Hot Air

Leaked CDC model projects: Next month daily deaths will reach 3,000, new daily cases will reach … 200,000; Update: “Unfinished projections?”

This is why it’s a bad idea for Trump to keep tossing projections into his public statements. The numbers are getting worse. The more he has to move the goalposts to keep up with them, the more he inadvertently highlights how much worse they’ve gotten.

If he’s going to do a forecast, he should set the bar high. “Anything less than 500,000 dead — wait, I mean a million dead — is a win.”

The Times somehow got hold of a FEMA chart based on numbers put together by the CDC. (How they got hold of it is itself an interesting question.) Most of us are already grimly inured to daily reports of 1,500-2,000 dead and 30,000 more infected. Imagine how much more inured we’ll be on June 1 if this data pans out. One death is a tragedy; 3,000 deaths every 24 hours is a statistic.

Note the difference between modeled deaths and reported deaths. Right now the U.S. is averaging around 1,700 daily, with numbers somewhat below that on weekends and above it on weekdays. The CDC model had expected fewer than that. If it’s trending conservative, how much worse might the daily death toll be on June 1 if than the 3,000 per day that the agency is expecting?

That is, the red line represents a 50th percentile projection. The model suggests that deaths could be lower somewhere in the red band — or higher.

Interestingly, the CDC’s projection on new cases doesn’t track with its projection on deaths. Remember that we’re averaging around 25-30,000 new confirmed cases per day right now:

The model expects daily deaths to rise by 40 percent or so over the next month but it expects daily cases to rise by around 750 percent. How does that square? All I can think is that the CDC assumes there’ll be a massive increase in testing across the U.S. this month, which would be great news, and that we’ll begin detecting many more extremely mild or even asymptomatic cases. Either that or they’re expecting the disease to become much less lethal over the next 30 days, and that can’t be right. Not even remdesivir has shown a mortality benefit in treating COVID-19.

Er, why is this leaking today? Theories are being kicked around:

I don’t understand Silver’s point. The projections hurt the White House more than help it by predicting that the crisis is about to deepen, and the fact that the CDC model *underestimated* the number of daily deaths to this point suggests that the feds have done an even worse job in mitigating the outbreak than expected.

As for Bernstein’s theory, maybe whoever leaked it did so to try to spook governors into rethinking their plans to reopen more widely. Jeremy Faust is an ER doctor in Massachusetts who’s worried that some states are reopening too soon even though he understands the urgent need to restart the economy. It’ll backfire, he thinks, by spooking consumers more than a delayed reopening would:

Faust is convinced that the average person doesn’t appreciate the risk of infection, partly because they don’t know anyone who’s had the disease (New Yorkers are different, of course) and partly because they’re gauging their personal exposure based on the number of confirmed cases locally. But confirmed cases don’t tell us much, he notes. We all understand that the number of asymptomatic carriers walking around is much greater than the number of people with confirmed infections. What does that mean in practice? Quote:

Here’s an instructive way to think about this. In Atlanta, the cars of MARTA, the rapid transit system, have a seating capacity of 64 and a total capacity of 96. Georgia had reported 26,000 cases as of Friday, though many experts assume the number is far higher. But assuming that is right, the chance of being on MARTA car with someone with the virus is 21 percent. Even if somehow only 10,000 people in Georgia have the virus, the chance of being exposed to someone with the virus during one single train ride is still around 9 percent. But if the real number of cases in Georgia is 50,000 or even 100,000, as some believe, the chances are 36 to 60 percent. And that doesn’t even take into account the viral particles left by the commuters who have already left the train. Nor repeated rides day after day.

Anyway. The White House is disavowing the CDC model — but not because it’s “fake news” or the assumptions are necessarily wrong. The claim is that it hasn’t been vetted by the coronavirus task force led by Mike Pence:

Fair enough, although it continues to be deeply weird that the CDC, the country’s foremost agency devoted to disease control, has been largely sidelined in the federal response. The model that leaked to the Times wasn’t from the Commerce Department or some other arm of the government that has no business making projections about the pandemic. It’s from the CDC. Why shouldn’t we take it seriously?

By the same token, why shouldn’t they be held to account for it if, a month from now, their projections turn out to have badly overestimated the number of deaths and new cases? We should want to know if the CDC knows what it’s doing or not. Based on the early evidence this year, it’s emphatically “not.”

Here’s Scott Gottlieb summing up the past six weeks. The lockdown era helped, but it didn’t quite achieve what we hoped it would.

Update: I leave you to judge whether this is spin by the feds to clean up a distressing leak or the actual truth.

The White House and the Centers for Disease Control and Prevention quickly disavowed the report, though the slides carry the CDC’s logo. The creator of the model said the numbers are unfinished projections shown to the CDC as a work in progress. The work contained a wide range of possibilities and modeling was not complete, according to Justin Lessler, an associate professor of epidemiology at the Johns Hopkins Bloomberg School of Public Health, who created the model…

“I had no role in the process by which that was presented and shown. This data was presented as an FYI to CDC … it was not in any way intended to be a forecast,” Lessler said.

Lessler insisted, however, the numbers show how moving to reopen the country could spiral out of control. He said 100,000 cases per day by the end of the month is within the realm of possibility.

Leave a Reply