The U of Washington/IHME model (
covid19.healthdata.org) was updated again and now predicts 1,488 deaths for Iowa by August, up from 1,367 yesterday.
Since Trump and Reynolds have now both cited this model, I wanted to read more about its assumptions. This is a long post, but the short version is that Reynolds's claims that the model's assumptions are wrong are...let's be generous and say "misinformed."
Reynolds's comments on the model from today's press conference
are in this video from 27:35 to 29:37 (
Register coverage here). They downplay the high death count on the grounds that the model's assumptions do not reflect the mitigation efforts we have been undertaking. This is an argument people here have made as well.
Not surprisingly, Reynolds didn't mention that the model assumes all of those mitigation efforts are in full effect within seven days. Whether she didn't know that (a scary thought) or she's just intentionally trying to mislead is anyone's guess. Regardless, in reality the model assumes a fairly small window where the state is not exercising mitigation.
This tweet explains why the projections took such a big jump, from 777 two days ago to 1,367 yesterday. They just simply didn't have the data they needed (or enough data) before. So it appears not to have been a jump at all, so much as previous estimates having been artificially low.
The assumptions built into the model include whether a state has 1) implemented a stay at home order, 2) closed schools, 3) closed non-essential services, and 4) severely limited travel. Reynolds would say we've basically done three of those four things.
But the model's assumptions are based on the
New Zealand government alert system, level 4, and it's clear we aren't anywhere near that level. We have no stay at home order with any teeth to it (I'm sorry, but the governor repeatedly begging people does not count when she openly admits she can do more and won't yet). We have not closed nearly enough businesses. And we are doing nothing to restrict travel. The one thing the model's assumptions really does get wrong is that we've closed schools - though oddly even in that instance Reynolds only recommended, not mandated, school closures and thankfully schools complied.
Further, because deaths
like this one are not counted in state/official statistics, deaths and hospitalizations are likely undercounted even in the model's data.
The one weakness I can pick up on in these assumptions is that they seem to be a binary choice - a state has either closed almost all businesses, or they all remain open - and we're somewhere in between so the model can't account for that. But again, even those murky conditions only exist for seven days before the model assumes all mitigation efforts are strictly observed.
Reynolds's total failure to understand data, along with those ridiculous metrics she keeps talking about - this story from the
Iowa City Press-Citizen details it further - are guiding her decisions. Nobody should be reassured that she knows what she's doing.