 weeks past their peak, and they also saw similar decreases in the second week following their peak..... averaging out to a 32.3% decrease in fatalities during the second week out from peaking.
Apply this to the US, and we can expect that two weeks from today, we should have seen the previous week average 1,500 fatalities per day.
Those numbers are all we really have to go by. If the US follows a similar pattern as those countries did, we should expect to see a total of 23,457 more fatalities during the next two weeks.
If you add that to our current total (45,313 deaths as of yesterday), then that means two weeks from now, total fatalities should sit around 68,770.
.... that's more than 3,000 fatalities than the IHME model is projecting us to have in August. And yet if we follow those other countries' patterns, we'll still be seeing 1,000+ fatalities per day two weeks from now.
Once again I think this model is way off. Unfortunately I think we're looking at at least 80,000 US fatalities by August. I hope they are correct and I'm wrong, but I don't see any reasoning for their low projections.
Posted on 4/22/20 at 5:33 pm to BHMKyle
quote:
Once again I think this model is way off. Unfortunately I think we're looking at at least 80,000 US fatalities by August. I hope they are correct and I'm wrong, but I don't see any reasoning for their low projections.
You have no idea how much I want, for once, your amazing ability to crunch numbers and analyze data to be off.
I don't think you are off however.
Posted on 4/22/20 at 10:15 pm to BHMKyle
I mean I think our definitions of “way off” are different. 15,000 deaths total is way off?
It’s hard to sit here and say you’re right when I don’t know the counter argument. I’m sure the modelers made that simplistic of an analysis (or at least I would hope they would)
I would guess the difference is you’re looking at the country as a total instead of area by area. For example, New York (very populated area) hit their peak 13 days ago according to the model while Florida hit it 10 days ago and Alabama just hit it today. So you’d start to see the -30% to -40% in the biggest populations now and get the small decreases in small populations. I could easily see that accounting for 15,000 deaths
It’s hard to sit here and say you’re right when I don’t know the counter argument. I’m sure the modelers made that simplistic of an analysis (or at least I would hope they would)
I would guess the difference is you’re looking at the country as a total instead of area by area. For example, New York (very populated area) hit their peak 13 days ago according to the model while Florida hit it 10 days ago and Alabama just hit it today. So you’d start to see the -30% to -40% in the biggest populations now and get the small decreases in small populations. I could easily see that accounting for 15,000 deaths
This post was edited on 4/22/20 at 10:44 pm
Posted on 4/24/20 at 11:01 am to BHMKyle
Every death is being reported as Covid so the State can get the Medicare funding ....
Posted on 4/24/20 at 12:21 pm to Weagle25
quote:
I mean I think our definitions of “way off” are different. 15,000 deaths total is way off?
I think 80,000 fatalities is a conservative estimate, and that's 23% higher than the IHME Model projects. Yes, 15,000 is "way off" if its missing it by at least 23%.... likely higher.
Posted on 4/24/20 at 12:30 pm to BHMKyle
(no message)
This post was edited on 4/24/20 at 12:31 pm
Page 1 of 2
Popular
Back to top
Follow SECRant for SEC Football News