Next step forward

Last week version 3.0 of coronavirus modeling for Minnesota was finally released, and the governor announced somewhat loosened restrictions.

I skimmed through the technical document for the model, and there are some interesting points, although it seems that the governor’s most recent order—beginning next Monday—was driven mostly by political considerations. One of them being having Wisconsin for a neighbor, where many things have been opened up, thanks to judicial intervention. A large fraction of Minnesotans live within easy driving distance of the Wisconsin border, and many were already accustomed to going to Wisconsin to buy fireworks banned by Minnesota. (Some of those fireworks are also illegal to fire off in Wisconsin, but can still be legally purchased there.)

So the governor is going to allow retail businesses to open, at half capacity and with other restrictions, and is also allowing small gatherings of up to ten people.

The modeling timeline is that they built the latest version of the model in April, took data from March 22 to April 25, did calibration and runs, and then stopped on May 1 to write up results, which were released on May 13. So data from the last two weeks was not included, perhaps for the better in terms of calibration, since recent weeks have been dominated by the spread of the virus in nursing homes, while the model does not yet have the capability of working with “hot spots”. Or co-morbidities. Or most combinations of interventions.

A couple of numbers from the technical document and slides keep appearing out in the media:  a 37.6% reduction in people’s social contacts under social distancing, and a 55.1% reduction in contacts under the stay-at-home order. The media portrays them as a failure of Minnesotans to achieve the government-planned reductions of 50% and 80%, but the numbers above simply did not come out of the real world, not at all. They came from the modelers’ attempts to tweak these and other parameters to make the model’s output match the data that they had.

One surprising thing I found in the technical document is that according to the Department of Health, 65% of the deaths attributed to the virus have happened outside of hospitals; at home or in a nursing home. Even the model predicts nearly 70% for this.

In an earlier post, I guessed a long-term care rate in Minnesota of 1.5%. It is actually under 1%. If I put that together with the dying-out-of-hospital rate, that suggests that over 6,000 Minnesotans are going to die at home over the course of the pandemic.

The one very definite indication from the scenarios that were modeled is that Minnesota should not follow the CDC guidelines for re-opening, which according to the model would delay the peak in coronavirus cases for several months, and then spread it out so widely that half of the 2200 ICU beds with ventilators that the state has arranged for would never be used.

Under more realistic conditions, the virus is expected to peak in Minnesota in late June or early July.