The pandemic has changed—and it’s worse than ever.
Cases of Covid-19 are spiking in nearly every state. The statistics are grim. With more than 100,000 new cases and 2,000 deaths every day, hospital intensive care units are filling up everywhere. It’s an entire country of New-Yorks-in-April. And yet Covid skepticism—over how to fight the disease, and sometimes even the reality of the disease itself—remains a hallmark of right-wing politics.
Read all of our coronavirus coverage here.
There’s a light at the end of the tunnel—vaccines may well be available to millions of people before the end of the year. That’s a scientific triumph, to be sure, but meanwhile, we’re still in the tunnel. Manufacturing and distribution challenges mean that it’ll take until at least June 2021 to vaccinate everyone, according to the head of logistics for the government’s vaccine-goosing Operation Warp Speed program. Until then, the same public health measures that slow the spread of the virus—curve-flattening “non-pharmaceutical interventions” like mask-wearing and canceling gatherings—remain the only tools in the toolbox.
So policymakers and leaders have a stark choice: Force everyone, again, to abide by effective but potentially onerous public health measures, or let hundreds of thousands of people die. It’s a tougher choice than it sounds. Anti-“lockdown” rhetoric and a year of stress has ground people down emotionally and nuked the economy. Plus, it’s a basic tenet of public health that abstinence enjoinders and shame don’t work. If curve-bending efforts aren’t politically and socially viable, they’ll fail—and hundreds of thousands of people will die. As Mike Ryan, executive director of the World Health Organization Health Emergencies Programme, said at a press conference last Friday: “Those countries currently in the fight of their lives, you have got to stick with this. You’ve got to try and control this transmission, or your health systems will not be able to cope.”
What you’d really like to know here is which specific interventions give the most bang for the buck, the greatest reduction in disease transmission with the lightest possible touch on the social fabric and the economy. Is it … a mask mandate? Closing bars and restaurants? Closing schools? Temperature checks at building entrances? It would be very, very good to know this, because all of those things have benefits, but also costs. But scientists and public health experts don’t have answers. They know these things work in the aggregate, but not how they overlap and synergize, how behaviors change in response to new rules, and how politics and sociology affect adherence.
That’s why nothing seems to make sense today—indoor dining open, places of worship closed; outdoor playgrounds closed, gyms open; outdoor dining areas open then closed; curfews implemented on bars. In many countries, early measures combined with financial support and rigorous testing, tracing, and isolation programs squashed the disease. The US and Europe did some of the first thing and almost none of the others, dooming those places to a bloody oscillation: Cases spike, lockdowns come, economies and mental health crash, places reopen, cases spike, repeat. And now, well, we are where we are.
In the initial months of the pandemic, everybody blew it. “Ideally, you want interventions that have maximal effectiveness with the least social and economic downside,” says Lawrence Gostin, director of the O’Neill Institute for National and Global Health Law at Georgetown University. “That’s the rational way of doing it. But there’s been no rationality around fighting this pandemic, particularly in the United States and Europe.”
Public health experts know, in the broadest terms, what measures will bend the curve, but the science of it is really more of an art. Every country around the world rolled out roughly similar sets of public health interventions to fight Covid-19 in roughly the same order, at roughly the same moments in their encounters with the disease. According to research led by Thomas Hale at the Blavatnik School of Government at Oxford, most countries started communicating to their citizens in February about the potential problems to come, and instituted border controls even before they had confirmed cases. About 25 days later, in mid-March, countries started canceling public events and closing schools, and then closing workplaces five days after that. Four days after that, on average, came stay-at-home orders and public transportation closures—lockdowns.
The total portfolio of interventions is familiar, and taken in total they’re still the best and only way to reduce Covid-19’s spread—that portfolio is the basis for new, sweeping recommendations from the Centers for Disease Control and Prevention, for example. But nobody studied those measures closely to see what was working. “I think there has been a missed opportunity to learn from the first-wave experiences,” says Ben Cowling, an epidemiologist at the University of Hong Kong who has tried to assess interventions for Covid-19 and other outbreaks. “A lot of interventions were applied, natural experiments in different places, and it’s been a missed opportunity not to study those more carefully to see what happened, what worked, and what didn’t.”
Take mask-wearing. I’m not going to focus too much on that here, because so many smart reporters already have, and because widespread mask-wearing clearly cuts transmission of Covid-19. But just consider how hard it is to really-and-for-truly test that proposition. It’s possible to, for example, test how well masks reduce the emission of the aerosolized particles that seem to be a major contributor to the spread of the virus. You can also compare places with mask mandates to places without and see what the virus does. In Kansas, for example, counties that put a mask mandate in place last summer saw Covid-19 incidence decline; counties that didn’t mandate masks saw incidence increase. The same correlation held in Germany; regions with mask mandates for transit and shops saw reductions in cases while non-mandate regions did not, leading researchers there to conclude that 20 days after mask mandates go into force, infections go down by about 45 percent.
That’s great, and those new CDC guidelines are all-in on masking, recommending people wear them everywhere other than their household, including indoors. President-elect Joe Biden told CNN that he’ll ask all citizens to wear masks for the first 100 days of his term—quite the turn from Donald Trump’s derisive attitude. Mask not what your country can do for you, I guess.
But mask studies still have some gaps. Was it really the masks themselves blocking enough virus emitted from the mouth and nose? That’s almost certainly true, but what if the population-scale effect of masking also includes more subtle effects? “Mask mandates usually arise when cases are rising. Does a mask mandate reduce transmission because masks actually stop transmission by blocking virus when we breathe out, or because wearing masks reminds us that there is a pandemic going on, and we don’t socialize?” says A. Marm Kilpatrick, an infectious disease researcher at the University of California, Santa Cruz. “Or, are masks completely irrelevant and the decrease in transmission that sometimes follows mask wearing is just because people were freaked out by news reports of cases rising? Disentangling these effects is extremely difficult.” It’d be good to know those things, if for no other reason than to be able to convince people to wear their damn masks.
A gold-standard, randomized controlled trial of mask use would answer these questions—you’d think. A recent Danish study attempted that. One group of people got told to wear masks outside their homes, and another group didn’t. According to the study, masks didn’t really help; the reduction in Covid-19 infections wasn’t significant. But as other analysts have written, the study wasn’t enforcing actual mask wearing. So it wasn’t actually a study of masks versus no masks; it was a study of the effect of telling people to wear masks. If anything, it was a study of mask mandates, not mask effectiveness. Even the lead author acknowledged that it didn’t really say anything about whether masks worked. And it wasn’t really a gold-standard trial. For one thing, it’s impossible to blind which people were doing masks or no-masks. For another, the study didn’t actually isolate the thing it set out to test.
Actually isolating the right variable and getting statistical power out of a double-blinded RCT on masks would be nearly impossible. “Doing a mask RCT with 30,000 people would be very, very challenging, and one would need to somehow measure mask usage, since a substantial fraction of people wouldn’t wear their mask all the time,” Kilpatrick says.
Masks are relatively cheap, and mask mandates are relatively easy to implement—especially if you’re not trying to put any power of enforcement behind them, like fines. Harder, society-level interventions with bigger economic and social implications are even tougher to study.
People have tried. When you can’t run an RCT, you have to be satisfied with natural experiments, observational and retrospective analyses, and computer models. In mid-November, The New York Times correlated Hale’s Oxford countermeasure data for US states with the extents of their outbreaks—new cases per 100,000 people and hospitalizations per 100,000 people. Sure enough, the states with the fewest and least potent restrictions—North Dakota and South Dakota, it turned out—had the worst outbreaks. Hawaii’s rules were the most stringent, and the state had the fewest cases.
That analysis makes intuitive sense, and it supports the everything-all-at-once strategy. But it isn’t perfect. Time plays a role here—tests were hard to come by early in the pandemic, which means states could have had thousands of cases that went undetected. Hawaii is a lot easier to keep potentially infected travelers away from than, say, Iowa or Kansas, so maybe the Aloha state had fewer patient zeroes to begin with.
Just a couple of days before the Times analysis, the journal Nature Human Behaviour published another attempt at developing a heuristic for non-pharmaceutical interventions. An international team led by Austrian network scientists looked at the rise of nearly 7,000 different countermeasures in March and April in 79 different countries. They found a pattern similar to the one that Hale’s team discovered. But then things got tricky. “Many countries implemented bunches of measures simultaneously, and this is statistically challenging, because if you have 10 interventions implanted the same day, it’s hard to disentangle the effects,” says Peter Klimek, a data scientist at the Medical University of Vienna who led the research. “And these individual interventions are closely related to each other. You can’t shut down schools without shutting down other areas of life in parallel. There is no such thing as the effectiveness of a single intervention.”
Klimek’s team tried anyway. They looked at the kinds of interventions that the countries ran and their concurrent reductions, if any, in the effective reproduction number, Rt—how fast the disease moves from person to person. Then the team ran four different kinds of analyses: They ran a complex regression model, basically comparing countries with specific interventions to countries that didn’t do the same thing; a time-series regression that assumed the interventions had specific effects they could find in the numbers; and a couple of machine-learning algorithms to look for patterns, too. (They also tried to build in ways to account for socioeconomic and political differences in the countries themselves.) “For some measures, these four different methods gave us completely different results. But there was a core of interventions on which they could agree,” Klimek says, “where they were implementing a measure and observing a reduction of the effective reproduction number.”
The winners? The most effective move was canceling small gatherings like weddings and parties, and closing shops, bars, and restaurants. Number two: school closures, which is controversial because Covid-19 seems to spread differently among younger kids, and because closing schools has all kinds of knock-on effects—developmental impact on kids, forcing caregivers to stay home as well. Klimek says his team saw hints in their data that closing high schools was more effective than closing elementary schools in terms of reducing disease transmission.
Making sure that health care facilities had enough personal protective equipment was number three, and good communication strategies were number four. Airport restrictions also helped, but only if countries put them in place in the very early days of the pandemic (so no help there, really). Mask-wearing showed up in some of the statistical approaches but not all, and might be subsumed in some of the other broad descriptors, like social distancing.
The losers? Improvements in testing and tracing didn’t seem to help—maybe because they weren’t implemented soon enough or at grand enough scale. Closing intracity public transit didn’t move the needle, either. “These are just not places where a lot of transmission happened,” Klimek says.
But as appealing as this assessment might seem, it’s not straightforward. All of those “winners” that Klimek laid out might (or might not) be part of a “lockdown” or stay-at-home order like the ones coming into force in California. Instituting one or several might enhance or limit the effects of others. “The measures do not associate in a linear way. They overlap partly,” Klimek says. On the plus side, that means removing a single intervention from the set doesn’t necessarily knock over the whole Jenga tower. Places could open schools, for example, but keep bars and restaurants closed and prohibit small gatherings, and still get a good effect. Maybe. “We’re starting to see this in many European countries,” Klimek says. “It’s not only black and white. We can do this now in a more differentiated way.”
It’s a good thought, and Klimek is working with the Austrian government to put it into effect. It’s also not clear-cut. “I’d caution you that regression analysis and modeling are very unreliable,” Gostin says. “You really need to do rigorous retrospective analyses to try to figure out what works and what doesn’t, which we’ve never really done.”
Worse, though, as the paper itself acknowledges, when it comes to population-level public health interventions during a crisis, subtext is everything. It’s the unmeasured changes in people’s behavior that confound work like Klimek’s, the second-order consequences of the rules—like caregivers staying home because their kids don’t have school to go to. If the government doesn’t subsidize staying home, let’s say, those people could lose their jobs, which crushes their finances and harms the overall economy. So school closures don’t “work,” as such, without all kinds of other supporting efforts. “When we think about other NPIs, like school closures, the effect of that kind of intervention could vary from one place to another,” Cowling says. “So I don’t think it’s possible to have an estimate of the effect of school closures on Covid. You could have the effect of school closures on Covid in April 2020 in the UK—in a particular place, at a particular time, with an intervention described in a certain way.”
That means research like Klimek’s faces a huge task in trying to disentangle what public health rules say from what they actually do. “What they are actually measuring is the direct effect of these interventions, plus the indirect effect of these interventions, plus the direct and indirect effects of other things happening at the same time,” Kilpatrick says. “None of the papers I’ve seen tries to quantify media coverage of local epidemics, even though I’d bet that’s one of the largest effects on peoples’ behavior.”
(It does seem to be true that perception of the seriousness of the pandemic changes how people respond to it. Simple perception of the threat—understanding that Covid-19 is dangerous—makes people more willing to comply with interventions, or more cautious overall, and so indirectly reduces Rt, or infections or death rates. This, too, makes intuitive sense. People who hear that their local hospitals are full or have Facebook feeds full of posts from sick friends and relatives might be more likely to simply stay home. In one model—so, caveats apply—a greater awareness of Covid-19 deaths reduced, in the short term, death rates overall. I tried to hint at this a little more bluntly last April, in a story that tried to figure out how many people would have to die of Covid-19 in the United States before everyone knew at least one person, and so might be more open to behavioral interventions. It turns out to be a harder math problem than it sounds, but the range was between about 500,000 and 1.6 million people, which seemed much less possible nine months ago than it does today.)
And the converse might be true too. Maybe people who don’t buy it, who get their information from sources that downplay Covid-19 or think that a laissez-faire herd-immunity strategy would work (it wouldn’t), might not respond to any interventions. They’d put themselves at risk, and since Covid-19 has significant transmission by people who have no symptoms, put others at risk as well. That could look like the intervention itself was failing, when the real problem was compliance, or lack of synergistic effects with other subtler, cautious behaviors. So which thing actually cut Rt?
No, seriously, I’m asking. Because no one really knows.
Even worse, different places and conditions might lessen or increase how well Covid-19 spreads. It’s “spatiotemporally heterogenous,” meaning the disease moves in spikes through populations, with super-spreading events having a disproportionate effect on where and when it crops up. One study, from researchers in Canada and England in the journal Proceedings of the National Academy of Sciences in September, even hypothesized that different interventions would affect the reproductive number differently at different kinds of events—smaller or larger, longer or shorter, and so on. The trick is knowing which kind of event you’re at, and getting people there to believe they should do the safest thing.
A catastrophic year of inconsistencies has led, in the US, to millions of sick people, a quarter million deaths, and mounting frustration. Even influential public health voices, like former White House Covid-19 task force leader Scott Atlas, seemed to be arguing for a “herd immunity” strategy that involved letting everyone get sick and hoping for the best. If nothing seems to work, why bother trying?
Of course, it’s not that nothing works. It’s that everything does, a model that some public-health-minded scientists liken to stacking slices of Swiss cheese. Layer enough of them, and all the holes get blocked. The sum is greater than the holes of the parts.
If the government wants people to keep building cheese barriers (if you see what I mean), the cheese has to come with social and financial support for staying home, for wearing masks—money to keep businesses and homes together until the pandemic ends—and clear, transparent communication from officials (and maybe some famous people, too) to explain what’s going on and why. None of that has happened. “The most effective interventions are population-wide risk mitigation measures based on changing the public’s risk profiles. That’s why the United States has failed so abysmally—because very simple behavioral mitigations have become politicized,” Gostin says.
Back in April I wrote that the point of communal public health measures was to hold the line, to keep cases and deaths low, to keep the hospital system from being overwhelmed until research scientists could get us out of this mess. I thought, perhaps naively, it would take months. It took a year, and cases and deaths are worse than they’ve ever been. But we’re almost there—months away from getting a shot that looks likely to stop the pandemic. Holding on is the only option. “It’s not the expressed goal, but that’s the absolute hidden goal,” Gostin says. “That’s literally all we’ve got left. We’ve failed at everything else.” Non-pharmaceutical interventions are the only tools in the toolbox, and they only work if you take them out and use them—even if you don’t know how to use them well. They work, and they’re the only way the country will get from here to the After Times. But it’s going to be a muddle, not a march.
More From WIRED on Covid-19
- 📩 Want the latest on tech, science, and more? Sign up for our newsletters!
- Two global efforts try to trace the origin of the Covid virus
- Warp speed, atomic structure, and the future of vaccines
- The free-market approach to this pandemic isn’t working
- Winter is coming. Could humidifiers help?
- The vulnerable can wait. Vaccinate the super-spreaders first
- Read all of our coronavirus coverage here
