Covid Models Get It Wrong


I admit I’m jealous. In my profession as an orthopedic surgeon, I’m not allowed to make mistakes. If I do, someone is sure to file a medical malpractice case against me. But in the world of meteorologists – and Covid models – mistakes are ignored with no loss of credibility. When are we going to learn?

Case in point is the University of Washington Institute for Health Metrics and Evaluation (UW), which has been consistently wrong about predicting Covid future experience but is still quoted by the media. It’s about time we stopped talking about their models.

The Wall Street Journal has also had enough. Their editorial board has called out their mistakes. Here’s some of their predictions from the recent past:

  • UW forecast New York would need 49,000 regular beds and 8,000 ICU beds at the peak of the pandemic.
  • N. Y. Governor Andrew Cuomo’s experts projected 140,000 regular beds and 40,000 ICU beds needed.
  • Actual regular beds needed was 18,825 and 5,225 ICU beds at the peak
  • Actual regular beds needed were only 38% and ICU beds 65% of the UW model predictions – even though New York was the epicenter of the pandemic.
  • Actual regular beds and ICU beds needed were only 13% of Governor Cuomo’s experts’ predictions


Governor Cuomo now admits their mistakes. “All the early national experts, ‘Here’s my projection model.’. . . They were all wrong. They were all wrong. . . There are a lot of variables. I understand that. We didn’t know what the social distancing would actually amount to. I get it, but we were all wrong.”

Despite the media’s attempts to overhype the resurgence of cases, the actual numbers are tell a different story. In Texas, hospitalizations have been climbing, but weekly fatalities are down 40% from a month ago. Covid patients occupy fewer than 55 of all hospital beds, and more than a quarter are available. Even in Houston, where the largest increases in hospitalizations have occurred, Covid patients occupy only 6% of hospital beds. More than 20% are unused.

Covid-19 patients take up a small share of ICU beds in most states that have reopened including California (16%), Texas (11%), Georgia (10%), Utah (9%), Wisconsin (8%) and Florida (7%). Nearly all states have plenty of hospital beds and ICU capacity.

Yet the UW models are once again predicting dire expectations and the media is promoting their claims despite their earlier mistakes. UW now projects that reopening will cause deaths to triple in California and increase six-fold in Florida and Arizona through September. I live in Florida – and I’m not worried.

Stanford epidemiologist John Ioannidis discusses models and why their predictions are often wrong in a new paper. He says models overshot by making faulty assumptions about virus reproduction rates and homogenous susceptibility. A Massachusetts General Hospital model predicted more than 23,000 deaths within a month of Georgia reopening but the state had only 896.

Ioannidis writes, “In the presence of strong groupthink and bandwagon effects, modelers may consciously fit their predictions to what is the dominant thinking and expectations – or they may be forced to do so. Forecasts may be more likely to be published or disseminated, if they are more extreme.”

In other words, the worse the prediction, the more the media will accept and promote it. No wonder the media – and the models -have lost all credibility.

No comments yet. You should be kind and add one!