The real number of coronavirus cases in the United States could be 6 to 24 times greater than the confirmed number of cases, according to a large federal study conducted by The US Centers for Disease Control and Prevention (CDC) based on data from 10 cities and states across the country.

The study, published Tuesday in JAMA Internal Medicine, was based on serological tests: blood tests that look for antibodies against the virus and that determine if someone was previously infected. They are different from diagnostic tests, which only detect people who currently have the virus.

Serological tests

Overall, it is estimated that 1% of people in the San Francisco Bay Area have had Covid-19, while 6.9% of people in New York City have had it, according to the authors of this study that includes researchers at the US CDC. In 7 of the 10 locations analyzed, the estimated number of cases was 10 times the number of reported cases.

The study was based on evidence from more than 16,000 people across the 10 locations, but with one limitation: It is based on old data. Samples from San Francisco were collected from April 23 to 27, while tests from New York were conducted from March 23 to April 1. The last tests were carried out in May … that is to say, that many things can be during two months in the course of an outbreak.

In South Florida, for example, researchers estimated that 1.9% of the population had antibodies to the virus. But that number is based on samples collected April 6-10, and given the spread of the virus since then in the state, the number now would certainly be a larger number.

Still, the data reflect what the CDC director, Robert Redfield, recently said: that the true numbers of cases are 10 times higher than the confirmed diagnoses. Confirmed cases in the United States amount to more than 3.8 million.

The data underscores two other points: that tests in the United States do not capture the full extent of the outbreak, and that even the worst-affected communities are nowhere near reaching a herd immunity threshold.