Covid Bombshells at the Minnesota Modeling Presser
On Friday, Minnesota unveiled a lot of details about its disease model for covid to the world. That press conference featured a number of bombshells, most of which seem to have received little or no follow-up from the media. (Are they distracted? Overworked? Are these revelations not interesting to the general public? Unclear.) I figured I should therefore blog about them, but I'm having a hard time figuring out how to do so efficiently.
Today, I'm giving up. I'm just going to list the revelations with no particular rhyme or reason.
(Here's the briefing, well worth watching: https://www.youtube.com/watch?v=iDqUjeIib-Q)
ITEM!
The "rule of thumb" for the past few months -- a pretty wild guess, frankly, and everyone knew it -- was that, for every 1 confirmed case of covid, there were 9 more going undetected. In other words, our covid detection rate was, at a guess, about 10%.
On Friday, Minnesota announced that it believes this rule of thumb is wrong. In fact, Minnesota believes that, for every 1 confirmed case of covid, there are closer to *99* cases going undetected -- a detection rate of 1%.
What this means:
We currently have 1,621 confirmed cases in a population of 5.64 million. (0.03% of the state confirmed infected)
Until recently, we believed that 1621 confirmed cases meant 16,210 actual cases. (0.3% of the state presumed infected)
But we now believe that 1621 confirmed cases means some much larger number -- something like 162,100 actual cases.
In their technical paper, the Minnesota team explained that they estimate a detection rate of 1%, with a possible range from 0.5% to 5%.
That's actually (mostly) really really good news. If lots and lots and lots of people have already been infected, that means two things:
The disease is less lethal than we thought. If 20,000 Minnesotans caught covid and 1000 died of it, that's a fatality rate of 5%. We could expect covid to kill hundreds of thousands of Minnesotans. But if there's actually 200,000 Minnesotans who've caught covid, and still only 1000 died of it, that's a fatality rate of only 0.5%, and we could expect deaths in the low tens of thousands. (These are numbers I pulled out of a hat to illustrate the point. You can't actually take current case/death counts and do napkin math with them very easily, because deaths lag cases by several weeks. But there's no other way to slice it: this is good news, and means fewer people will die.)
Meanwhile, if more people have had it, we're closer to developing herd immunity than we think. As I have argued strenuously elsewhere, herd immunity is not something we should aspire to, because it means consigning a lot of people to die. But it's reasonable to expect that we would acquire herd immunity if 75% of Minnesotans caught covid. That's about 4.1 million people. If 20,000 Minnesotans have had it already, we're still just a tiny fraction of the way to herd immunity. If 200,000 Minnesotans have had it, then we're already 5% of the way there. That means that the large peak Gov. Walz is expecting and planning for in summer will come a little faster and be a little lower. Not much lower, but it's something.
The bad part about the low detection rate is that it implies covid is more transmissible than previously believed. Let's talk about that.
ITEM!
Until the past week, most estimates of the basic reproduction number (or "r0") for covid have fallen in the low 2's. The famous (or infamous, depending who you talk to) Imperial College London report assumed that the r0 for covid was between 2.0 and 2.6, with a median assumption of 2.4. That means that, without intervention, the average person who catches covid will spread it to 2.4 people.
The purpose of interventions like social distancing and stay-at-home orders is to lower the effective reproduction number, or "rE", from that "natural" rate of r0 to something more manageable. As long as rE stays above 1, that means the average person who catches it spreads to more than 1 person, and the epidemic may be slowed, but will continue to grow. So we've been hoping, basically, that social distancing was lowering covid's reproduction number from something like 2.4 to something like 0.8.
The Minnesota modeling team no longer believes an r0 of 2.4 is plausible. Based on the latest data, which has pointed toward more infections and lower lethality than previously believed, they now believe r0 is at least 2.5, and possibly as high as 4.7 (!), with a median assumption of 3.9. That means the infectiousness of covid is less like the 2009 pandemic influenza (r0 = 1.4-1.6) and more like smallpox (whose r0 has been estimated in the range of 3.5-6).
That means covid may be much, much harder to suppress, test, and contain than previously hoped. Lowering r0 from 2.4 to 0.9 seems plausible, especially with half the state staying home all the time and going out only for groceries. But lowering r0 all the way from 3.9 (or 4.7!) down to 0.9 may be futile. I still think the alternative is so bad that it's worth devoting extraordinary resources to the attempt -- with huge disruption and damage to employees and businesses -- but seeing these numbers gives me more sympathy for Gov. Walz's decision to write off tens of thousands of Minnesotans as doomed to die and focus instead on keeping our hospitals from getting overwhelmed.
There's bad news here for mitigation, too. If the virus is more transmissible, it'll take us longer to reach herd immunity. The most naive formula for herd immunity is [(r0 - 1) / r0]. So, if r0 were 2.4, like we thought last week, then we wouldn't have herd immunity until (1.4/2.4=) ~58% of the population caught covid. If r0 is 3.9, then we don't acquire herd immunity until 74% of the population has had it -- which adds up to another 900,000 Minnesotans getting sick. This isn't bad enough to outweigh the very good news that the virus is less lethal than feared, but it still means -- unless we switch to a suppress, test, and contain strategy -- a lot more Minnesotans have to get sick before this goes away.
Just to put in perspective how big a deal this higher transmissibility is: back in January, on Facebook, I shared a Twitter thread by Dr. Eric Feigl-Ding where he cited a Chinese paper that suggested an r0 as high as 3.8. After he was roundly criticized by the entire mainstream media and much of the scientific establishment for way-over-the-top alarmism, I retracted my Facebook post and Dr. Feigl-Ding issued a correction to the thread. And now the State of Minnesota officially considers the r0 to be around 3.9.
To some extent, this is a measure of how much America wanted to stick its head in the sand about covid: at the time, the media wanted us focused on impeachment, which was reaching its climax at the time, and they accused all the marginal right-wing voices who were worried about the "Chinese coronavirus" as alarmists and xenophobes. And, to some extent, it was fair deuce: Dr. Ding's thread was way over-the-top in several other ways. But it also shows how much we have learned about this disease in just the last 90 days -- and how much those of us who aren't epidemiologists but wanted to keep tabs on the "Wuhan virus" (as it was called back then) were depending on the sketchiest of information mere weeks ago.
SIDEBAR!
There are still some people out there saying things like, "Half of us already had the coronavirus, so we're actually very close to herd immunity," or, "Lots of you had coronavirus in December and didn't realize it." They will often cite the so-called "Oxford model," which implied that herd immunity in the U.K. was imminent because only 1 in every 10,000 cases was being detected. Then they'll tell you about how they (or their godparents' roommates or whatever) had a really bad cough that just wouldn't go away a couple weeks before Christmas and wasn't the CDC saying the December flu season was worse than expected? Surely that was the first wave of the 'rona!
These people are fools.
We now know, with absolute certainty, that the so-called "Oxford model" is false. We know this because the U.K. has 88,621 confirmed covid cases. If the Oxford model were true, then there would currently be 886 million undetected cases of covid in the U.K., with another 500,000 being infected every day -- which represents a problem, because the total population of the U.K. is only 67 million. We are juuuuuust starting to get decent serology tests, and even the sunniest findings suggest that we're light-years away from the optimism of the "Oxford model."
We now know, with absolute certainty, that there were no cases of covid-19 in the United States before, at the extreme earliest possible date, 5 January 2020. We know this because we've decoded the genome of many different covid viruses from all over the world. Since viruses are constantly mutating very slightly, we've compared their mutations in order to create a kind of viral "family tree." Thanks to this "tree," we know, with a surprising degree of precision, where each strains of the virus came from and when.
The American infections came from many different sources, some in Asia, some in Europe. But our early infections all came from China, and we know that this particular strain of the disease did not leave China until at least 5 January 2020. There's simply no way anyone in the United States had the coronavirus before then. Even once it arrived here, it started in only a handful of people, mostly on the coasts, and then started to build. So the odds that you, some random suburban Minnesotan who hasn't been overseas in a decade, had the 'rona at any time in January, are not quite zero, but let's just say it's more likely that you were struck by lightning.
(This is another body blow to the "Oxford model," which implies that not only are we missing a lot of cases, but that covid arrived on our shores weeks earlier than could possibly be true.)
Minnesota is saying that it now believes that covid has been spreading more widely than previously expected, that it is less lethal than previously believed, and that we are closer to herd immunity than we thought. This is all excellent news. But don't confuse this with vindication for those who believe this will turn out no worse than a severe flu.
Minnesota usually has fewer than 100 flu deaths in a year, although we can lose as many as 500 in especially bad years.
Thanks to its higher lethality, higher threshold for herd immunity, and our complete lack of ability to create artificial herd immunity with vaccines (which keeps influenza contained to about 15% of the population every year), Minnesota expects to see not fewer than 9,000 deaths if the stay-at-home order is cancelled, expects 22,000 deaths, and fears 36,000. That's even with "commonsense" social distancing (keep high-risk people at home, don't have large gatherings) kept in place after stay-at-home ends.
That's a lot better than the 50,000 Minnseotan deaths we were talking about a week ago. The State has cut its projection in half. But it would have to cut its projection in half six more times to achieve the hundredfold reduction it would take for us to get down to the "just a flu" threshold. (Bear in mind, also, that Gov. Walz reports that the Mayo Clinic and other hospitals have their own models, and they are not as sunny as the state's.)
Feel free to hope for further downward adjustments, but don't count on them.
ITEM!
The State of Minnesota officially believes it will not have enough ICU beds to take care of all patients at peak, which it expects to arrive in June, July, or August. I can't believe this conference was so full of bombshells that this one only gets mentioned 1500 words in the article, because this, too, is a very big deal. Gov. Walz's strenuous planning effort (which deserves considerable praise) looks like it's going to be close to meeting our needs, but close is no cigar. We expect to need somewhere between 2,700 and 4,800 ICU beds. We expect to have only about 3,000 ICU beds available.
We could get lucky, and even our worst-case scenario doesn't look quite as bad as Italy -- but we're still operating on the expectation that our medical system will be overwhelmed, and that some people who need critical care are going to be left to die, as has happened elsewhere.
The good news for us epidemic hawks is that, unless this supply situation resolves itself in the next couple weeks, Gov. Walz will likely either extend the stay-at-home order further into May. Walz's whole strategy has been to acquire herd immunity while ensuring that every infected Minnesotan has a hospital bed to die in. Since he doesn't have the beds, yet, he may not have a choice but to buy more time for his supply chain. That also buys more time for our preferred solution (massive testing, contact tracing, surveillance, and targeted quarantines) to come online as an alternative to waiting for herd immunity.
ITEM!
Gov. Walz and the Minnesota modeling team are still refusing to model what would happen if covid were suppressed with an indefinite stay-at-home order until a vaccine, treatment, or test-and-contain strategy became available. The Minnesota team modeled five different scenarios, but the most aggressive of them merely extended the stay-at-home order to May 17th.
Walz and the modelers should change course and model indefinite stay-at-home.
First, it's just not a good thing to withhold data about a strategy that lots of Minnesotans are talking about and what to consider... even if it may not be the governor's preferred strategy
Second, the stupid-but-popular IHME / University of Washington model is based on the assumption of more-or-less permanent stay-at-home orders. When people look at the IHME model, they almost always ask why Minnesota expects 22,000 deaths when the IHME model expects only 656 Minnesotan deaths (at this writing). A very, very big part of that discrepancy is the fact that the Minnesota model assumes the stay-at-home orders are lifted before there's a treatment or widespread testing available -- May 17th, at the very latest -- while the IHME model assumes the opposite.
Another reason is that the IHME model is based on an "incredibly optimistic assumptions." Those are the exact words Stefan Gildemeister used in this briefing. There's a lot wrong with the IHME model, not just its poor communication. Gildemeister alludes to the unreliable data it depends on. I would point instead to the fundamentally questionable way the model operates: it doesn't use a real simulation of the epidemic based on what we actually know about it. Instead, it takes observed death counts and tries to fit those deaths to a mathematical curve, then guess where that curve is going next if it follows a normal distribution. (EDIT: We might also point to its horrendous performance, even on its core metric of death counts, to say nothing of its often-outrageous hospitalization projections.)
But it's still a very, very popular model, probably because it has such a slick interface. A lot of powerful Minnesotans who have absolutely no excuse are depending on IHME's projections to make sense of what's going on. Many of them are insisting that, because the IHME model projects so few deaths, we should immediately reopen the economy (which the IHME model says we absolutely must not do, or deaths will skyrocket).
Gov. Walz should model the same thing so that we can directly compare the models -- and so he can give lawmakers a clear choice between a prolonged economic closure and relatively few deaths or a shorter economic closure and 10,000-30,000 deaths.
ITEM!
We confirmed in this conference that the State of Minnesota believes we are currently cresting the "first peak" of the epidemic, and that cases (and deaths) will start to decline over the next couple of weeks. We expect very few total deaths at this stage -- well under 1,000, if I'm reading the chart right. That's actually fairly consistent with the IHME estimate.
However, as soon as we lift the stay-at-home order, currently scheduled for May 4th, we begin building toward a much higher "second peak", with (ultimately) somewhere between 9,000 and 36,000 Minnesotans dying.
(A friend asks: why would the second peak be higher than the first peak? Shouldn't the second peak be less bad, because more of the population will have immunity? I gave him a long answer, but the short answer is: the first peak was made artificially tiny by our massive stay-at-home interventions -- that was the whole point! -- and not nearly enough people will have immunity in the second peak to make a significant difference.)
I would have emphasized this "first peak" much more than the State did, since, as soon as case counts start to decline, the fools will start to crow about how the epidemic is over, it was all a huge overreaction, and we should chill out. The fools will be wrong, but it would be smart for Minnesota to get out ahead of their criticism.
So, yeah, extraordinary press conference, lots of huge bombshells.
The next day's headline was the "feud" between Sen. Gazelka and Gov. Walz, wherein Gazelka made an ass of himself. Because it's always about the personalities, not the reality. That's just the world we live in. Hopefully, your desire for more is why you're reading my blog.
Okay, gotta run for now. Questions or comments to combox or email. G'night!