Minnesota yesterday released a variety of information relating to the model used to support the shutdown order in the state. The release occurred following a commentary I wrote that was published in the Star Tribune on Friday and pressure from a variety of other sources for more transparency. The materials are available here (Mn. Model) and if you click on the links at the bottom of that page you will see a video of the briefing, the technical paper supporting the model, a powerpoint discussing the scenarios run and other information. The state has promised to release the actual code underlying the model but not for a couple of weeks, which is strange, because apparently R was used so they should be able to release it now.
The construction of the model is good, I think it takes a reasonable approach. The primary author is appropriately disdainful of the IMHE model, which has been so widely used, and so far off-base, and I would agree that the approach the Minnesota modelers took is better. But like all models it is only as good as the accuracy and completeness of the information available to support assumptions. So a number of the assumptions they have made continue to look very flawed. The most important is that they understate the ability to create additional ICU beds quite easily, as has happened in NYC, and to adjust ICU capacity accordingly. People who haven’t worked in a hospital probably don’t understand that an ICU isn’t some mysteriously high-tech medical space, it really is just taking a typical ward and adding some additional monitoring equipment and increasing staff levels. It is fairly simple to convert existing wards in a hospital to additional “ICU” beds. If access to a ventilator is the issue, you don’t need to be in an ICU to be on a ventilator. Almost all the deaths in the scenarios the Minnesota modelers run result from supposed excess ICU need over capacity, which simply isn’t the case.
The other errors in assumptions, which results in apparent over-run of ICU capacity, leading to all those deaths, relate to the percent of people who get infected and have serious illnesses, the mortality rate, and a failure to account for the likelihood that the most susceptible become infected, seriously ill, and die at the earliest stages of the epidemic, and the rate of deaths among this group, or any sub-group, does not persist at the same rate throughout the course of the epidemic. An appropriate mortality rate and an accurate assessment of hospital resource would lead to far lower projections of deaths. And please note that since the initial run of the model, projected deaths dropped by over 50%. This just shows how wrong it is to use these models for decision-making in the first place.
What they did do that is very useful, is to run various mitigation strategy scenarios. That is exactly the approach that should be taken. How else do you know what the relative benefits and harms will be of different sets of mitigation of spread measures? So in the scenarios the modelers run, you get basically the exact same number of deaths regardless of the mitigation strategy. My favored approach, use basic mitigation for the bulk of the population, but isolate the at-risk groups, has the same outcome as making everyone stay at home. This would allow the working age population to go back to work. The economic harms due to job loss and other non-economic harms flowing from job loss would be substantially reduced. Regular readers know that I have repeatedly made the point that analysis of mitigation of spread measures needs to be viewed incrementally, so that we understand the benefits of each in terms of illness and deaths deferred (again, not saved, the virus isn’t disappearing), and we can measure the comparative economic and non-economic harms of each possible mitigation strategy.
And that is the other part of the story, that is being ignored by the Minnesota press. The Governor did no analysis of the economic, job loss or other harms that might flow from the actions he took. I am flabbergasted to say the least. I simply can’t comprehend not wanting to have information about that, for planning purposes at a minimum. I have tried to think of a good analogy, and the best I can come up with is you are the lead bull for a large herd of cattle. Your herd is spooked by some threat and you all take off running, chased by the threat. There is dropoff ahead and you think if you go over that, maybe the threat will be lessened, but you don’t know how far you will fall and how many of the cattle might die or be injured in the fall. You have pictures available that if you looked at them would tell you what might happen in the fall. You just go ahead and run off the cliff. And, oh by the way, the threat is following you right off the cliff and is still present.
The Governor just took us off the cliff. He didn’t have to, he could have used a different strategy that would have incurred far less economic damage and the same amount of coronavirus damage, according to his own modeling. But he didn’t even bother to consider alternatives. That is dreadful, dreadful decision-making and judgment.