Slate is making its coronavirus coverage free for all readers. Subscribe to support our journalism. Start your free trial.
“A small but important subset of people with the coronavirus also really hate puppies.” That’s what I keep thinking to myself as I see story after story linking various, so-called surprising symptoms to COVID-19: pinkeye, loss of hearing, loss of smell, brain impairments, and “staring off into space.” These pieces provide varying levels of evidence, from a Twitter dispatch to a patient to surveys conducted by researchers, to propose that the mere correlations their authors are writing about might actually have some sort of firm, causative relationship.
Each of these symptoms lacks substantial peer-reviewed evidence connecting them to the novel coronavirus; at best, there are a couple studies. (Loss of smell has perhaps the strongest case, especially given that it’s an established symptom of regular upper respiratory infections, but even that case is weaker than some of the data circulating suggest.) They all involve doctors and reporters scrambling to make sense of how to spot the novel coronavirus as we experience a shortage of the things that would actually help us spot the novel coronavirus: reliable lab tests, consumer thermometers (fever is a consistent symptom of many cases), and a robust understanding of how pre-symptomatic transmission helps spread the virus.
The latest in this emerging symptom-free-association genre is an opinion piece by Seth Stephens-Davidowitz, a data scientist, that ran in the New York Times titled “Google Searches Can Help Us Find Emerging Covid-19 Outbreaks.” Searches can “help detect unknown Covid-19 outbreaks, particularly in parts of the world with poor testing infrastructure,” writes Stephens-Davidowitz, referring to the United States of America. He sifted through data for searches for symptoms that spiked along with a rise in COVID-19 cases. He presents his case that, based on a spike in searches for “my eyes hurt,” eye pain might in fact be a potential new symptom of the coronavirus.
This is truly bananas!
What Stephens-Davidowitz has is at best a whisper of a suggestion that eye pain might possibly be sometimes associated with COVID-19. He acknowledges that his research project does “not necessarily” mean eye pain is a symptom. But he also argues that other possible explanations for spikes in eye pain searches don’t explain it, as searches for eye pain do not correlate with pollen concentrations or a decline in movement, as shown by data collected by cellphones (sure wish he’d linked to that data). And yet, the rise in “my eyes hurt” searches could still be due to several other things that also correlate with what it’s like to live through a pandemic. Maybe the people searching are stressed, sobbing, or trying to spare themselves a trip to the ophthalmologist. It’s hard to tell what, exactly, Stephens-Davidowitz considered or didn’t, since he’s writing in an op-ed section, not a scientific paper. We’re sort of forced to take his word that the data are as robust as he implies.
Might those eye pain searches end up being directly caused by the coronavirus, along with some of those other weird symptoms? Sure, they might. Maybe eye pain will be linked to the coronavirus in some unexpected way. Maybe it will be a random coincidence that some people experiencing neurological issues—another potential symptom written up in the Times, based on anecdotal evidence and a preliminary paper—also have the coronavirus, just as it might be accurate to say that some people who have the coronavirus also hate puppies. This is exactly what we mean when we say that correlation doesn’t imply causation—two trends might be rising in tandem while being totally unrelated or maybe only linked by some third factor. (For example, perhaps elderly patients living in crowded nursing homes are both more likely to contract the virus and also more likely to have neurological conditions.)
Science is moving incredibly fast right now—much faster than usual as we try to figure out this virus. That’s as it should be, but “speed remains the enemy of rigorous science,” writes Retraction Watch founders (and, disclosure, my former bosses) Ivan Oransky and Adam Marcus in Wired. They point out that many papers posted on pre-print servers right now “will likely turn out to be at least partially wrong.” Oransky and Marcus propose affixing a black box warning to papers that says as much. If only we could do the same for news stories. For now, the best I can do is continue to remind you to take all of these possible symptoms with an enormous grain of salt.
For more on the impact of the coronavirus, listen to Monday’s episode of The Gist.