Medical Examiner

Science by Press Conference

Andrew Cuomo shared the results of widespread antibody testing in New York on Thursday. Here’s what they mean.

Andrew Cuomo sits at a table and raises his hands.
New York Gov. Andrew Cuomo gives a press briefing about the coronavirus crisis. Matthew Cavanaugh/Getty Images

Slate is making its coronavirus coverage free for all readers. Subscribe to support our journalism. Start your free trial.

On Thursday morning, Gov. Andrew Cuomo announced the preliminary results of antibody testing efforts in New York state. Of 3,000 people, recruited as they shopped in stores in 19 counties, nearly 14 percent tested positive for antibodies, suggesting they have had and recovered from the novel coronavirus. Among those tested in New York City, that number was 20 percent. That puts total infections in the state at 2.6 million, and the death rate at 0.5 percent.

It is hard to sit in Brooklyn, read those numbers, and not feel hopeful, like there is some significant chance that I or my loved ones have already been quietly infected and spared. But what do those numbers actually mean? How much should we trust them at all?

Fundamentally, it’s simply hard to tell. A politician on television is not the ideal way for scientific data to make its way into the world; at the very least, there should have been a methods section ready in the wings. “Sorry, Governor, but that is BS,” tweeted Florian Krammer, a virologist at Mount Sinai Health System in New York City. “Which test is this based on?”

Antibody tests produce false positives and false negatives—for example, by picking up signs that a coronavirus has been in someone’s body, perhaps one that causes the common cold, and mistaking them for the coronavirus, as in the one that is wrecking everything right now. Without more information on which test was (or tests were) used, it’s hard to know what the error bars on the estimates should be. We don’t have that much information at all on the study, which was, again, briefly summarized on TV. “It would be helpful (for scientists, essential) to have the details on methods in order to evaluate a study, including the sampling plan and the test characteristics,” Stephen Morse, an epidemiologist at Columbia, wrote me in an email.

There was one clear shortcoming in the methodology, which Cuomo acknowledged in the press conference: The sample isn’t an exact representation of the population. People who are out shopping are not likely to be those who are stringently isolating and therefore less likely to catch the virus. They definitely aren’t people who are incarcerated or in retirement homes, where infections rates can be high.

The study did succeed at capturing data from many places in the state, which requires some ingenuity. Natalie Dean, a biostatistician at the University of Florida, is one of the experts who participated in a phone call to help a New York state task force design the study. “We were trying to suggest cheap and easy strategies,” she says. The stores idea was raised on the call by Bethany Hedt-Gauthier at Harvard. While Dean notes that it could select for people who were eager to participate because, say, they’d had a nasty cold and wanted to check whether it was the coronavirus, stopping people is much better at eliminating this bias than recruiting subjects via Facebook, as an antibody survey in Santa Clara county did.

Both Morse and Dean said that the numbers from the study and estimates for statewide infection made sense to them, even though they might feel high to the average person. (They do to me.) Based on data from other countries, and the fact that many people have not been tested and are riding out the virus at home, it’s clear that the numbers are much higher than the number of cases that we’re seeing in hospitals and in official coronavirus counts. “I’ve generally used as a rule of thumb that the confirmed cases would be perhaps 10–15 percent of the actual number infected,” wrote Morse. By that rule, with about 260,000 cases confirmed in New York State, 2.6 million infections statewide, as the testing suggests, makes sense.

And there’s one important point that the data Cuomo shared bears out: The virus is disproportionately affecting people of color. Nine percent of white people tested had antibodies, while just over 22 percent of black and Latino people had them. “This reflects more the regional breakdown,” Cuomo said Thursday, pointing out that there are more white people upstate, where infection rates are lower. Still, the survey should underscore that everyone is not suffering equally from the virus; among other things, being rich enough to flee an area with a lot of infections helps.

Aside from feeling mad, and maybe hopeful, what we should individually do with this information is nothing. An infection rate of even 20 percent, the estimate from this survey for New York City, is “still a ways from herd immunity,” noted Dean. Roughly 70 percent of us would have to have been infected and have immunity before enough people have had the virus that it stops reliably spreading around the community. Even if the survey data were not preliminary (which it is), and even if we knew enough about coronavirus immunity to say for sure that you can catch it once and not again (which we do not), we would still basically stuck inside. For now, all we can do is watch the slow crawl of science play out on television.