Pandemic diary 2020-05-25: Two big questions

My broad take on the pandemic has been and remains pretty grim. Early on, I thought that after decades of prioritizing efficiency over resilience in our supply chains, the pandemic could lead to outright collapse. Gratefully, I think that less likely now, although the possibility is not off the table. It feels like we are in the middle of this thing, but by the numbers we are still in the beginning. According to serological tests in Spain, which has endured much worse mortality than the US, only 5% of the public has been exposed and should be expected to have some degree of immunity. If we take the infection fatality rate to be 0.7% (see a range of estimates here) and if we imagine — very optimistically — that only half the US population becomes infected before “herd immunity” sets in, then we should expect a toll of more than a million dead should we fail to suppress the epidemic.

So far we are “only” one hundred thousand tragedies in. Just the beginning. This is the thinking that places me squarely in the camp of coronavirus hard-liners. I think that we as a society should place a very high priority on infection suppression. For now that means people should either be at home, outdoors and socially distant, or carefully protected when performing essential work. (Whether that “should” needs to be legally enforced is a different question, which I’ll put aside for now.) We should reduce the burden of restricted lives with generous financial and social support. But we absolutely should not encourage people to return to restaurants, bars, theaters, malls, theme parks, gyms, or salons until we are capable of suppressing outbreaks (à la Korea, test/trace/isolate), or until we develop a vaccine or effective treatment.

Is my view wrong, harmful and overdone? One of the poles of our heartbreakingly polarized political system says it is. Restaurants, salons, and gyms are all open here in Florida, where my family is isolating. (Long story.) The benefits that a functioning service economy provides to customers, workers, and business owners are very real. The question is, can “reopening” be done safely? If it can be, to the degree we can reopen without accelerating infection, of course we should. But we should not undertake any reopening that would place us back on a path of exponentially increasing infection. So we come to the key uncertainty: How dangerous is reopening? Right now the US infection rate seems stable, even gently declining. But are we suspended by a delicate thread thousands of feet above a rocky canyon? Or are we held by a firm cable just a few feet above grassy earth? I don’t think we know. A month since states began to reopen, we don’t see much evidence of accelerating infection. Internationally, there has also been some loosening of restraints, without obvious remissions. Sweden, which has been among the least cautious developed countries both in terms of legal restraint and actual behavior, is seeing its (once very high) mortality rate decline despite remaining far from estimated herd immunity. Outbreaks in the developing world, thank god, have not so far been as severe as we might have expected, at least as we can imperfectly observe them. On the other hand, benign outcomes may be the result of continuing caution by the general public. Nowhere are service economies thriving. “Reopenings” are happening mainly in low prevalence places, drawing customers from currently low prevalence populations. But prevalence can change. As we continue to loosen restrictions (legal restrictions, or just restrictions on our own behavior), sparks that have so far died out may suddenly ignite. Some of our sort-of success may be attributable to factors like seasonality and mask-wearing, but both of those are in different ways temporary.

There are I think two potential explanations of benign outcomes that would really change how we understand the pandemic. But I don’t think we know if either of these explanations are true, which turns them into very big open questions.

Explanation I: The spread of COVID-19 is overwhelmingly driven by “superspreaders”, and superspreading is determined much more by behavior and circumstance than by the biology of the superspreader.

If it is the case that the vast majority of infecteds don’t spread at all, while the average number infected (R0) of roughly three is due to a few superspreaders that compensate for the nonspreaders, then preventing infection by these superspreaders could be sufficient to suppress the epidemic. If superspreading is mostly an invisible biological phenomenon, if some recently infected people just shed a whole lot more virus than others, this fact doesn’t help us much with “reopening”. Most of the time most restaurants will operate safely, but occasionally a superspreader will walk in, and unbeknownst to her or anyone else, infect half the room. However, if superspreading is less about biology and more about behavior and circumstance, then we can simply limit the activities during which superspreading occurs. We know indoor choir practice is probably an activity that risks spreading the virus, so we don’t do that. Superspreading seems associated with congregate living arrangements (prisons and god help us nursing homes), loud crowded workplaces (meatpackers), crowded indoor social gatherings, etc. We know that we need to protect or reorganize congregate living places and crowded workplaces — although disgracefully, we seem instead to be opting for a “herd immunity strategy within many of these settings. Following Japan, we could encourage people to avoid the “Three Cs”. We’d keep places like nightclubs and sporting events closed, open up but limit crowding elsewhere.

But this only works if less close, less crowded interactions really are safe. Definitively resolving this is a strong case for aggressively pursuing contact tracing by human interviewers. If infection is a matter of avoidable circumstance rather than an ineluctable result of prior infections (as it might be in simple models of “open”), then actually learning the details of the circumstances under which people became infected is our surest path to freedom. Are indoor restaurants mostly okay? How should we arrange the air conditioning? In our dismal, ridiculous, politics, contact tracing has become associated with Democrats who (like me) want to see the US pursue a South-Korea-style strategy and criticize the President for our botched test-and-trace ramp up. Contact tracing is therefore dismissed by Republicans, who argue that Korea-style containment is unworkable in the United States, so what’s the point? But Republicans eager to reopen should be especially devoted to contact tracing. Reopenings that prove catastrophic to public health will not help anyone, economically or electorally. Contact tracing with detailed interviews is how we learn to reopen safely, if it can be done.

Explanation II: Immunity not reflected by serology

Until a few weeks ago, a big open question was whether the coronavirus was spreading much faster than the disease we actually observed. This, oddly, was more a hope than a concern. If mild infection was in fact much more widespread than we knew, then infection fatality and severity rates — the fractions of people infected who die or experience severe illness — would be much lower than we thought. For individuals, it would mean the disease is less dangerous. In aggregate, it would “lower the ceiling” on mortality. (In the bad case where the epidemic runs its course, a lower infection fatality rates means a lower total casualty count.) We thought that serological studies would answer this question of unobserved spread, and probably they have. Using Spain as an example again, we’ve learned that for every infection we formally detect, roughly 9 people other people who were never diagnosed show antibodies. That may sound like a lot, but it was only enough to bring the infection fatality rate down to about 1%, which still would mean more than a million US deaths without suppression. Undetected cases would have had to be closer to two orders of magnitude more common than detected cases, rather than just one order of magnitude, to keep the death toll in the ballpark of a few bad flu seasons. According to the serological data we have so far, I think this hope is thoroughly dashed. Studies that suggested much more widespread infection, like the Stanford Santa Clara County study, have not proved persuasive.

A last ray of hope lies in the possibility that people who undergo very mild infections do not end up with reliably detectable levels (“titers”) of antibodies in their blood, but still end up with “memory” cells in their immune system that would let them ramp up production of antibodies and specialized T-cells quickly if they were re-exposed to the virus. In other words, there may be people who have a degree of immunity that is invisible to our antibody tests, so that despite our serological disappointments, the true infection fatality rate is lower than we estimate. This is pure hope: There is no evidence that I know of for this idea, beyond circumstantial handwaving about ceilings in the levels of mortality per population we’ve observed so far, which more likely reflects a successful behavioral response than a speculative immunological one. This unobserved immunity hypothesis stands on the less persuasive side of a war between parsimony and hope. Nevertheless, I’m willing to hope.

But I think it’s the wrong call to bet people’s lives on this kind of hope until we have some evidence for it.


6 Responses to “Pandemic diary 2020-05-25: Two big questions”

  1. Kien writes:

    Thanks, very interesting, and I agree these are arguably the most important unknowns. The greatest difficulty with Covid-19 is how much uncertainty there is about how this virus infects humans. Even now we are still ignorant about very important issues like the ones you have identified.

  2. Detroit Dan writes:

    So the lockdown has largely worked, and the probable reason is that superspreaders have been contained. This seems like the most likely guess so far.

    If it is the case that the vast majority of infecteds don’t spread at all, while the average number infected (R0) of roughly three is due to a few superspreaders that compensate for the nonspreaders, then preventing infection by these superspreaders could be sufficient to suppress the epidemic. 

    I imagine the superspreaders doing a lot of sneezing and coughing during choir practice. Could it be we’re stumbling toward some sort of rational response as a nation, despite our apparent dysfunction? Nobody wants to die, so even the Republicans are applying common sense? The Democrats may be overly cautious, but this is acceptable for many of us individually and best for society as a whole?

  3. Unanimous writes:

    Some jurisdictions have put quite an effort into tracing including attempting RNA sequencing of all detected cases to link up cases as far as practical. There was one preprint article from the Doherty Institute on cases in Victoria, Australia for example.

    It’d be surprising if health authorities in these jurisdictions didn’t have the answers to your questions based on tracing results and cluster analysis. You might look for further preprints on cluster analysis and tracing, but absent that, watch the restrictions that remain in Taiwan, South Korea, Singapore, Australia, New Zealand, Iceland and Norway.

  4. C writes:

    There is some other nascent scientific work backing up your first hypothesis. NYMag linked to the attached study which seems to suggest that indoor crowds and some superspreaders account for a lot.

    I have not read the paper enough to tell how conclusive it is but it appears that there is some evidence along those lines.

  5. albatross writes:

    I think it’s likely that lots of things have to line up just right to get a superspreader event–you have to be producing a lot of virus, producing/spreading around a lot of droplets, and be in a crowd in a place where you can maximally infect people.

    Telling people to stay home when they’re sick is one way to try to deal with the first one of those, but apparently you’re most contagious the day or two before symptoms start, so it’s probably not so useful for most superspreader events.

    Masks (especially if we can get everyone using decent ones instead of makeshift T-shirt-material masks) probably help with the second–fewer droplets get out and they don’t get as far.

    The third is the hard part, because a lot of things people like to do seem to be good ways to have a superspreader event–weddings, funerals, parties, crowded restaurants, going to church and singing along, etc.

  6. How will all the rioting impact the spread of COVID-19?