• Home
  • About Us
  • Guest Posts

Sunday, March 22, 2020

Sunday, March 22, 2020

Coverging Crises Part I: Covid Shutdown Theory (Updated)

Shutdowns are now spreading as fast as the coronavirus. On March 19, Gov. Gavin Newsom ordered 40 million Californians to stay home, claiming that the infection rate puts the state on track for 25.5 million infections.  The order has no end date.  New York and other states and counties have since followed suit: by noon on March 21st, 75 million US residents were under some kind of lockdown.

In this post I'm going to talk about what I've learned during a sustained effort to apply analytical expertise to a topic outside of my normal subject areas, as I try to build a base for a series of citizen judgements about health policy, and also the related areas of educational and economic policy that I know more about.

This learning process has changed my mind about a number of Covid-related issues: for example, when I learned March 10th of UCSB's shutdown--at the end of my senior seminar, thanks to Jenna, multitasking on her email again!--I was a skeptic about the benefits of widespread closures. Now I'm a believer: I think that widespread social distancing is our only chance to avoid levels of infection that would overwhelm hospitals and clinics and lead to much excess death.  At the same time, I'm also more optimistic about reducing infections than I was a week ago.

The main part of this post close-reads the one published infection model that I've been able to find-Neil Ferguson et al.'s paper, from Imperial College London.  The U.S. Centers for Disease Control and Prevention (CDC) has not released its modeling, though it was discussed in a bootleg version by the New York Times.  My caveat up front is that the SARS-CoV-2 infection model I analyze does not offer any certainty about the future. But I will talk about the powers of the suppression regime we've entered into, and how the disease might be made less deadly than many of us now assume.

An overview:
  • The policy of virus suppression does appear to reduce Covid-19's spread. I'll define this and other terms below, since terminology is all over the place in media reports. (The one journalist I've found to have interviewed Neil Ferguson--Nicholas Kristof of the New York Times--conflates mitigation and suppression.)  Suppression has worked well in South Korea, Singapore, and post-lockdown Hubei in China when social distancing is combined with mass testing. 
  • The U.S. simply does not have the testing capability to do the most effective form of suppression.  (Santa Barbara County has brilliant and frequently exercised emergency services.  As of March 22nd it has 13 confirmed Covid-19 cases, a shortage of test kits, and 200 tests out whose results won't be in for awhile.) The U.S. has not been able to do contact-tracing, which would have allowed a much more efficient form of isolation than the mass version we're doing now.  In spite of some encouraging reports of new equipment coming on line, the U.S. is in the midst of what statistician John A. Ioannidis calls an "evidence fiasco," and its public health capacities have been downsized (personnel down 20 percent since 2008, according to David Himmelstein) to the point that we're likely stuck with the crudest, most disruptive, and most economically damaging form of suppression. 
  • This has implications for rebuilding social and public capabilities that I'll save for a later post on how SARS-CoV-2 is putting neoliberalism out of its misery--and how to keep that from causing further misery for diverse publics.
  • A theory point: public officials are using projections of high infection and death rates to install suppression regimes, but these suppression regimes are designed to invalidate the numbers that justify them (by producing much lower rates of infection and death).  Either you infect 81 percent of California by doing nothing, or you lockdown California and get a much lower infection percentage.  You don't do both.  I elaborate on this point because it's important for people not to think lockdown = death (regardless), but to think the opposite.  
  • A policy point: public officials must not bullshit the public with exaggerated numbers, withheld models (CDC!), and mashup policies that will encourage cheating. Newsom did the right thing, but he didn't give clear, honest reasons for it.  That has to change.
To take the last point first: Where did Newsom get the number that he used to shut down most of the state economy without an end date? We don't actually know. The LA Times reports, "the governor’s office declined to provide an explanation of the state’s projection that 25.5 million Californians will be infected with this virus. Instead, a spokesman for the governor said the state’s mitigation efforts could lower that estimate."

The last part is true (though "mitigation" is the wrong word, as I'll explain), but the public should be told the source.  In the meantime, I'll guess that Newsom's people got that number from the now-famous pandemic modeling paper I mentioned at the top, Ferguson et al. Their baseline reproduction number (Ro) for the disease is 2.4--meaning each case typically goes on to infect 2.4 other people. You can get to 25.5 million Covid-19 infections by taking California's Covid infection count when Newsom spoke--around 1000--and giving it an exponent of 2.45.  (Updated: See Akos Rona-Tas's correction of this speculation below, under March 23.)

The Ferguson paper derived that Ro in part from from the spread of the virus in Wuhan, China, before the government began its many non-pharmaceutical interventions (NPIs)--forced quarantining, widespread testing, etc.  (Wuhan's Ro was previously reported as 3.11).  The projection that 56 percent of the California population will become infected appeared as a math error in Newsom's letter to Trump requesting a hospital ship: it's actually 64 percent, or alternately, 39.56 million Californians * 0.56 = 22.15 million inflections.  The point isn't the bad math but the need to offer credible numbers and explain clearly where they come from.  People will take honest, fully disclosed estimates more seriously.  Health policy needs to be open to establish the trust that government now desperately needs, to discourage cheating, and to allow meaningful democratic judgment about overall policy. 

Public officials, including Newsom, seem to be now focused on using big numbers to stampede the masses into social distancing, RTFN. This is understandable, since, in the suppression arsenal, social distancing is pretty much all we've got.  But one major effect of their statements is to muddle the difference between mitigating and suppressing a pandemics: the former allows infection rates like 55 percent. The latter slows growth rates and can put them into reverse.  Suppression also requires a rigor that people won't pursue if they don't understand the massive difference it can make.

To put this in the form of a question, could the U.S. and the European Union (and other regions) achieve suppression and thus decline in the number of new cases?  The current tracking in California is not good.



But look at  the South Korean case pattern.



South Korea had our hockey stick and has now bent it down into slower growth of new cases.  As is now widely discussed, South Korea, Singapore, Hong Kong, Taiwan, and now Wuhan have slowed the spread.  This is the effect of suppression strategies.  There's some important news here, which is that Covid-19 infections rates can be reduced, and its case-mortality rate can be kept low (not the 3.4 percent reported by the World Health Organization, but about 1 percent in South Korea, or 0.54 percent for cases under age 60).  Germany currently has a 0.3 percent case-mortality rate. SARS-CoV-2 kills people by doing horrible damage to their lungs (see the images around 0'30" in this Santa Barbara Cottage Health grand rounds lecture).  And yet the virus does so little to so many other victims that 86 percent of cases in China were undocumented prior to travel restrictions. 

On to the model: the Ferguson et al. paper draws on previous work with influenza pandemics to compare three responses-- doing nothing, mitigation, and suppression.  Doing nothing seems to have been the preferred option of the Boris Johnson and Donald Trump governments until about March 15th-16th  (Johnson, Trump), with the Johnson government allegedly working on a trust that infection would create "herd immunity" without disrupting the economy.  At least in the UK, they seem to have taken on board the Ferguson et al. calculations that "doing nothing" will lead to infection in 81 percent of the population (at  2.4 Ro), producing 510,000 deaths in the UK, plus 2.2 million deaths in the United States, both over a 2 year period.

With doing nothing now ruled out, the alternatives that Ferguson et al. modeled are mitigation or suppression. Suppression is China after January 23rd and South Korea, among others; Britain is moving to suppression with one escalating announcement after another (which may defeat the purpose).  Some parts of the U.S. are now doing suppression, including New York and California. The Ferguson paper divides these two strategies into two groups of non-pharmaceutical interventions (NPIs).

 The most effective set of mitigation measures are:
  • Case isolation in the home (CI): symptomatic cases stay at home for 7 days.
  • Voluntary home quarantine (HQ): all members of a household with a case stay home for 14 days
  • Social distancing of those over age 70 (SDO).
Note that this falls short of "lockdown," which includes social distancing for the whole population (SD) and, in most cases, closures of schools and universities.

Mitigation is the famous "flattening the curve." The serious cases that need hospital services are pushed out over time, with the goal of relieving some of the stress on the health care system. Mitigation is "predicted to reduce peak critical care demand by two-thirds and halve the number of deaths" (8).  Assuming the ratio of infections to critical care cases is constant, and that the syntax means mitigation yields 2/3rds of the "do nothing" infection rate, this leads to 54 percent of the population being infected, and to 1.1 million deaths in the U.S.  (When Kristof quotes Ferguson saying his best case is 1.1 million deaths, I think he ran Ferguson et al.'s two regimes together: in my view, the sentence should read, "his best case for mitigation" is 1.1 million deaths.

Clearly mitigation isn't good enough.  A million deaths in the U.S. is unacceptable, and the model suggests that under mitigation health care systems will still be overwhelmed (10). Since something like Italy's hospital crisis and high fatalities are the combination everyone wants to avoid, the UK, the EU, California, and now several other U.S. states have moved into suppression.

A side note: I would normally read the quotation to mean that mitigation reduces peak care demand (and infections) by 2/3rds, down to 1/3rd of their previous level, which is a 27 percent infection rate.  I don't know if that's what Ferguson et al. meant, but it's still more than double this year's seasonal flu rate (so far this season, flu has killed 22,000 Americans). 

Much of the U.S. is now following Italy, France, Spain, and other countries into suppression. The key benefit is that it reduces the reproduction number (Ro) to close to 1 or below, which China has shown is feasible.  Here's a nice stretch goal for the West.



 In the Ferguson et al. model, suppression adds to mitigation's measures:
  • school and university closures (PC)
  • social distancing expanded to the whole population (SD)
I've reproduced the table that shows the results. I'd recommend starting in column 1 with the baseline Ro of 2.4 (510,000 "do nothing" deaths) and look at the medium case of 200 (which means that the full suppression program is suspended when ICU cases fall below 200 in Great Britain, and are re-engaged when they rise above that number). (The paper does not have a similar table for the U.S.) 
California is now doing the full suppression program.  If you look at the right-hand column under Total Deaths you can see the results.  Deaths in Great Britain drop from 510,000 to 24,000, or by a factor of around 20.  The U.S. equivalent would be 110,000 deaths, not Kristof's 1.1 million.

Note two other features of this model.  The interventions all have finite periods: mitigation is modeled over 3 months (to mid-June 2020) and suppression over 5 months (to mid-August 2020).  They don't extend to the full 18 month "vaccine" period, nor are they open-ended.

Second, they are adjusted according to thresholds of infection and hospitalization that can be selected and monitored.  Governments have a great deal of agency here.  In other words, this new coronavirus is bad, but it is not an irresistible event like a giant asteroid hitting the earth.

A big catch is that the versions of suppression in South Korea, Taiwain, Hong Kong, Singapore, and China include mass testing.  Neither the US nor the UK have done this, nor do we seen to have the capability to ramp this up.  There's been much excoriating commentary on this point.  I had been hoping that UC Health could make a big difference to California public health. A potentially exciting March 14th headline, "UC has a solution to the national shortage of coronavirus testing," didn't, with our weak public sector, mean UC is gearing up mass testing for the public, but that it has a private test for its own patients.  I've heard ambitious UC plans--in this week's board meetings, one UC regent suggested for the installation of MASH hospitals on empty land that UC owns. But because of testing and equipment shortages, UC medical centers have to focus on protecting themselves (see 0'44"-0'49" or so in this very useful UCSF infectious diseases division' grand rounds). I'll end by adding a few items to the summary list above:
  • The virus is going to be terrible for public health workers, who deserve not only massive sympathy and support but also personal protective equipment, which they may now have more hope of getting.  Mass testing also depends on cranking out PPE.
  • Public health interventions in Asia have had enough success with suppression to give  credibility to the Imperial College model--most interestingly, its suggestion that deaths can be reduced by an order of magnitude. 
  • On the other hand, hospital access remains a potential catastrophe.  Full suppression reduces ICU need to 1/3rd of "doing nothing."  In the bootlegged C.D.C.’s scenarios, "2.4 million to 21 million people in the United States could require hospitalization, potentially crushing the nation’s medical system, which has only about 925,000 staffed hospital beds. Fewer than a tenth of those are for people who are critically ill."
  • Still, suppression seems to make a big difference even if it is leaky: the Ferguson et al modeling assumed incomplete success and still got major reductions (see Table 2 on page 6).
  • The US has a weak health system (or no health "system" at all, as Robert Reich rightly observes). This is a big problem. But the US has some other advantages: a lot of really good, dedicated health personnel, lower population density than Europe's or East Asia's and, ironically, dependency on the self-isolating feature of private cars.  Our version of suppression might be more successful than we now expect.
  • Officials should give expiration dates to the current suppression regimes. They can be extended later, depending on conditions.  As I noted, the Ferguson et al. model assumes a kind of regular adjusting depending on infection numbers. (Hong Kong has reimposed quarantine and testing on arrivals after an uptick in cases.) Indefinite lockdowns are bad for both people and the economy.  Once people are scared indoors, and the infection curve is bent like South Korea's, governments should throw the lockdown into partial reverse, lest they create another Great Depression x 2.4.
I'll move on to political, economic, and university dimensions in other posts.  From the Haley Street Bunker: stay well, and keep your distance! 
Monday March 23rd

Statistical chemist Michael Levitt hammers on one of this post's key points: "The virus can grow exponentially only when it is undetected and no one is acting to control it."  The media, he says, should focus not on total number of cumulative cases but on rates of growth of new cases. 
Speaking of which,  South Korea's number dropped again, so the chart looks a bit better today.

The coming U.S. health crisis will owe much to a social system that can't anticipate non-market public needs.   That's not what this WaPo piece says in so many words, but it has all the raw material--shortages of masks, gowns, tests, ventilators.  What aren't we short of Covid-wise?

This piece, by a Mass General physician, specifies how the market power of large hospitals will mal-distribute emergency equipment: "We are currently taking an every-hospital-system-for-themselves approach, in which some hospitals will surely say “we’ll take them all” while others will lack the capital to make such large purchases in advance and therefore will be reliant on FEMA, which will be forced to ration scarce, lifesaving equipment. These already cash-strapped hospitals serving poorer populations will soon be put in even greater jeopardy.

From Akos Ronas-Tas (Prof of Sociology, UC San Diego): How Newsom got his numbers (over half of Californians being infected) is a mystery, but it is surely not by raising 1000 to the power of 2.45. I am no epidemiologist either, but the Ro produces an estimate only if you specify how many generations of infections you count. So if the base (generation 0) is 1000 and Ro is 2.4 (used by Ferguson), the first generation will be 1000*2.4=2400, the second generation 2400*2.40 = 5760 and so on. The total number infected will be by then 1000+2400+5760=9160, adding up generations 0,1 and 2. In the Ferguson paper they use a 6.5 day generation time. The key here is that Newsom made his prediction for 8 weeks out. So he is counting roughly 8 generations. The number of newly infected in the 8th generation will be 1000*2.4^8=1,100,753. You have to add to this those from the earlier generations. That will give you the total number of those infected (roughly, 1.9 million). Some of them will have recovered by then and happily immune, others would have died. I don’t see how this adds up to 25.5 million, either as the number of all people who have ever been infected, let alone all people needing care at a certain date.  You would get to a cumulative 26 million in 11 generations with 15 million new infections. That is 71.5 days, 10 weeks, still only late May.

You can make the model more complicated. Ferguson assumed a variable R in each generation and it should also vary across generation as the number of people getting immunity increases.

Here is a nice calculator that adds a few other considerations.

The real scary numbers come from the healthcare system. There are only 74,000 hospital beds in California, and 6,300 in SD  county, only 32% of which are available. This is probably similar in the state overall. But what you really need is ICU beds (only 800 available in SD county). There are about 50,000 ICU beds in the entire US and about 100,000 respirators. And you also have to add to this that beds, even ICU beds are useless unless you have trained personnel attending to them. So if we suppose only 2 million people being sick at the same time in CA, and only 10%  (100,000) needing hospital beds and only 4% (40,000) ICU beds, we have a major catastrophe. 


Tuesday, March 24

On the duration of the shutdown, Jeffrey Sachs invokes the example of China. Their ironclad version of suppression, including mass testing, suggests the spread of SARS-CoV-2 can be stopped in 60 days.  Sachs says 60-90 days.

This is not what's happening in Italy, where exasperated mayors berate their citizenry.

Buzzfeed does funniest home videos for the Covid quarantine

As India's government orders a 3-week "total lockdown,"  nearly 60 percent of the U.S. population is not under stay-at-home orders or being mass-tested.  The U.S. is therefore not, overall, doing suppression, but mitigation of SARS-CoV-2.  Note that this predicts some "flattening of the curve" of infection--reducing but not eliminating the overload on health care-- but not reversing the spread of the disease (Ro stays above 1). Some red state politicians are actively resisting social distancing (Texas, Mississippi), as is POTUS himself.

Speaking of testing, California is way behind New York, working "piecemeal."
This piecemeal approach, said Harvard epidemiologist Michael Mina, is a key problem with testing in California and nationwide.
“We have a decentralized healthcare system and we have no way to scale for government means,” Mina said. “Everything is privatized, everything is individualized in our country and it’s become our Achilles’ heel in this case.”
   
Wednesday, March 25  It's Bailout Day!

NYT summaryEssential first take by David Dayen. Trigger warning: wow will this analysis not reassure you that any economic reforms are in the offing.

Yes we have no protection: "A very American story about capitalism consuming our national preparedness and resiliency"  Painful contrast between the American scramble for the most basic equipment and Germany's highly successful health system for radically minimizing fatalities.

Half-assed LAT reporting on the coming fiscal crisis of the state of California.  No real info, and other annoying stuff. How do you find the school lobbyist who will say this will be really bad for the schools, and then add, "under current law, it is likely that schools could withstand a total statewide revenue loss of around $5 billion. But more than that and schools will face significant problems."  So your own lobbyist just told the state that a 7 percent cut is fine. 

Where's higher ed in the stimulus bill? Inside Higher Ed's summary:
Six-Month Loan Deferment in Senate Bill
March 25, Noon. Student loan borrowers would be allowed to defer making payments for six months, without interest, through Sept. 30, according to a summary of the $2 trillion stimulus package Senate leaders agreed to at 1 a.m. Wednesday morning. The full bill is still being written and hasn’t yet been released.
But according to summaries of the bill making the rounds among education advocacy groups and obtained by Inside Higher Ed, the measure will also include changes sought by advocates such as not requiring Pell Grant students to repay money to the federal government if their terms are disrupted by the coronavirus emergency.
However, the bill is expected to disappoint advocates who had embraced Democratic proposals in the House and Senate, in which the federal government would have made the payments on behalf of borrowers, reducing their balances by at least $10,000. The summary did not mention any loan cancellation.
A separate summary contains $30.75 billion in grants to “provide emergency support to local school systems and higher education institutions to continue to provide educational services to their students and support.” That amount appears be about $29 billion less than what higher education institutions could potentially get in the bill proposed by House Democrats, but $21 billion more than what Senate Republicans had initially proposed, one higher education lobbyist said.  Associations representing institutions that were disappointed with the previous proposals were still waiting for the full bill before they commented on the level of funding.
The bill requires the secretary to defer student loan payments, principal, and interest for six months, through Sept. 30, 2020.

Thursday, March 26

Covid revealing America's rear guard place in the world 

Zero Hedge's mashup of hostility to the shutdown, mixing vulnerability of SARS-CoV-2 to treatment (it isn’t a superbug) with statistical problems (extensive) with lockdown’s effect on the economy (bad but unavoidable). Playing rural roulette because lockdowns are Democrat.

Suppression works, says none other than Neil Ferguson!!
He said that expected increases in National Health Service capacity and ongoing restrictions to people’s movements make him “reasonably confident” the health service can cope when the predicted peak of the epidemic arrives in two or three weeks. UK deaths from the disease are now unlikely to exceed 20,000, he said, and could be much lower.
But don't go back outside! Because, on the other hand,
This measure of how many other people a carrier usually infects is now believed to be just over three, he said, up from 2.5. “That adds more evidence to support the more intensive social distancing measures,” he said.
Special bonus for modeling fans: Oxford now has a model too. More on this coming soon.

Hope for a UK Covid-19 home test within two weeks.

And We're Number 1 - in Covid-19 cases.

4 comments:

AndrewD said...

Nice poet! Thanks Chris. My cavil is that I feel your terms "mitigation" and suppression" are likely not unambiguous, but rather regions in the space between "do nothing" and "solitary confinement" for all. These regions will change from country to country depending on the individual country's social structure, culture, and governance.

Chris Newfield said...

@AndrewDvery true! same issue within the US, as lots of states still have open bars, big weddings, etc. the variations can be deadly.

Sara Schoonmaker said...

Thanks for this post, Chris. I agree that Newsom did the right thing by going with the suppression approach, but that we need to have more evidence about why he made that call and also an end date. As you continue to work on this, I'd be interested to know best practices on how to come out of these suppression regimes. Sara Schoonmaker

Chris Newfield said...

@Sara Schoonmakeryes hopefully the unemployment filing spike will start a more sophisticated phase. but first we need some masks and test kits. I heard China was willing to send us a couple of mercy shipments . . .

Join the Conversation

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