Minnesotans are preventing tens of thousands of coronavirus deaths by making unprecedented sacrifices of personal freedoms and commerce in the face of the pandemic, according to the model that informed Gov. Tim Walz’ stay at home order.
The model suggests the number will easily be in excess of 20,000 deaths prevented, although those behind the model are wary of such specifics. And actually, a lot more deaths will almost certainly be prevented as the state and nation continue to adapt — but the model has yet to estimate numbers for that, according to those who built and continue to work on the model.
The 20,000 figure — a rough estimate with high uncertainty — comes from a COVID-19 model developed by a team at the University of Minnesota’s School of Public Health and is believed to be among the first models to incorporate Minnesota-specific data. Walz leaned heavily on the model in his statewide address March 26, in which he announced his stay at home order and pledged to Minnesotans that his decisions would be made not on fear but on the best data available.
Models of epidemics are fraught with uncertainty, and that’s especially the case now, as statisticians attempt to game out the behavior of a virus that was unknown to science four months ago among a population that has never experienced anything like it.
Nonetheless, such models are among the few tools policymakers have to inform them as they make decisions for how to respond, such as whether to order residents to stay at home — and for how long — and how many hospital beds will be required to treat the critically ill.
Several members of the modeling team and officials with the Minnesota Department of Health, which contracted with the U to develop the model, discussed it with the Pioneer Press.
WHAT THE MODEL SAID
By the time Walz addressed the state Thursday, the model, which had been rapidly developed in the week prior, had only analyzed two scenarios in depth, each of which sought to project what might happen moving forward from Sunday, March 22.
Scenario 1 envisioned what would have happened if the state had done little and had returned “to normal.” Under that scenario, the number of seriously ill would exceed the state’s capacity of intensive care unit beds within five weeks or so, and the death toll in Minnesota could reach as high as 74,000. Walz discussed this number.
Scenario 2 envisioned what Walz ultimately decided to go with: a three-stage plan. That plan included a stay at home order for two weeks; followed by a continuation of closures and restrictions on entertainment venues and restaurants until May 1 and keeping schools closed until May 4; followed by an undetermined period of restrictions regarding the most vulnerable, including the elderly and people with underlying conditions that make them more susceptible.
Walz did not say how many would die under that scenario. The model pegged the number at between 50,000 and 55,000 and suggested the number of deaths prevented could be around 20,000.
But those involved with the model wince at giving that figure too much stock, calling it an “incredible mismatch” for the reality on the ground.
More importantly, they emphasize, the model projected that the restrictions would buy time — six weeks or so. Instead of the state reaching ICU capacity in about five weeks under the do-nothing scenario, it would reach capacity in about 11 weeks — give or take two weeks on either side.
TIME EQUALS LIVES SAVED
This is an important distinction: The model suggested that “flattening the curve” — spreading out the period of time that people would be infected and ill to lower the peak — would be far more difficult than a mere two-week stay-at-home. The model doesn’t currently attempt to project economic impact, but Walz has said that it’s an important consideration of any response, and there’s little question that the stay at home order is exacting a daily toll on the economy, not to mention the psyche.
“The idea of flattening that curve requires a pretty long period of social distancing,” said Eva Enns, an associate professor at the School of Public Health and part of the team that developed the model. “When we’re talking two weeks, three weeks, a month, two months, we’re still not seeing a flattening of the curve, but rather a delay of the curve, which is why the focus is on what you do with that additional time.”
So instead of flattening the curve, the model projects the curve gets pushed to the right, and in that gap is time.
That time is critical for saving lives because it allows ICU capacity to increase.
COVID-19 most often kills by infecting the lungs and bringing about a crushing type of viral pneumonia where the body’s own immune system appears to work against it. The lungs are unable to convert enough oxygen from the air into the blood. For patients in the most serious stages, the only way to keep them from dying is to run a tube down the windpipe (intubation) and use a machine — a ventilator — to help them breathe.
The good news is that if this can be done, the likelihood of dying decreases dramatically. If no ICU bed is available, the chances of dying increases by between 1.5 and 16.5 times, according to numbers used in the model that are based on what has been seen elsewhere.
The bad news is that there’s a limited supply of ventilators, staff trained to use them and available beds to bring it all together. But that number is increasing daily, as the flu season wanes and as hospitals attempt to stand up COVID-dedicated facilities.
ICU CAPACITY IS KEY
According to the model, the time the state could reach ICU capacity is some time between the weeks of May 24 and June 21.
But modelers were conservative, only estimating a peak ICU capacity of 1,000 beds. In fact, the state already has more than 1,200. Although most of those beds are occupied on any given day — the number of available ICU beds has hovered around 235 recently — health and hospital officials are hoping to increase that number significantly, but they haven’t provided clear numbers yet.
Every additional ICU bed saves multiple lives.
“What you and your readers absolutely need to understand is that we hope as ICU capacity goes up and is built up over the next week or the next two weeks, these numbers will change because we expect as more people who will need ICU care and are very sick have appropriate ICU care, more will survive their horrendous illness,” said Stefan Gildemeister. He’s the state’s health economics director who’s acting as a primary liaison between the Department of Health and the modeling team at the U.
“So to say 50,000 (deaths) is to say the bed capacity will stay around 1,000 ICU beds,” he said. “To say that means there’s an incredible mismatch between what we know is coming and are already working on. … The 74,000 (deaths if nothing was done) is not the final number, but please understand that the 50,000 to 55,000 number is also not a final number.”
Still, the state’s hospitals, National Guard and emergency management officials have their work cut out for them.
The U’s model estimates the number of patients needing ICU beds at the height of Minnesota’s epidemic — regardless of when that comes — will be 5,000.
A different model, created by a team at the University of Washington’s medical school, attempts to forecast what each state’s death toll will look like. That model projects a peak demand in Minnesota coming around April 24, but with a lower ICU demand than the U model. It anticipates that Minnesota will be close to meeting that need, vastly reducing the mortality rate. Total deaths by early August, according to that model, would be fewer than 1,300.
It’s unclear if that model accounts for some of the Minnesota-specific data that the University of Minnesota’s model includes, such as health and demographic data about Minnesota — how many people of what age have which underlying conditions. It’s those sort of details, Walz and state health officials have said, that can make the U’s model so valuable.
Health Commissioner Jan Malcolm on Sunday noted that some health systems that operate in Minnesota are using their own models, and that Minnesota’s modeling team is planning to get together with other modeling teams to “compare notes.”
IS IT WORKING?
In addition to factoring for ever-increasing ICU capacity, the University of Minnesota modeling team hopes to integrate more Minnesota-specific data, such as geographic population densities and travel habits.
One of the model’s most important variables is how infectious the coronavirus is — how many other people one person typically infects. That number, known as the “R-naught,” is currently estimated to be 2.5, although in some countries the rate of transmission appears to be closer to 3.
It’s unquestionable that that number is drastically reduced by social distancing. That’s why the stay at home order is presumed to buy time: People who stay home can only infect others in the household. People who live alone would have an R-naught of zero if they could stay home and recover until they’re no longer contagious.
But there’s no practical way to know — statistically — how well Minnesotans are heeding the order right now. Because it generally takes about five days for people to start showing symptoms, and because it takes often more than a week after that for those who will need hospital attention to reach that point, it will be hard to know how well Walz’ strategy is working until it’s close to expiring, around April 10.
But it’s likely the U’s model will play a pivotal role in whether he decides to extend the order.
Malcolm on Sunday warned of the limits of any model.
“It’s critical to say over and over again that the models really aren’t intended to be any sort of a crystal ball or prediction, per se, but really more of a tool to model out what interventions would do,” she said.