As states are opening up, here's how to assess whether restaurant or other public spaces are taking the proper COVID-19 safety measures. USA TODAY


In early March, University of Florida research scientist Thomas Hladish put together a simple mathematical model to forecast the spread of the novel coronavirus.

The Florida Department of Health had requested his assistance to predict the demand for hospital beds in the state, and Hladish, an epidemiologist and mathematical modeler who specializes in the study of the dengue virus, dropped his normal research to provide quick answers about the COVID-19 pandemic.

The results of his work were clear: Florida was in line for a huge public health crisis.

“We looked at what had happened elsewhere, we looked at how rapidly the epidemic was growing in Florida, and we said, ‘OK, we’ve seen this story play out before,’” Hladish said.

But while COVID-19 has infected more than 50,000 people in Florida and killed more than 2,200 as of Saturday morning, the state has fared better than most experts predicted. Hladish acknowledges the virus “has not progressed the way I expected,” pointing to cell phone mobility data showing Floridians hunkered down weeks before Gov. Ron DeSantis ordered much of the state to close. His model didn’t anticipate that.

Read more: Cellphone data shows Americans respected stay-at-home orders but are starting to move again

Infectious disease modelers like Hladish, who is building a more detailed model for Florida, say that despite their inevitable flaws, the models they make are important tools for leaders making decisions about COVID-19. They give government leaders a way to make sense of the data they have, and they can help predict what a virus might do in the future.

But as he guides Florida through a phased reopening, DeSantis has been openly dismissive of models, saying they’re making unreasonable assumptions and don’t take unique state efforts into account.

You may also be interested in: Florida scientist said she was fired for 'refusing to manipulate' COVID-19 data

At a news conference at Gulf Coast Medical Center in Lee County in mid-May, DeSantis questioned if “any of the models have been accurate so far?” Rather than focusing on models and “conjecture,” he said he is making decisions “based on facts,” like hospitalization rates and infection rates.

“I think that you go by the facts when you actually know what’s happening on the ground,” he said. “Let’s just be honest, the models have not been accurate.”

Projecting massive fatalities

DeSantis has called out one model in particular for its dire forecast.

In late March, the Stanford University-based modeling group COVID ACT NOW released projections that showed Florida hospitals could be overloaded with more than 465,000 patients by April 24.

The actual number of hospitalizations was just a fraction of that, less than 5,000.

“For those who know this business, Florida’s only got 70,000 licensed hospital beds,” DeSantis said, noting that if that projection was right, there would be hundreds of thousands of sick people with nowhere to go. “So you’re really projecting massive, massive fatalities.”

Show Thumbnails
Show Captions

But the COVID ACT NOW model’s projection of 465,000 patients was never a likely scenario. Rather, it was a worst-case scenario if Florida’s leaders and residents took little action to stop the virus’s spread. With mandatory shelter-in-place orders, the model projected about 18,000 hospitalizations by July, a number more in line with the roughly 8,200 people hospitalized through mid-May.

“We proved that sheltering in place actually worked,” Leo Nissola, a cancer immunotherapist who works with COVID ACT NOW, told USA TODAY Network — Florida in late April.

The London-based Imperial College model also has received intense criticism for its dire mid-March forecast of up to 2.2 million COVID-19 deaths in the U.S. But that too was a worst-case scenario, based on what could happen if no one changed their behavior to curtail the spread of the virus.

“You have to be able to say, ‘Look, if you do nothing, this is what we think will happen.’ It’s not that we think you will do nothing,” Hladish said. “The worst possible outcome is not really possible because people won’t let it happen. People will shut themselves in their houses before you could get to Armageddon.”

Hladish said those worst-case scenarios are not nonsense, even if they are highly unlikely to materialize.

They are important reference marks, he said, but “it sort of ends up playing into the story that the models are so pessimistic and the modelers have no idea what they’re doing.”

In case you missed it: US coronavirus curve may be flattening, but estimated death toll keeps rising. What does this mean?

Overly optimistic forecasts

One prominent COVID-19 model in the U.S., produced by the University of Washington’s Institute for Health Metrics and Evaluation, has been criticized not for its pessimism but for being overly optimistic and for projections that sometimes fluctuate wildly.

The IHME projections have been used to guide White House decisions. Jason Mahon, a Florida Division of Emergency Management spokesman, said in an email that it is a model the agency looks at regularly.

In late March, the IHME model was forecasting about 80,000 deaths nationally by August. But in early April it dropped its estimate to about 60,000 — a mark the country hit just a few weeks later, by April 29.