Minnesota researchers provided the first detailed look Friday at the predictive modeling that influenced Gov. Tim Walz's "stay-at-home" strategy — which began Friday at 11:59 p.m. — to slow the spread of a novel coronavirus that has caused four deaths in the state.

The good news is that the strategy could reduce predicted COVID-19 deaths by up to a third, according to the modeling, and even more if it buys the necessary time for researchers to come up with vaccines or drugs. But the death toll could still be steep, and the modeling still predicts a point this spring or summer when an overwhelming 2 million Minnesotans are infected all at once.

"My instinct was, 'that can't be right. That is so incredibly high it can't be possible,' " said Stefan Gildemeister, the state health economist who conducted the modeling along with researchers from the U's School of Public Health. While "shocking," he said the results are only estimates based on assumptions about a still-unfolding global pandemic.

Health officials on Friday also reported two of the state's four deaths from COVID-19, the respiratory illness caused by the virus. Both involved people in their 80s who lived in long-term care facilities in Hennepin and Martin counties. Minnesota's confirmed case count is now 398 — with 34 people hospitalized with the illness and 17 receiving intensive care.

Walz issued a two-week stay-at-home order, with plans to follow that with three weeks of lesser restrictions — which will still include dine-in restaurant, bar and school closures. After that, restrictions could be reduced except for the elderly and people with other health problems who are most at risk of severe COVID-19 infection or deaths.

Two of the models by the state and university researchers compared the impact of this strategy vs. doing nothing at all and found that it could reduce deaths by as much as a third.

Both estimates are stark. The do-nothing model predicted 74,000 deaths in Minnesota over the entire course of the pandemic, but the model for the current state strategy still predicted 50,000 to 55,000 deaths, Gildemeister said.

State officials urged extreme caution in interpreting these figures, which are based on a number of assumptions, such as the amount of face-to-face time by Minnesotans that could spread the virus. Researchers also used the global average for the infection rate of the coronavirus, but it might not spread as quickly in Minnesota as in states or cities with more population density. Researchers added that the numbers weren't created to predict death tolls, but rather to assess how much change could be expected under different social policies and restrictions.

"We get nervous when we focus on the number," said Shalini Kulasingam, one of the U researchers. "It's an estimate based on the data we currently have in hand.

"What we're all hoping for is at some point that we can actually deploy a vaccine" that dramatically changes the predicted outcomes, she added.

The goal of the "stay-at-home" intervention is to delay the peak of COVID-19 cases by as much as five weeks, which will buy hospitals time to acquire more supplies and ventilators and give researchers more time to study potential drugs and vaccines.

"The scenarios that had been run so far really do not reflect the precise mix of strategies the governor has chosen, and certainly don't reflect the impact of any of those strategies," said Jan Malcolm, state health commissioner.

Ventilators are of particular need. While 80% of those who contract corona­virus infections have only mild to moderate symptoms, as many as 5% suffer severe symptoms and breathing problems that require intensive care.

While the state now has 1,268 ventilators for adults, Walz said that might not be enough at the peak and that "we are working on procuring what we think is going to be needed."

Under the do-nothing model, the state would run out of intensive care beds in six weeks and see a peak in COVID-19 cases in nine weeks. Now, state officials hope the peak won't come for 14 weeks.

Other national models haven't predicted such severe outcomes, even for Minnesota, including those by the University of Washington Center for Health Trends and Forecasts, for the next four months. But each analysis is unique and based on different sets of assumptions.

Key assumptions by the Minnesota researchers included that face-to-face contact — the most likely method of virus transmission — will be reduced by 80% for the next two weeks due to the number of people staying at home and away from large gatherings.

Walz said Minnesotans have already practiced social distancing and that the results have shown up in various ways, including a 49% decline in traffic accidents.

Total predicted infection estimates for the state also rely on the current thinking that people are inoculated after initial infections and can't get sick again — at least not right away.

Researchers in China recently infected monkeys with the virus, and then challenged them again and found no second cases of illness, said Michael Osterholm, director of the U's Center for Infectious Disease Research and Policy. "We now have evidence to some immunity that occurs after infection, which is very important in terms of stopping ongoing transmission."

Minnesota's modeling shows a sharp decline in cases after the outbreak peaks, because so many people will have recovered and will no longer be infection risks. Osterholm said that presents an opportunity to mobilize people who have recovered.

So far, the state Health Department reports 180 confirmed cases of people who have recovered and are no longer required to be isolated.

"These are the people who can walk in the mouth of the lion and not worry about it," Osterholm said.

State researchers have conducted dozens of models, including looking at ways to "flatten" the peak of cases through the extension of stay-at-home conditions for months. The researchers did not reveal those models on Friday, noting they are unrealistic in a U.S. democracy compared to countries with authoritarian governments that could force those conditions.

Osterholm said data modeling results can vary dramatically but still don't change the looming impact of COVID-19 or the need for people to take it seriously right now.

"This is going to be a very serious public health challenge that's going to stretch our health care system to the limit. That I have no doubt about," he said. "And whether it's 5,000, 20,000 or 50,000 deaths, realizing every one of those is a real person, not just a number, that's an important point — but either way you're going to stress the health system immensely."

Staff writer Glenn Howatt contributed to this report.