The Minnesota Department of Health on Wednesday unveiled the latest version of a model projecting COVID-19’s trajectory in the state.
The model was built by MDH in conjunction with the University of Minnesota, and the latest version, “3.0,” takes into account new information that wasn’t available the last time the model was publicly released, on April 10.
Here’s what you need to know about Version 3.0.
What’s different about this model versus the last one?
In a web tutorial (you can watch it yourself here or see the slides here), State Health Economist Stefan Gildemeister outlined some key changes to the model since it was last publicly unveiled, including but not limited to:
- Asymptomatic cases: The previous version assumed 25 percent of cases were asymptomatic. The new version assumes 41 percent are asymptomatic.
- Out-of-hospital deaths: The new model accounts for the fact that some Minnesotans who get very sick are dying outside of hospitals. For example, some older Minnesotans have health care directives asking drastic measures not be taken to keep them alive.
- Social distancing: The new model downgrades the effectiveness of Minnesota’s mitigation strategy. Previously, it assumed initial physical distancing reduced contacts by 50 percent and the stay-home order reduced contacts by 80 percent. Those numbers have been revised to 38 percent and 55 percent, respectively.
- New Minnesota data: When the model was last unveiled, Minnesota had seen 57 deaths. In the new version, data has been fitted to actual hospitalizations and deaths reported in Minnesota up to April 25, by which time there had been 797 hospitalizations and 244 deaths in the state.
- New U.S. data: Other updates include new parameters due to new U.S. data. Previously, much of what was known about the virus came from China, where different demographics and social structure mean different effects.
In the aggregate, these changes relative to Version 2.0 suggest slightly less time until Minnesota sees infections and ICU needs peak and higher mortality.
What does the model say about when the disease will peak now? And how many deaths are expected?
Gildemeister walked through several scenarios the modeling team put together. Please note that all the below numbers are estimates — not precise predictions.
[image_credit]Minnesota Department of Health[/image_credit][image_caption] Where are scenarios 2 through 4? They’re outdated now because they were constructed under Version 2.0 assumptions, but you can read about them here).[/image_caption]
- ICU peak: May 11
- Top ICU/ventilator demand: 4,991 beds
- Mortality by the end of pandemic: 57,035
- Mortality by the end of May: 42,032
Scenario 5: The stay-at-home order ends May 18, followed by three weeks of physical distancing similar to what was in effect before stay-at-home, and continued physical distancing for vulnerable people.
- ICU peak: June 29
- Top ICU/ventilator demand: 3,397
- Mortality by the end of pandemic: 29,030
- Mortality by the end of May: 1,441
Scenario 6: Stay-at-home ends May 31, followed by three weeks of physical distancing similar to what was in effect before stay-at-home, and continued physical distancing for vulnerable people.
- ICU peak: July 6
- Top ICU/ventilator demand: 3,006
- Mortality by the end of pandemic: 28,231
- Mortality by the end of May: 1,388
Will more testing for COVID-19 have any effect on how bad the disease gets?
[image_credit]Minnesota Department of Health[/image_credit]
Gildemeister noted these are hypothetical scenarios, since testing protocols are still being developed, but stressed they are designed to show a range of possible outcomes. The model assumes people who test positive isolate to reduce contact — and transmission — with others.
Scenario 5a: A worst-case testing scenario layered on scenario 5 where the tests are low-sensitivity and fewer people are tested.
- ICU peak: June 29
- Top ICU/ventilator demand: 3,150
- Mortality by the end of pandemic: 26,914
- Mortality by the end of May: 1,430
Scenario 6b: A best-case testing scenario layered on scenario 6 where tests are highly sensitive and widely available.
- ICU peak: July 13
- Top ICU/ventilator demand: 2,444
- Mortality by the end of pandemic: 22,589
- Mortality by the end of May: 1,375
Currently, the model does not account for those who came into contact with a positive case also isolating themselves to avoid further spread. The state’s contact tracing program aims to promote that by informing people if they’ve come into contact with someone who’s tested positive for COVID-19.
What would happen if we stayed at home way past the end of May? What if some new treatment becomes available?
[image_credit]Minnesota Department of Health[/image_credit]
Scenario 7: This scenario looks at how following federal guidelines to, among other things, wait to open up until there’s a 14-day downward trajectory in cases would affect the COVID-19 curve.
Gildemeister said that would likely mean a stay-at-home order into September. Relative to scenario 6 (where the stay-at-home order ends May 31), it has modestly fewer deaths spread out over a longer period of time.
- ICU peak: July 6
- Top ICU/ventilator demand: 1,034
- Mortality by the end of pandemic: 26,294
- Mortality by the end of May: 1,388
Scenario 8: This looks at following the federal guidelines, plus accounts hypothetically for recent research on remdesivir, which early data suggests can reduce hospital stays and possibly mortality.
- ICU peak: July 6
- Top ICU/ventilator demand: 1,034
- Mortality by the end of pandemic: 25,392
- Mortality by the end of May: 1,388
So is the model right?
Critics of MDH’s model have often pointed out that the number of deaths it predicts tends to be higher than other models — like one developed at the University of Washington.
And some say the numbers just aren’t plausible. The scenarios presented estimate there could be between 1,375 and 1,430 deaths by the end of May in Minnesota. So far, Minnesota has seen 638 confirmed COVID-19 deaths, and nine presumed COVID-19 deaths.
To hit the estimated deaths by the end of the month would require, on average, between 41 and 44 deaths per day for the rest of the month. Minnesota has averaged 22 deaths per day in the last week.
“We hope these are pessimistic estimates, but it’s certainly a plausible outcome with the assumption of [continued increased ] spread and wider spread among the population,” Gildemeister said.
Thanks for the analysis!
Thanks, Greta. Better to be pessimistic than optimistic in this context.
Better to be realistic (with a slight negative bias).
The ~25K death estimates for the “end of the pandemic” are highly questionable. Better to have estimates for specific dates such as “end of 2020” and maybe “end of 2021” so we can evaluate the actual results.
I rarely read about the overwhelming number of MN deaths among cases that resided in long-term care or assisted living facilities. Currently over 81% (568/700 on 5/16). Much higher percentage than other states.
MN government officials have done a truly awful job of protecting our most vulnerable who are confined to faciliites and instead spent significant resources sheltering-in-place the rest of the state. Shame on them.
With a 30-55% compliance rate for the “mild stay at home order” we allowed the state government, the burden of responsibility for those senior deaths you pretend to care so much about lies with the citizens of this state who still don’t believe there is a difference between a coronavirus and a “mild flu.” No government can fix stupid and we have too much of that all across the state.
A nice breakdown but they’re still way off the mark. Any scenario claiming 1,000+ deaths by the end of May is way off the mark. There is no chance the total deaths will nearly triple in 2 weeks. This is why the models and model makers need not be given any more say in this. After 2+ months of data they still can’t even come close to reality.
Bob,
The doubling rate is every 10 days right now, so there is every chance we will see 1000+ deaths by the end of may. We are in an exponential growth phase right now. Look at the actual data, it really doesn’t take much to see us hitting that number.
Bob, you weren’t one of those who were thinking we would be nearly done with the virus by the end of May, right?
I agree that every model put out there will be incorrect, and is formulated on data that has already been deemed out of date at time of release. The models provide guidance, not certainty. Bottom line, we don’t enough about the virus or how people will act, but we can be sure that the spike will be higher to extent the expert advice is not heeded.
… your math is not right. “So far, Minnesota has seen 638 confirmed COVID-19 deaths, and nine presumed COVID-19 deaths.” An additional 362 cases would bring us to 1000. That is a ~57% increase from our current numbers, not a tripling…
I believe the word you were actually looking for was prescient. We’re at 672 deaths as of today, with 34 deaths announced. At current pace, we hit 1,000 deaths on the 24th or 25th of May, with a solid week to spare. Or by “off the mark” did you mean, “wow, those modelers were sure optimistic?”
The scenarios above all have >1375 by the end of May. Why are you writing about 1000?
Reading is a another vanishing skill in an all-subjects illiterate nation. They are responding to “any scenario claiming 1,000+ deaths by the end of May is way off the mark” at the top of this discussion.
Also, as the article indicated, many Covid-19 deaths will go unreported/uncounted because of patient DNR orders and a variety of situations that will cause deaths to go uncounted for several years. So, any count you read will be low. Remember when ” Puerto Rico reported that only 64 people died in the hurricane” and how, much later, that turned into an official death toll of 2,975? 15 years after Hurricane Katrina, the death count there is still described as “somewhere between 1,245–1,836, perhaps more.” This is not rocket science because not nearly enough rocket science-quality people are involved.
Mr. Barnes, you were wrong. It’s May 28th and the number of deaths in MN stands at 967, with 35 deaths since yesterday. It’s hard for people to grasp how epidemics work, there is a tendency to underestimate and minimize them. And the danger is that when a people do not pull together in the face of an epidemic, the consequences are even more deadly.
Well, you were incorrect and the data was correct!! There were actually 1040 deaths as of May 31.
As far as the plausibility of the estimates for deaths by the end of May, it is worth noting those are in line with the “ensemble” estimate formed from other models. Some of the individual other models suggest lower numbers, some higher, but on the whole, the Minnesota model is right in the middle. As mentioned, the one developed at the University of Washington is one of those suggesting a lower number; however, it now isn’t dramatically lower, unlike what was the case earlier in the epidemic. To explore all of the models, go to the Reich Lab COVID-19 Forecast Hub: https://reichlab.io/covid19-forecast-hub/
So the summary is vastly different social distancing timelines produce roughly the same number of cases/deaths. Which is why they’re choosing to end it – because they never consider a test/trace real solution.
Given that we have a fairly constant new case rate of a few hundred per day in our current fairly weak stay home rules, it does seem like there’s an outside chance that if people wore masks indoors and we continue to keep bar/restaurants/events closed, we may be able to keep the r value close to 1 for months. It’s just so dramatically affected by tiny changes in the transmission rate, especially when it’s near 1.
Can someone (knowledgeable, please) comment on why the estimated effect of social distancing was reduced? There must have been some basis for that change. I’m not here to challenge, just understand.
https://www.nytimes.com/interactive/2020/05/12/us/coronavirus-reopening-shutdown.html?searchResultPosition=1
This would be my guess. There was no change in the stringency of the requirements or recommendations, but a fairly significant reduction in its effectiveness. If people are starting to ignore the request, then it will have less impact on slowing the spread.
Looks like while there was a lot of congratulation about “pulling together”, the reality is that even the largely sensible state of MN gave lip service to social distancing and stay at home orders.
Apparently the cell phone data giving MN a stay at home grade of F or D was correct.
You don’t have to travel far from the Cities to find that Minnesotans are not pulling together at all. The far right is and always has been about “me first, you never.”
In Cook County, there are no reported Covid cases. Our county has the highest unemployment rate and its may industry is tourism. We have~5100 residents, half of which are over 55 yo. We have 1 portable ventilator with emergency health care services are located 2-2.5 hours away in Duluth.
Residents are very concerned as businesses are anxious to open to that industry and visitors want to escape to a less dense population. Our county could be a test county where the U of M and Rochester uses its partnership to test and trace all its residents and its visitors immediately to track the infection and its outcomes. Please take some action now before the infection spreads, destroying our small, currently healthy community. I am begging that we take this aggressive action. Lives and our small community are at risk!
Why is every scenario worse than reality so far? MN averages 22 deaths a day in May. The model would require daily death rates to average 47-50 per day for the rest of May to be accurate. Very thankful this model is WRONG and probably showing the worse to scare people. Do any scenarios expect 1,000 deaths or less by the end of May?
There were 1,040 cases on May 31st.
Love the arguments about statistics of death. It really comes down to how much do you care about the numbers? If one person 50,000 dies, you could argue that statistically insignificant. If you’re the one person who died it’s pretty damn significant.
Statistic is the discipline of trying to predict the future based on the past and modeling the equation to fit the data. I love the people who argue about the numbers will probably never taken a statistics class in their life. The reality is everyone tried to do the best they could and tried to minimize the risk to health care providers who cared for these people. That is why the state did what it did. And It kept hospitals from being slaughterhouses.
Anyone who disputes that frankly is a self-centered bastard only concerns herself with his or her own trivial concerns and life needs. It would be sad to be dead and considered insignificant. wonder how all the GOP sceptics would feel about their life being insignificant..
Of course the person deciding that would probably be a GOP voter, and they would have the same fate as most poor, people of color or social preferences different from their own biases about what normal is. That is the ultimate irony. They become what they despise, insignificant.