"We’re going to lose anywhere from 75-, 80- to 100,000 people," Trump said.A day or two later, when the New York Times presented evidence that the White house had been warned about a much higher death toll, the White House immediately rejected the report. According to an NPR article, White House spokesman Judd Deere said:
"The president's phased guidelines to open up America again are a scientific driven approach that the top health and infectious disease experts in the federal government agreed with. The health of the American people remains President Trump's top priority"Let us do a quick reality check. The "top health and infectious disease experts in the federal government" are working for or with the CDC. The CDC maintains a web page that lists a number of different computer models, and explains clearly why forecasting COVID-19 deaths is critical.
The web page also includes a link to a downloadable spreadsheet file that contains the numbers for all the models listed. I looked at some of the models (which now do not include the terrible IHME model anymore), and then downloaded the data file to have a closer look. Some models only predict one week into the future, but most models predict between 4 and 7 weeks ahead. Here are the total numbers of COVID-19 deaths that the models predict:
Model
|
Projected COVID-19 Deaths |
|---|---|
| CU-80 contact |
336,930
|
| LANL-GrowthRate |
242,352
|
| CU-70 contact |
242,043
|
| CU-60 contact |
219,258
|
| YYG-ParamSearch |
218,893
|
| COVIDhub-ensemble |
189,105
|
| UT-Mobility |
160,977
|
| MOBS_NEU-GLEAM_COVID |
143,314
|
| UMass-MechBayes |
135,688
|
| MIT_CovidAnalytics-DELPHI |
119,337
|
There is quite a variation in the predictions, which range from 119,337 deaths to 336,930 deaths over the next 4-7 weeks.
Even the most optimistic model predicts close to 120,000 COVID-19 deaths. On average, the models predict about 200,000 deaths; half of the models predict more than 200,000 deaths. There is not a single model that supports the numbers Trump gave; all numbers are substantially higher. A "scientific driven approach" would list the range that the models indicate, perhaps leaving out the highest and lowest predictions: 135,000 to 242,000 deaths.
Note at all of the model predictions cover a period shorter than the next 2 months, and that all models still predict a substantial daily death rate at the end of the prediction period. In my model runs, the total deaths toll at the end of 2020 was about 50% higher than for the next 6 weeks. Clearly, the overall expected death toll in 2020 is even higher than the numbers in the table, quite possibly twice as high.
Furthermore, most of the models assume that social distancing measures remain in place. There is absolutely no doubt that COVID-19 transmissions and therefore deaths will increase when states "re-open", as many states have started to do. Nobody knows exactly by how much, but even a modest increase of transmissions by about 20% would result in more than 400,000 additional deaths. A more dramatic increase could lead to more than a million additional deaths in 2020. The only thing that will hopefully prevent such an enormous increase in deaths will be state governors who "lock down" states again when they see case numbers increasing. But it looks like they will have to do so against the resistance of the White House.
