Issues with the IHME model.
-
From a blog post by Greg Cochran, an outline of issues with the model:
First, they accept CCP data. I am not sure what actually happened in Wuhan, but I am sure that A. things were at least as bad as admitted and B. the CCP (eventually) reacted with overwhelming energy, in a way that seems to a fit a much bigger disaster than admitted.
AW thinks that deaths are significantly underestimated in several countries (certainly Italy, Spain, and France) and that IHME errs by not considering this. Correct.
IHME model assumes that state that do not yet have infection measures will do so with a week. That would be nice.
They assume that once controls are imposed, the epidemic will shrink just as fast as it previously grew. If the original R0 is 3, the post-control R0 must be 1/3rd. Uh, why?
“They assume that state-wide infection controls are equivalent in effect with the controls the CCP imposed in Wuhan.” AW calls this assumption crazy, but surely a stronger word is called for. In their first try, the CCP tried these measures, which are more intense than any yet in the US:
-Blocking outward transportation from Wuhan
-Closing public transit and vehicular traffic inside Wuhan
-Compulsory mask-wearing in public places
-Cancellation of public events
-Self-quarantine of confirmed or suspected casesThat dropped the R0 from around 3.5 to 1.25 – which wasn’t enough. They moved on to
-Full quarantine of confirmed or suspected cases (i.e., extraction to a separate quarantine site), including contacts of confirmed cases
-Temperature monitoring of all residents
-Universal and strict stay-at-home orders for all residentsAll of these errors are in the same direction.
-
IHME generally lowered their projections across the board. Feels like Bill Gates got an early view. NY is down a lot in particular.
Comparing some other screenshots I took as we speak.
Their model was down for 5 days, I thought there must have been a big flaw so I was surprised that the revisions went in the direction they did.
-
THey actually moved NY's peak earlier by a day.
They now say RESOURCE use will peak Wednesday.
Seems like we already know for sure that isn't true, since it must lag peak case count by a considerable amount of time - weeks perhaps.
-
Active cases in Italy are still growing by 3k per day. That's (new cases - recoveries).
But NY resource utilization will peak in 48 hours?
-
Yet look at NY vs Italy cases on a log scale. We're going to peak in 2 days and they haven't peaked yet?
-
8 weeks ago I predicted that at the end of this, those that were very concerned about this would be saying I told you so. I also predicted that those that said this isn’t that bad and we’re overreacting would be able to say the same. I stand by those predictions.
-
https://www.sydney.edu.au/data-science/
n international group of statisticians from CTDS, Northwestern University and the University of Texas have released a paper (pdf, 2.5MB) investigating the predictive performance of the model developed by Institute for Health Metrics and Evaluation (IHME). The IHME model is is used to predict ventilator use, hospital bed requirements and other resourcing for US states response to COVID-19.
The key findings are:
-
Over 70 percent of US states had death rates that were inconsistent with IHME predictions
-
Ability of IHME model to make accurate predictions decreases with increasing amount of data
-
Improved predictive modelling needed for adequate provision of ventilators, PPE, medical staff at a local level
-
-
A widely followed model for projecting Covid-19 deaths in the U.S. is producing results that have been bouncing up and down like an unpredictable fever, and now epidemiologists are criticizing it as flawed and misleading for both the public and policy makers. In particular, they warn against relying on it as the basis for government decision-making, including on “re-opening America.”
“It’s not a model that most of us in the infectious disease epidemiology field think is well suited” to projecting Covid-19 deaths, epidemiologist Marc Lipsitch of the Harvard T.H. Chan School of Public Health told reporters this week, referring to projections by the Institute for Health Metrics and Evaluation at the University of Washington.
Others experts, including some colleagues of the model-makers, are even harsher. “That the IHME model keeps changing is evidence of its lack of reliability as a predictive tool,” said epidemiologist Ruth Etzioni of the Fred Hutchinson Cancer Center, home to several of the researchers who created the model, and who has served on a search committee for IHME. “That it is being used for policy decisions and its results interpreted wrongly is a travesty unfolding before our eyes.”
The IHME projections were used by the Trump administration in developing national guidelines to mitigate the outbreak. Now, they are reportedly influencing White House thinking on how and when to “re-open” the country, as President Trump announced a blueprint for on Thursday.
The chief reason the IHME projections worry some experts, Etzioni said, is that “the fact that they overshot” — initially projecting up to 240,000 U.S. deaths, compared with fewer than 70,000 now — “will be used to suggest that the government response prevented an even greater catastrophe, when in fact the predictions were shaky in the first place.”
That could produce misplaced confidence in the effectiveness of the social distancing policies, which in turn could produce complacency about what might be needed to keep the epidemic from blowing up again.
Believing, for instance, that measures well short of what China imposed in and around Wuhan prevented a four-fold higher death toll could be disastrous.
The most notable bounces in the IHME projections have been for the eventual total of U.S. deaths by early August, which is when many epidemiologists believe the outbreak will be tailing off. (Many expect daily deaths in the U.S. to fall to 10 or fewer by early June, from 2,000 or so in April.) Death projections for individual states have also fluctuated significantly.
More interesting stuff later in the article.
-
Please, it's not like the model was so bad that college dropout piano pimps were pointing out flaws... Oh, wait..
-
@George-K said in Issues with the IHME model.:
@Loki said in Issues with the IHME model.:
IHME has been updated.
Florida should not be ignored just because it is inconvenient politically. Who cares if the GOP was “asleep the wheel” and didn’t apply the social distancing as coronavirus as spreading rapidly.There is real information in a different result that we can learn from.
-
I think we’ll only see that low a number if there’s significant seasonality in transmission.