ProPublica is a nonprofit newsroom that investigates abuses of power. Sign up to receive our biggest stories as soon as they’re published.
There’s a pandemic spreading across the U.S., and amid ongoing struggles to scale up testing, it’s still unclear how many people are actually infected.
In the absence of real numbers, projections have filled the void, and I’ve been struggling as a reporter to know which forecasts to trust. Last week, the Ohio Department of Health said that over 100,000 people were infected, at a time when there were five confirmed cases in the state. Some epidemiology experts critiqued that estimate as too high, and the department’s director later said she was “guesstimating.”
Should reporters second-guess their health department’s figures? When is it safe for me to retweet? I set out to talk to some experts and learned just how fuzzy those forecasts can be. But there are ways we can think about them that help give us more clarity.
It’s impossible to know how many people are infected right now. Just know that the number is far higher than the running case count.
It’s easier to do big, sweeping projections on a question like how many people will be infected by the time this is all over and done with. Marc Lipsitch, head of the Harvard T.H. Chan School of Public Health’s Center for Communicable Disease Dynamics, has been running projections to figure out how many adults across the world will be infected before a vaccine hits the market (one won’t be available for at least a year) or herd immunity kicks in — when enough people have developed immunity to the virus, from having caught it, so that it can’t easily be transmitted any more. He concluded that between 20% and 60% of adults worldwide will ultimately get infected. (ProPublica has created a tool based on his models and it’s worth checking out.)
It’s much harder to figure out how many Americans are infected at this very moment. Some states have little information because they’ve barely started testing and don’t know when community transmission began in specific places. Many people with symptoms consistent with COVID-19 were turned away from getting tests.
Caitlin Rivers, a computational epidemiologist at Johns Hopkins Center for Health Security, said she’s “not sure we’re in a place where we’re able to estimate that,” though she’s confident that there are far more infected people than the number of reported cases at this time. Computational epidemiologist Maia Majumder said that without widespread testing, this is a “really challenging question to pin down.”
In short, the best way to actually know is with hard evidence — which means testing. As World Health Organization director general Tedros Adhanom Ghebreyesus said on Monday: “You cannot fight a fire blindfolded, and we cannot stop this pandemic if we don’t know who is infected. We have a simple message for all countries: Test, test, test.”
Think in orders of magnitude. The numbers are less precise than they seem.
Of course, computational epidemiologists do have ways of coming up with ballpark estimates. I asked the experts to discuss this sentence with me, from a March 14 post about New York City’s looming crisis by a Facebook data scientist:
“We estimate that between 1,281 and 2,280 people are infected as of yesterday.”
That range seems to be both wide yet oddly specific. What is a reader supposed to take away from a sentence like that?
Reporting precise numbers allows epidemiologists to replicate each others’ work; rerunning the same equations and arriving at the same results helps to validate findings. Public health officials also need as much information as possible because it’s helpful for decision-making about preparing hospital beds or closing schools.
My concern, however, is that a sentence like that can give readers a false sense of precision, as if it’s possible to know down to a difference of one or two people. When I asked Rivers what a regular person should take from these counts, she said people should focus on the order of magnitude: “This means there’s a couple of thousand people, it’s not 200 or 4 million. It means we’re in the low thousands.”
“If the forecast doesn’t state its assumptions, I’d be wary,” Majumder said. For example, a popular modeling approach, called the SIR model, assumes that each individual is equally likely to come into contact with any other individual in the population; you don’t have to have a Ph.D. to see how that’s not very realistic.
Majumder also noted that just because someone is good at math doesn’t mean that they’re equipped to do epidemiological forecasting, which comes with its own nuances. “Has the author published epidemiological modeling studies – preferably of other emerging infectious diseases – before in peer-reviewed literature?”
My advice: Check who the numbers are coming from before repeating them.
Don’t get hung up on the specifics. The big picture is clear.
Trying to get clarity on exactly how many people are infected in your city shouldn’t be your goal, if you’re a regular member of the public. Rivers and Majumder agreed on this: There’s no difference in what action you need to take, whether the models say there will be 10,000 or 20,000 infections in your state within a certain number of days or weeks.
There isn’t a single expert I’ve talked to who said case counts won’t continue to soar. There are two reasons for this: As testing becomes more available, cases that already exist will be revealed. Secondly, of course, the virus is continuing to spread. The trends are crystal clear, and the call to action is indisputable. ”If your state has reported community transmission, the message is the same no matter the number of cases: engage in social distancing immediately,” Majumder said.
