This model is being used to dictate coronavirus response policies in the U.S. Why is it so far off?

Wednesday, April 8.

That’s the date a COVID-19 projection model by the University of Washington’s Institute for Health Metrics & Evaluation projects will be a dark day in Tennessee.

It’s the day the model says the demand for hospital beds will surpass the availability of those beds in Tennessee.

In just three days, the model says, there will be so many Tennesseans ill with COVID-19 and needing to be hospitalized that there won’t be enough beds available for them.

Tennessee has 7,812 hospital beds, and 629 ICU beds, available. But on Thursday, April 9, there will be 8,626 hospital beds needed. Even worse, there will be 1,544 ICU beds needed.

That seems outlandish, considering that as of Sunday afternoon, there had been a grand total of 328 coronavirus patients hospitalized in Tennessee — a number that includes those still in the hospital, and those who have been discharged or who have died.

And if that’s not enough, consider this: The model projected that Tennessee would need 4,870 beds, and 1,028 ICU beds, by Sunday.

Today is Sunday. There haven’t even been 4,870 patients diagnosed with coronavirus in Tennessee. There have been 3,633. And only 328 of them have required hospitalization.

The University of Washington’s coronavirus model paints an incredibly grim picture for Tennessee. By April 13, the model suggests, there will be more than 100 Tennesseans dying of coronavirus every single day. By the peak of the virus outbreak in Tennessee a week later, there will be 165 deaths per day, leading up to more than 3,400 total deaths by May.

But here’s just how pitifully wrong the model is. It projected that by today, there would be 27 deaths in Tennessee. In a single day.

Instead, there was one (1).

By tomorrow, it said, there will be 34 deaths in a 24-hour period. And, by the next day, 42 in a 24-hour period.

As of today, April 5, the model suggested there would be 105 total deaths in Tennessee. Instead, there have been 44.

So why is this model being used to dictate public policy as America combats the coronavirus outbreak? It’s been mentioned by President Donald Trump’s administration, specifically cited at press briefings by Deborah Birx, the respected physician who is leading the White House’s coronavirus response. It’s also been cited by Tennessee Gov. Bill Lee, and many other political leaders across the U.S.

This isn’t to suggest that the coronavirus isn’t a legitimate threat. Clearly, it is. Look at New York. Look at New Orleans. And Illinois, and Michigan, and etc. But that’s all the more reason to be sure that our response — which is inevitably going to cost millions of Americans their jobs and usher in a dark economic era for the United States — is being based on correct data.

Not only is the UW model faulty data, but no one is questioning it. And when I say no one, I mean my newspaper — the Independent Herald — published a story about its latest projections for Tennessee on Sunday evening.

I was skeptical of the model’s projections. I sat down after the story was published to record a podcast about the model, and how it stretches the imagination to think that more than 100 Tennesseans will be dying per day within one week, based on the current data that we have in front of us. As I digested the data, I realized that the projections were even more enormously erroneous than I had first realized. So erroneous that I decided to un-publish the story. I firmly believe in presenting all available information to keep readers informed. Less info is never a good thing. But information this erroneous doesn’t deserve to see the light of day. By the time I got to the newspaper’s Facebook Page to delete the post promoting the story, BJ Gislason — a DJ at the local radio station — had correctly pointed out that the model’s projections were so far off that total hospitalizations should’ve already overwhelmed capacity in Tennessee … when, in fact, nothing could be further from the truth.

At one point, when it was first published, the UW model projected that Tennessee would be one of only nine states across America where the coronavirus outbreak did not overwhelm health care capacity. At the peak of the outbreak, the model projected, there would be 35 Tennesseans dying in a single day.

The original version of the model seemed much more believable. But given how outlandish the updated version is proving to be, it’s safe to say that every state’s numbers should be called into question.

The model projects that 262,092 hospital beds will be needed across America when the virus outbreak peaks — exceeding capacity by more than 87,000 beds. It projects that 39,727 ICU beds will be needed — exceeding capacity by nearly 20,000 beds. When the virus peaks in 11 days, the model says, more than 2,600 Americans will be dying per day. An eventual total of 93,500 deaths are projected.

For the record, 2,600 Americans dying per day does not seem like too much of a stretch. The single-day death toll topped 1,000 on Sunday — and the outbreak is still growing. And it’s projected total for this date — April 5 — isn’t far off. It projected 9,800 deaths by now, and we’re at 9,600.

But, all of a sudden, there’s no way to put much faith in any of the model’s projections.

How could it be so wrong for Tennessee? And if it’s so wrong for Tennessee, how wrong is it for other states, and what does that say about the national projections? After all, it projected 179,000 hospital beds to be needed nationally as of today. There have only been 337,000 cases in the U.S. to date. Even if all 337,000 cases were active (they aren’t), that projection would assume that in excess of 50% of all coronavirus patients require hospitalization. In Tennessee thus far, about 9% have required hospitalization.

By the way, the model isn’t just over-estimating the COVID-19 burden. It’s under-estimating it in some states. In Kentucky, it projected there would be 21 total deaths by today. There’s been 45.

Again, the grand total of 93,000 American deaths projected by the model doesn’t seem to be far from reality. But given how woefully bad this model is performing with individual state case totals, how can it be trusted at all? And how can it be used to dictate policy?

Ben Garrett

Ben Garrett is a journalist from East Tennessee. He is publisher of the Independent Herald, a weekly newspaper serving the Big South Fork region of the Cumberland Plateau, with a sideline in website development and digital marketing. He is also an erstwhile blogger.

Add comment

Follow us

Don't be shy, get in touch. We love meeting interesting people and making new friends.

Most popular

Most discussed