STATEWIDE - The numbers are stark and sobering:“If we just let this thing run its course and did nothing, upwards of 74,000 Minnesotans could be killed by this.” – Gov. Tim Walz, …
STATEWIDE - The numbers are stark and sobering:
“If we just let this thing run its course and did nothing, upwards of 74,000 Minnesotans could be killed by this.” – Gov. Tim Walz, March 25
“So you’re talking about 2.2 million deaths, 2.2 million people from this.” – President Donald Trump, March 29
Both men were describing what could happen during the current COVID-19 pandemic without preventive steps. But where did they get their estimates?
The answer is, from sophisticated computer-driven mathematical models developed by epidemiologists, specialists who draw on diverse sources of data to understand the nature of a disease and its potential spread and to develop scenarios to guide public health decision-makers.
Trump’s decision to extend federal guidelines through April was reportedly driven by a dozen different scenarios presented to him by members of his coronavirus task force. For Walz, however, national models weren’t sufficient to make decisions about the lives of Minnesotans.
“We don’t look exactly the same in demographics as Italy or China or New York City but some of the patterns that emerge can be modeled,” Walz said in a March 25 telecast.
So the governor enlisted the aid of experts from the University of Minnesota School of Public Health and the Minnesota State Department of Health to come up with scenarios specific to the state’s population demographics.
Dr. Shalini Kulasingam is an associate professor in the U of M Division of Epidemiology and Public Health with extensive expertise in decision modeling. She came to the U of M in 2012 after working at the Duke University Center for Clinical Health Policy and Research for seven years and she spoke to the Timberjay this week.
“When we did this, our first iteration of the model, which was just two weeks ago, is to say, ‘What would happen to them if they became exposed to the virus, if they become infected,’” Kulasingam said Tuesday. “How many of them would recover and not really have any symptoms? How many of them would be symptomatic and go to the hospital, and basically what would happen in the hospital, how many would end up in the ICU, how many would end up potentially dead? What we can then do once we’ve mapped out the Minnesota population and we’ve said this is what we think will happen to them if they were to get infected, then we can do these what-if scenarios.”
All the scenarios they ran projected approximately 2.4 million Minnesotans will contract the virus, with nearly 2 million having relatively mild symptoms that do not require hospital care, Walz said. With the “no action” scenario, infections would peak in nine weeks, but the state’s 235 intensive care unit beds would be filled in only six weeks, leading to disastrous consequences for the most seriously ill.
“That time from six weeks to nine weeks is exactly what is happening in Italy where you see the number jump from 10 to 50 to 500 a day dying because there is no ICU capacity,” Walz said. “That’s what we need to avoid.”
The models produced by the state’s epidemiologists provided insight that helped Walz and his public health advisors determine a course of action that would push both the peak infection rate and ICU demand out an extra five weeks. Beginning last Friday, the governor implemented a two-week shelter-in-place order, closing all non-essential businesses. Then, for three weeks, the state will return to the limitations on the size of social gatherings and closures on various public services, such as restaurants and bars, ending May 1. Schools were directed to implement distance learning plans through May 4. Additional physical distancing for vulnerable populations also will be likely, Walz said.
“That will buy us enough time that working with the Corps of Engineers we will be able to transform our arenas, our stadiums into hospitals,” Walz said. “We will be able to stockpile the PPE and get the ventilators into the system that raise that critical number of what hospital ICU capacity is, while keeping the number of infections low enough that when people go in and we’re reaching peak capacity somebody gets better and moves out.”
While these actions are projected to decrease the possible death toll by about a third, between 50,000 and 55,000 Minnesotans may still die from COVID-19, according to remarks made Friday by MDH health economist Stefan Gildemeister, who is part of the modeling team.
A moving target
While the models provided the impetus for the governor’s actions, one word looms large over every predictive coronavirus model right now: uncertainty. Kulasingam was confident about the model she helped to develop but said it would be constantly updated as new data becomes available.
“When the governor made the announcement, it was the worst-case scenarios,” Kulasingam said. “We’re taking very limited data from international sources, and then what we have is a series of math equations that say this is how we think the virus spreads through the population. It’s really important that people understand that these models can tell you in general what we think will happen with some kind of time frame but with a lot of uncertainty, which is why people are seeing big fluctuations in numbers, including our own.”
Unlike viruses Kulasingam has dealt with in the past, the new COVID-19 virus has no history to inform their modeling. Information and research changes daily, and with this new data the Minnesota model can be constantly refined.
“The biggest thing for us right now is that there is limited data coming out of the U.S.,” she said. “The initial data is coming out of China, and we’re not sure how applicable that is here. Even with the data from Europe some of the things we have to consider are will the dynamics be the same in our populations?”
This week provided one example of how data can change. On Feb. 28, the World Health Organization reported a fatality rate in China of 3.8 percent. This Monday, a more refined evaluation of the Chinese situation reported in the medical journal The Lancet put the fatality rate at 1.38 percent overall.
One piece of data working in favor of the Minnesota model, Kulasingam said, is a study that was conducted in Minnesota on comorbidity, the death rate when people have multiple medical issues at the same time. That information relates directly to data indicating that COVID-19 patients are at greater risk of dying if they have an underlying medical condition.
Still, for epidemiologists like Kulasingam and her colleagues, it’s an ongoing race to scour more data to answer important questions to make their models better.
“One of the unknowns for us when we started this process and that we’re hoping to get a better handle on is that we don’t know already what portion of the Minnesota population has been infected,” Kulasingam said. “We have a limited number of tests and those are being used for people who are being admitted.”
Antibody tests designed to detect if a person has already been infected are being developed by the U of M and Mayo Clinic to help answer that question.
Researchers also haven’t yet pinned down key questions about the virus itself.
“How fast does it spread?” Kulasingam said. “Does it infect two people for every one person that’s infected, or do they infect four people? How many people does one person infect?”
However, epidemiology isn’t just about disease. It’s also about how that disease operates within the social dynamics of a group of people, and here again is an unanswered question.
“One of the things that’s actually lacking is if you divide up the population by age, who has contacts with whom by age?” Kulasingam said. “Is it physical contact? Is it just talking to that person, are they walking by that person? When you start to put in these kind of scenarios, they’re really asking people to move away and to not interact.”
Kulasingam said the U of M is working with MDH on a study to fill that gap in which people will be interviewed about their interaction patterns on regular days and when people are staying in place.
One question the model isn’t refined enough to answer yet is what differences may exist between the spread of the virus in rural vs. urban areas.
“Would it play out differently?” Kulasingam said. “[Rural] people are really spread out, they’re more in their houses. Where are they congregating? Does that mean a delayed spread, so there’s just a longer lag between when it would occur there, or would it play out differently in that fewer people would be infected at all because there’s just a different pattern of how people go about their day? That’s something we’ve been asked about and that’s a next iteration, once we get more comfortable with the dynamics we’re seeing at the state level.”
And for the foreseeable future, Kulasingam and her colleagues will continue to research and refine the model to provide even better information to guide critical decision-making.
“The more data that comes out the better we will be at being able to narrow down our uncertainty,” she said. “The thing I would love people to know is that we’re doing our very best, and just to bear with us and to know that we’re all in this together.”