Sunday, March 29, 2020

Coverging Crises Part I.1: Covid Encouragement from University of Washington Modeling (Updated)

This prepublication paper (March 25th) uses newer data than that of the Ferguson et al. paper (March 16th) that I drilled into last week.  It also focuses on the U.S. situation, broken out by state. And the executive branch seems to be using it to design shutdown policy. The headlines:

The University of Washington paper projects 81,114 deaths, with a very large Uncertainty Interval (UI) belying the precision of those numbers (at left).  This is in the ballpark of the lower bound of Anthony Fauci's estimate of 100,000-200,000 deaths. These are high numbers.  The bottom is about  3 times typical death counts from influenza.  But they are 20 times lower than the "doing nothing" estimate in the Ferguson et al. paper. The two papers agree on the scale of reduction that suppression measures achieve.

(UPDATE: This number of total fatalities was revised to 93,765 on April 2nd. It was revised downward in the wake of sustained social distancing measures, and has stood at around 60,400 since April 9th).

The authors also modeled health care capacity, estimating demand for hospital beds peaking at 7 percent above the national bed count, and Intensive Care Unit demand peaking at 25 percent above the national ceiling.  This means that suppression measures do not eliminate the health system crisis (this looked better in the Ferguson et al's projections for the UK).

The  UW team couldn't estimate ventilator shortages because they couldn't find reliable counts for these. Bed and ICU shortfalls are expected to vary greatly by state, with "peak excess demand" for beds being particularly bad in New York, New Jersey, Connecticut, and Michigan, and excess demand for ICUs being more widely spread across the country.  California is in comparatively decent shape on both fronts, though that isn't saying much.  Excess ICU demand is figured below.



All the estimates (reduced deaths, very serious but perhaps non-catastrophic excess hospital demand) assume social distancing that approximates Ferguson et al.'s suppression regime.  I find this to be the least convincing aspect of this paper.
For states that have not implemented 3 of 4 measures (school closures, closing non-essential services, shelter-in-place, and major travel restrictions), we have assumed that they will be implemented within 7 days, given the rapid adoption of these measures in nearly all states. At this point in the epidemic, we have had to make arbitrary assumptions in our model on the equivalency between implementing 1, 2, or 3 measures – and we have implicitly assumed that implementing 3 of 4 measures will be enough to follow a trajectory similar to Wuhan – but it is plausible that it requires all 4 measures. (8) 
I see two obvious problems with this. First, we are not seeing "the rapid adoption of [suppression] measures in nearly all states."  Duringthis past week, many Republicans politicized social distancing as anti-business.  Although some Republican governors have not followed this line, Covid suppression has now been polluted by the country's toxic political discourse. POTUS is angling for ways to pin an extended shutdown (anything past Easter Sunday on April 12) on the libs, and his proposal for a NY-NJ-CT quarantine didn't have to be implemented to tar the country's leading Democrat stronghold as a disease-carrying Gommorah. Florida governor Ron de Santis has set up border checkpoints partly aimed at excluding New Yorkers; at the same time, Rhode Island's Democrat governor is proposing door-to-door searches for infected New Yorkers.  The American political system may not be good at public health, but it is world-class at finger-pointing. This does not bode well for a national suppression regime and its 20x reduction in mortality.

Second, I see no reason to assume, as the authors do, that "implementing 3 of 4 measures will be enough to follow a trajectory similar to Wuhan."  That is the best-case trajectory, in which full shutdown in  Wuhan on January 23rd led to no new cases by March 15th.  (That included a peak rate of daily deaths 27 days after that shutdown date.)  Wuhan implemented all 4 of 4 measures and then some--Wuhan authorities also instituted an app that used personal data to restrict movement for people rated "red" or "yellow," and broke disease clusters through family separations, in which some members were sent to medical dormitories for isolation.

In other words, the UW model may be way too optimistic--landing us back in a mitigation model which halves deaths (to about a million in the U.S.) while overwhelming the medical system.

This study notes but, as far as I can tell, ignores two other major differences between Wuhan (and South Korea) on the one hand and the U.S. on the other.  One is our lack of mass testing, including testing of people without Covid-19 symptoms.  Testing allows the health care system to identify people who need total isolation and/or treatment, making social distancing much more efficient.  The U.S. seems to have missed the crucial testing window (the New York Times' investigation of the serious failure of Trump's executive branch is worth reading in full). The other is China's massive mobilization of equipment and facilities--the new hospital constructed in two weeks and the like.  In stark contrast, the U.S. story is of shortages--of ventilators, of masks and gowns, of swabs, and, soon, of trained and healthy medical personnel.  All this also casts doubt on our powers of suppression. We will have to fall back on brute isolation, which is of course is the longest and the most costly mode economically--and educationally.

The paper ends on this note:
Our estimate of 81 thousand deaths in the US over the next 4 months is an alarming number, but this number could be substantially higher if excess demand for health system resources is not addressed and if social distancing policies are not vigorously implemented and enforced across all states.
We are thus encouraged to continue the most stringent version of a difficult lockdown with the clear possibility that the disease itself will be much less lethal than we have been assuming.

I do take this as encouragement.  We could start seeing a real ebbing of fatalities and infections in the Wuhan 60-day period: by mid-May in the New York region, and not too much later elsewhere. (See Bryan Alexander's Three Scenarios for other possibilities.)
Stay strong--indoors.

April  3rd.   The Washington Post has a big piece entitled, "Experts and Trump's advisers doubt White House's 240,000 coronavirus deaths estimate."  There's nothing new that we haven't covered here, but it's interesting to note the ongoing refusal of the White House "to explain how they generated the figure" of 100,000-240,000 national Covid-19 deaths, beyond the mashup of the two studies we've analyzed here - Ferguson et al. at Imperial College and the UW model discussed here, from the Institute for Health Metrics and Evaluation, with its user-friendly projections, one of which appeared in Deborah Brix's briefing.   "But what remains unclear and alarming to many modelers is whether the White House is using their data to create a coordinated, coherent long-term strategy." The answer is no they're not, for example, implementing a national stay-at-home order called for by Anthony Fauci.

April 5th Covid-19 continues to receive saturation coverage, but the single most interesting piece to me was NYT coverage of the German response to the virus. So far, Germany's case-mortality rate is 1.4 percent, nearly a tenth that of Italy and about half that of the United States. Some reasons why: the German infections "started as an epidemic of skiiers" coming back from Italy and Austria, so younger people were infected first. Second, Germany has run many more tests, so they capture a higher share of infections than Italy or the U.S. (their denominator is bigger). The piece goes through other crucial factors:"early and widespread testing and treatment, plenty of intensive care beds and a trusted government whose social distancing guidelines are widely observed."  The U.S. can't really draw on any of these.

I thought ruefully of UC Health when I read this passage:
Before the coronavirus pandemic swept across Germany, University Hospital in Giessen had 173 intensive care beds equipped with ventilators. In recent weeks, the hospital scrambled to create an additional 40 beds and increased the staff that was on standby to work in intensive care by as much as 50 percent.
“We have so much capacity now we are accepting patients from Italy, Spain and France,” said Prof. Susanne Herold, the head of infectiology and a lung specialist at the hospital who has overseen the restructuring. “We are very strong in the intensive care area.”
All across Germany, hospitals have expanded their intensive care capacities. And they started from a high level. In January, Germany had some 28,000 intensive care beds equipped with ventilators, or 34 per 100,000 people. By comparison, that rate is 12 in Italy and 7 in the Netherlands. By now, there are 40,000 intensive care beds available in Germany.
Separate item: People say the infection numbers are bad but the death counts are good.  A WaPo piece reminds us that they're bad too.  Covid-19  caused deaths are likely undercounted worldwide, perhaps by a lot.
 
April 12th

Today is Easter Sunday, which POTUS wanted to be America is Open for Business Day.  It isn't.  For example, Kansas Gov. Laura Kelly's executive order to close churches was opposed by Republican legislative leaders, but yesterday they lost before the Kansas Supreme Court.

Yesterday was also Peak Resource Use in the UW (or "Chris Murray") model I've discussed here.  Peak fatalities were to occur April 10th.  Early infection zones--Washington State, New York, and California--are all heading in the right direction.  Other areas (New Orleans, Detroit) are not yet.  The South remains a wild card.  Here in Santa Barbara County, we had 264 cases as of 5 pm on April 11th, with two deaths, a number that has held steady for a week.  Many of the new infections are in the Federal Prison in Lompoc.

Some states have released some incarcerated people in response to the pandemic (like Gov. Tom Wolf of Pennsylvania).  The Federal Bureau of Prisons is screening prisoners to release some into home confinement.

California is going to be closed at least through the end of the month, and LA County among others through mid-May.  But the debate on openings is heating up in Europe.

April 17th

STAT has an interesting overview of the criticisms of the methodology of the University of Washington model covered here.

No comments:

Post a Comment

Note: Firefox is occasionally incompatible with our comments section. We apologize for the inconvenience.