As the coronavirus pandemic sweeps the world, shutting down entire countries and threatening the worst economic crisis since 2008, developed nations are asking their citizens for the kind of sacrifices that normally are restricted to wartime. A shelter-in-place order now covers millions residents in California and Illinois. Colleges and universities around the country are shuttered. Are these unprecedented efforts to “flatten the curve” really worth it? One renowned researcher – John Ioannidis, a specialist in scientific methodology at Stanford University – is skeptical, arguing that there’s not enough data to justify wrecking the economy by putting the country on lockdown. His analysis is being taken seriously by scientific luminaries such as Steven Pinker. But it has some big shortfalls.
Ioannidis is famous for his 2005 paper, “Why Most Published Research Findings Are False.” This article was a stinging rebuke to the cultural underbelly of science, where researchers often prioritize flashy findings over accuracy, bend results to fit funders’ agendas (a particular problem in pharmaceutical research), and rely on sloppy statistical analyses. The problem was that this underbelly wasn’t just a small, seedy portion of the scientific world. In fact, a lot of scientific research – including that which drives public policy and healthcare decisions – is probably deeply flawed: hard to replicate, biased, and spurious. Since Ioannidis released this bombshell article, it’s been cited more than 8,000 times and has helped spark massive revolutions in how sciences such as psychology carry out their research.*
So Ioannidis has his bona fides. People – important people, influential people – trust him. He’s also an inveterate skeptic and data nerd, the world’s go-to expert for sniffing out the holes in fallacious scientific arguments. One of his biggest pet peeves is when arguments aren’t built on a foundation of rock-solid data. In the midst of the public-policy debates surrounding the coronavirus pandemic, he thinks that the case for preemptively closing down cities and regions to contain the spread of the illness is just such an argument. Writing in STAT, he argued on Wednesday we simply lack the data to determine whether such extreme measures are a good idea. But his hard-hitting analysis illustrates a unique danger of data wonkishness: putting so much stock in scientific abstractions that reality itself becomes invisible.
Hard Data in Uncertain Times
Let’s look charitably at his argument first. Problems with implementing testing mean that we have no real idea how lethal the bug is, how contagious it is, what proportion of patients need hospitalization, or what would really happen if we just let it run its course. And since most patients tested are those with symptoms, yet the majority of carriers of the COVID-19 virus have mild illness or are asymptomatic, the actual lethality of the virus may be much lower than the 3.4% originally estimated in Wuhan or even the .6-.9% in Korea. Given the extraordinary economic costs of what essentially amounts to a global shutdown – some financial experts are worried not just about recession, but about a possible depression as a result of coronavirus containment efforts – is it really worth acting on the basis of such flimsy and uncertain data?
There has been one opportunity to examine the spread and lethality of the COVID-19 virus in a closed population: the Diamond Princess cruise ship, which was quarantined off Japan for all of February. Ioannidis draws on the the Diamond Princess numbers to derive what he believes is a more accurate picture of the virus’s lethality:
The one situation where an entire, closed population was tested was the Diamond Princess cruise ship and its quarantine passengers. The case fatality rate there was 1.0%, but this was a largely elderly population, in which the death rate from Covid-19 is much higher. …Projecting the Diamond Princess mortality rate onto the age structure of the U.S. population, the death rate among people infected with Covid-19 would be 0.125%. …reasonable estimates for the case fatality ratio in the general U.S. population vary from 0.05% to 1%.
A population-wide case fatality rate of 0.05% is lower than seasonal influenza. If that is the true rate, locking down the world with potentially tremendous social and financial consequences may be totally irrational. It’s like an elephant being attacked by a house cat.
This critique mirrors the rhetoric of coronavirus skeptics around the United States, who see virtually nothing happening, disease-wise, in their actual neighborhoods and lives, and conclude that the whole crisis is overblown. Why shut down the entire economy and put millions of people out of jobs just to avert a bug that’s basically the yearly flu?
Ioannidis is – and is right to be – seriously worried about these newly unemployed people:
we don’t know how long social distancing measures and lockdowns can be maintained without major consequences to the economy, society, and mental health. Unpredictable evolutions may ensue, including financial crisis, unrest, civil strife, war, and a meltdown of the social fabric. At a minimum, we need unbiased prevalence and incidence data for the evolving infectious load to guide decision-making.…One can only hope that, much like in 1918, life will continue. Conversely, with lockdowns of months, if not years, life largely stops, short-term and long-term consequences are entirely unknown, and billions, not just millions, of lives may be eventually at stake.
These are chilling, if vital, warnings. No civilization in history has attempted this sort of totally coordinated response to a plague. In the past, plagues usually meant that every family, town, or village was on its own. Entire regions would see their populations devastated in just a few months – the Black Death infamously killed at least 1/3rd of all Europeans in the 1340s – and then afterward, slowly, crawl out of the ruins to start society up again. So it might seem pretty incongruous that we’re thinking of mobilizing on a World War II-type scale and potentially torching the global economy to stave off a much more innocuous plague that kills many fewer of its victims, the majority of whom are elderly or already have preexisting conditions.
Forest, Meet Tree
What Ioannidis doesn’t take into account, though, are the actual facts of what happens in real places when governments let the coronavirus spread naturally through the population. His anodyne projections of a flu-like epidemic are incongruous with boots-on-the-ground reports in, say, northern Italy. More specifically:
1. His back-of-the-envelope calculations focus on the relatively low lethality of the virus, but they don’t take into account its infectiousness, which we know for a fact to be much, much higher than the seasonal flu.
So when, for example, Ioannidis muses on a scenario in which the virus peters out naturally after infecting only 1% of the U.S. population and killing around 10,000 – a rounding error during an already-bad flu season – it’s hard to take him seriously. The growth chart for coronavirus is exponential in every country except those that have made aggressive moves to lock down populations or track and quarantine infected people – basically, China, South Korea, Singapore, and Japan.Exponential growth means that, for example, the U.S. only took
In reality, the doubling rate will slow as the virus reaches saturation, so it’s very unlikely that the whole country will get infected. But if we just let it rage out of control, it will infect much, much more than 1% of Americans – and very quickly. That’s what exponential growth means. It feels like nothing’s happening until suddenly everything is happening, and your hospitals are cracking under the weight of gasping, blue-faced patients who you’re forced to let die because there are no beds or respirators left.
Which brings me to the second point:
2. Ioannidis completely overlooks the temporal dimensions of the coronavirus plague.
Why is it so hard to get a bead on how lethal this bug is? Data from China suggest a simple explanation for the inconsistencies across regions: when good hospital care is plentiful, fewer than 1% of patients die. But when hospitals are overwhelmed and patients can’t get care, up to 5% succumb.
In Wuhan, where the virus first broke out, hospitals were quickly overwhelmed by the burden of new patients. Many of the victims could easily have been saved by proper care, but the surge of patients meant that the carrying capacity of the region’s healthcare facilities was crushed. By contrast, in other parts of China, the death rates for coronavirus infections remained well below 1%, because the rates of infection remained low enough to keep the inflow of patients well within hospitals’ carrying capacity.
It’s a bit like trying to catch water from a leak in a bucket. If the leak is just a slow dribble, you can stick a bucket under it, empty it out every hour or so, and carry about your business. But if the ceiling is pouring down a gallon of water every ten seconds, a single bucket just isn’t going to cut it. Water is always just water, but the flow rate makes a big difference for how much damage it can do. Coronavirus is the same way.
This is why the scenes in Italy have been so nightmarish. The newspaper in the town of Bergamo normally prints a couple of pages of obituaries per week, but last week it printed ten pages of obits – almost all from coronavirus.
This is not the flu.
When this coronavirus hits, it hits hard, and it lays entire regions low. This is because it overwhelms the hospitals with sheer volume of desperately ill patients. In northern Italy, doctors are being forced into wartime triage decisions – deciding who will get lifesaving treatment and who will die based on age and health. From an article at Reuters:
On Friday, the mayor of Fidenza, a city just outside the Lombardy region, shut access to the local hospital for 19 hours. It was overcrowded with COVID-19 patients and hospital staff had worked 21 days without a break. While the closure was aimed at keeping the hospital going, it meant some people “died at home,” said the mayor, Andrea Massari.
If coronavirus spreads unchecked across the United States, every hospital will become that hospital in Fidenza: overworked beyond its breaking point, with patients sucking up air in respirators while would-be patients are turned away to die at home. The New York Times reports on the rapidity with which hospitals in New York are already getting rammed:
“The most striking part is the speed with which it has ramped up,” said Ben McVane, an emergency room doctor at Elmhurst Hospital Center in Queens. “It went from a small trickle of patients to a deluge of patients in our departments.”
A recent study at Imperial College London warns that, in the absence of any suppression measures to “flatten the curve” (reduce the number of simultaneous infections) this overshoot of healthcare resources would mean as many as 4.4 million Americans dead – far more than every war the U.S. has ever fought put together, but with the deaths condensed into a single year. (For reference, the U.S. was involved in World War II for four years, and lost 400,000 dead in total.) The problem is that the curve needs to get very, very flat indeed to avoid the worst scenarios. Tomas Pueyo, a tech industry wunderkind, argues that extremely strong measures – total lockdown of the entire country, widespread testing and tracking of infected people – are the only way to ensure that we escape a cataclysm. If we enact those measures now, the worst of the nationwide quarantine might be over in as soon as a couple of weeks. Even better, we might escape with only a few thousand deaths.
That would feel anticlimactic for everyone who’s huddling inside their apartments in San Francisco, New York, or Boston right now. We spend weeks inside and then the plague never happens? But that’s the point. We’re trying to head off a once-in-a-century cataclysm. In an under-review white paper critiquing Ioannidis’s cool armchair analysis, Harry Crane, a statistician at Rutgers University, observes that
Given the severity of what we’ve already seen and the uncertainty of where we might be headed, the prudent approach is most definitely not to wait to sharpen our estimates. If there’s a chance that we’re underestimating cases by a factor of 300 (as Ioannidis alludes), then let’s assume we’re off by 300 and act accordingly. Whatever the case numbers are today, they’re likely to double (or worse) within a week if serious steps aren’t taken. The situation is bad, and it will only get worse if we’re lulled into a false sense of security by Ioannidis and others making similar calls for calm.
Science as Escapism
Ioannidis is a brilliant guy. He has more than 200,000 scholarly citations, and is widely looked to as one of the world’s top experts in critiquing scientific methods. But sometimes the quest for data blinds smart people to what’s front of their faces. It could be the case that the virus isn’t all that deadly after all, once all is said and done. It could be that its contagiousness has been overestimated. Those are totally realistic and reasonable possibilities. But, in fact, when the virus is simply allowed to spread in a natural human population, hospitals collapse. Small towns print five times the normal number of obituaries in a single week. Healthcare workers are pushed to the edge of breakdown. This isn’t what the flu does.
As I’ve written here before, science itself isn’t reality. Reality is reality. Scientific abstractions are just useful tools for getting a mental grip on phenomena. We can visualize the skyrocketing number of worldwide COVID-19 cases in a chart and make projections about what the caseload will be in a week, or a month, or two months. We can calculate the contagiousness of the virus in terms of its R0 value. But in this particular case, a very smart guy is falling victim to one of the classic scientific blunders: using data not a way to make contact with the reality we’ve got, but as a way to construct a more convenient alternative one.
* Not that psychology has exactly conquered the problem yet.