“If it looks like you’re overreacting,” Dr. Anthony Fauci said in March about the developing COVID-19 response, “you’re probably doing the right thing.”
So how long before people think you’re overreacting (answer: not long), and how do you remove the “probably” from “doing the right thing”?
In lieu of certainty, you project. You build statistical models with whatever data you can acquire and update them as better information comes along. Such models aren’t definitive but can guide preparations.
We see something similar with weather forecasts — the splay of spaghetti lines on a map that converge as a hurricane approaches landfall, or disagreements among models over how much snow will fall where, as was the case leading up to Minnesota’s most recent April affront.
The new strain of coronavirus has a lower fatality rate than those that produced SARS or MERS, but it’s also harder to contain. It’s deadlier than the seasonal flu by an order of magnitude (1.4% by one estimate, compared with 0.1%) and more ecumenical in its targets. Some who are infected have minor symptoms; others, none, but each must entertain the possibility of a terrible outcome. Thus, economy-crushing restrictions have been put in place to impede the disease. Doubts about these actions and the models that guide them are to be expected. With all that in mind, let’s look at the basics.
What model is Minnesota using?
The Minnesota model, now on its second iteration, was developed by the University of Minnesota School of Public Health and the Minnesota Department of Health. It combines data from elsewhere with information specific to the state. Some places are using similarly unique models, while others seek consensus from a range. Is the first or second approach better? Yes.
Ideally, the Centers for Disease Control and Prevention or the National Institutes of Health would have provided an official national model, but that’s now a task for the future. It’s too late this time.
Minnesota officials have emphasized that they pay attention to other models, including projections by hospitals that have been more pessimistic.
How does the Minnesota model work?
Models require inputs — both known data and assumptions. This quickly gets complicated, but here are some of the factors that have evolved enough to bring worst-case projections down:
• R0 (pronounced “R-naught”): The average number of people who will catch a disease from an infected person. The assumption of 3.87 for COVID-19 has increased as it’s become clear that some people carrying the virus do not have symptoms, suggesting that we’re somewhat closer to an initial peak than previously thought. (R0 for the seasonal flu is 1.3. A disease begins to die out when each person who has it infects fewer than one other person.)
• The number of days patients spent in the hospital (13.3) and the ICU if necessary (10.3). These shorter-than-anticipated stays are good news for capacity.
• Mitigation: Projections consider several scenarios for social distancing over varying periods, with reduction in contacts ranging from 20% to 80%.
All of this converges on a primary goal: delaying the peak number of concurrent infections in Minnesota to allow time to build health care capacity. The lack of an ICU bed and/or ventilator when needed is a key factor for mortality, increasing the risk by anywhere from 1.5 to 16.5 times, depending on a patient’s age. The modeling estimates that the state can have 2,200 ICU beds for COVID-19 patients specifically by the time infections peak in June or July, still below but closer to the anticipated demand.
Minnesota has made its model quite transparent. Go to mn.gov/covid19/data/modeling.jsp and see for yourself.
But aren’t state officials projecting an implausible number of deaths?
Under the modeling scenario most closely resembling current practices in Minnesota, 22,000 deaths are projected (in a range of 9,000 to 36,000). How could that be possible even when the much more densely populated New York City, in the crucible over the last month, has had just 11,477 so far?
It’s important to note that Minnesota’s projections are for the duration of the outbreak — 12 to 18 months. State officials also are concerned about placing too much emphasis on specific numbers as opposed to trends.
Another model, from the Institute for Health Metrics and Evaluation at the University of Washington, has been shrinking its estimates and as of Friday predicts 195 deaths in Minnesota, from a range of 95 to 605. (There have been 111 here so far.) The IHME projects only through Aug. 4 and predicts that Minnesota already has reached peak use of health care resources. It does emphasize: “For hospital administrators and government officials, it is important to pay attention to the full range of values in our forecast, especially the upper values.”
If you suspect the state of sandbagging, a question: In personally stocking food and supplies, have you erred toward worst- or best-case scenarios?
Wouldn’t it be better just to get it over with and develop herd immunity?
This refers to the point at which a sufficient number of people have antibodies preventing them from contracting or transmitting the disease. Two ways to reach it: vaccination or exposure. A COVID-19 vaccine is at least a year away, and the strength of immunity for those who have overcome exposure is undetermined. Between 55% and 80% of the population would need to be inhospitable to the virus if it is to begin to die off. Even with asymptomatic cases, former FDA Commissioner Scott Gottlieb thinks that a “small percentage of the population — certainly in the single digits — [currently has] the level of antibodies needed for immunity.”
Inherent in doubts about COVID-19 projections is the notion that we can — and for the sake of the economy must — relax restrictions. In finding a balance, economic modeling would be as valuable as the epidemiological kind. But for the moment, let’s set aside the variables and just do some crude math about letting loose: Minnesota has 5.6 million people. To reach even the low-end estimate of 55% for herd immunity, 3.1 million people must be exposed. If the 1.4% fatality rate endures, we’d lose 43,000 souls.