The New York Times had a story at the end of last week: a Department of Homeland Security document (a PowerPoint slideshow, in fact) saying that a “best guess” death toll, if shelter-in-place policies are lifted after 30 days, would be 200,000. That contradicted Donald Trump’s own claim that the count would come in under 100,000.
I’ve taken a look at the slides, and they do indeed say what the NYT reported them as saying. But I wonder if they’ve buried their own headline.
That is: the slides do not only predict the number of deaths; they predict the number of people infected. Under a do-nothing scenario, the “best guess” is 195 million infected and 300,000 dead; under shelter-in-place for 30 days followed by other “steady-state mitigation measures”, it’s 160 million and 200,000.
The “infection fatality rate”, remember, is deaths divided by infections. For the do-nothing scenario, 300,000 divided by 195 million is 0.0015, or 0.15%. For the shelter/mitigation scenario, 200,000 divided by 160 million is 0.00125, or 0.125%.
An IFR of 0.15% would be below even the most optimistic estimates. The Imperial model has assumed an IFR of 1%. Early findings from Germany and Iceland put it more like 0.4%, but people generally expected that to be an underestimate.
If the IFR is really low, that means that more people have had it, which in turn means that we might be further through the epidemic than we thought.
I’m uncomfortably aware that this is exactly what I want to believe, so I ought to be extra wary. Also, I don’t know what the Homeland Security numbers are based on. You’d hope that it’s not just plucked out of the air and that they have some sort of testing data that they haven’t released, but who knows. I certainly don’t suggest that we ought to change course on the back of one leaked PowerPoint slide.
But if it’s anywhere near true — even though it mentions 200,000 deaths — it’s actually good news, not bad. It would be down near the most optimistic estimates. Of course, if they’re wrong and the IFR is much higher, then the predicted death toll should be correspondingly higher.