Welp.
The CDC issued a new estimate of the prevalence of Omicron vs. Delta in the United States last week that “has dropped significantly from 73.2% w/ a 95% prediction interval of 34-94.9% down to 22.5% w/ a 95% prediction interval of 15.4-31.5%.”
CDC's estimate for the prevalence of Omicron last week dropped significantly from 73.2% w/ a 95% prediction interval of 34-94.9% down to 22.5% w/ a 95% prediction interval of 15.4-31.5%.
12/25 estimate is 58.6% with a 95% prediction interval of 41.5-74%. https://t.co/eaMCVk3TEc pic.twitter.com/ZjEvREb7Gs
— David Lim (@davidalim) December 28, 2021
This is a MAJOR revision:
That’s a pretty massive adjustment https://t.co/UDCFQkxge4
— Maxim Jacobs, CFA (@MaxJacobsEdison) December 28, 2021
What’s 51% between friends?
https://twitter.com/toomuchguitar/status/1475881599929438211
Nate Silver went as far as to say the “CDC’s method is crap and should be ignored going forward”:
If they revised last week's figure from 73% to 22% (!!!!!!! Seriously WTF?!?!?) I think we have to assume the CDC's method is crap and should be ignored going forward. https://t.co/6aDc522JdF
— Nate Silver (@NateSilver538) December 28, 2021
Here’s what the CDC says changed:
CDC just responded to my initial inquiry:
"There was a wide predictive interval posted in last week’s chart, in part because of the speed at which Omicron was increasing. We had more data come in from that timeframe and there was a reduced proportion of Omicron." pic.twitter.com/kokHyYVCtb
— David Lim (@davidalim) December 28, 2021
It also means that many of the infections we’re seeing are Delta and potentially more likely to lead to hospitalization and/or death:
Setting aside the question of how the initial estimate was so inaccurate, if CDC’s new estimate of #Omicron prevalence is precise then it suggests that a good portion of the current hospitalizations we’re seeing from Covid may still be driven by Delta infections. https://t.co/ZacVjEhk5x
— Scott Gottlieb, MD (@ScottGottliebMD) December 28, 2021
We’re two years into this whole pandemic and the data is still for s*it?
That "omicron is 73% of U.S. infections" number, as only a few people pointed out at the time, was based on an incredibly imprecise CDC model. https://t.co/8uk7qu6vvy (I'm not knocking CDC for having such a model, worth a try, but we need to acknowledge when we don't know things)
— Patrick Brennan (@ptbrennan11) December 28, 2021
There *should* be a number of corrections right now:
IMO newsrooms that published the 73% number on their front page should be doing a retrospective about how a number with a 95% confidence interval of 34-94% got such prominent play https://t.co/Ni05CGkK7C
— Simon Fondrie-Teitler (@varlogsimon) December 28, 2021
Keep in mind, this has real-world consequences:
HHS zeroed out Regeneron for 10 days while Delta raged based on a CDC Nowcast model that a week later has already missed its (very wide) confidence interval.
Government monopoly purchasing and distribution of medicine is a disaster.
— Phil Kerpen (@kerpen) December 28, 2021
***
Join the conversation as a VIP Member