New Model Shows COVID More Widespread, Less Severe

S E E D E D   C O N T E N T

April 18, 2020—Infection with SARS-CoV-2, the coronavirus that causes COVID-19, is already much more widespread in the U.S. than current testing data suggest, according to a new report.

The report, which uses CDC data of cases of influenza-like-illness, or ILI, estimates that at least 8.7 million people were infected across the U.S. during the 3-week period they studied in March. (Earlier, the researchers had estimated it could be as many as 28 million, but revised it when they re-examined the data after publication.) The research has not yet been peer reviewed.

In comparison, as of March 28, the CDC had reported more than 122,000 confirmed cases.

The researchers say their findings are in line with another new study that looked for COVID-19 antibodies in the blood of 3,300 Santa Clara County residents. Based on their tests, researchers estimate that between 2.5% and 4.2% of county residents have antibodies against the virus, a number that translates into 48,000 to 81,000 infections, or 50 to 85 times as high as the number of known cases. That study also has not been peer reviewed.

For their study looking at ILI rates, the researchers looked at several sources of information to determine the infection rate:

• Every week, about 2,600 U.S. health care providers report the numbers of patients who have ILI to the CDC. These patients have a fever of at least 100 degrees, a sore throat or/or a cough, without a known cause other than flu. Colds, flu, other respiratory viruses, and COVID-19 can all cause these symptoms.
• The researchers then excluded people who were eventually confirmed to have flu.
• Of the remaining group, they assumed that those numbers above the season average of the nonflu cases could be attributed to COVID-19.

Study co-author Justin Silverman, MD, PhD, assistant professor of information science and technology at Penn State University, cautions that the data about infection rates needs to be verified yet by other methods, such as testing blood samples and taking swabs.

The findings point to the need for more testing, and not just of those showing up at the doctor with symptoms, says Alex Washburne, PhD, co-lead author of the study and a research scientist at Montana State University. What's needed, he says, is random population testing.

"I would like to see 100 random people in each state on each day" tested, he says, both with swabs and antibody testing. "That would give us a sense of prevalence."

Ruiyun Li, PhD, a research associate at the Imperial College London, who has studied undocumented COVID-19 infections, agreed that the findings stress the need for more testing. "The finding is important as it indicates the urgent need to improve the testing and detection capability - the more we test, the more cases could be confirmed," Li says. He also said their approach is interesting because it's impossible to test everyone. "The estimates based on current ILI surveillance system could help."

The lack of testing is one of the reasons many countries have employed social distancing measures to prevent the spread of the COVID-19 infection. In the U.S., officials have expressed optimism that ramping up widespread antibody testing may help relax distancing measures.

Li echoes the researchers' caution that the estimates are just that - estimates -- and the findings do need to be validated by blood tests to show exactly how many people test positive.

Death Rate Questions

The finding of much more widespread infection suggests the infection to fatality rate from COVID-19 might be less than current estimates, which range from 3% from the World Health Organization to 1.4% reported recently by other experts. But the researchers stop short of estimating a revised death rate from their findings.

Fatality rates also depend heavily on how overwhelmed hospitals get and what percentage of cases are tested. The New York Times reported that China's estimated death rate was 17% in the first week of January, when Wuhan was in its peak, but only 0.7% by late February.

Washburne says while their findings suggest a lower fatality rate is possible, it's premature to calculate a death rate from the infection rate the researchers found. Death rate, he says, ''is probably the most important and contentious number for assessing the risk-benefit of costly interventions, such as statewide lockdowns versus more diffuse interventions such as mask wearing."

One bit of missing information may throw off the estimate of the death rate, Washburne says. "If we're off by 0.1%, that's 200,000 deaths."

Washburne says what they've done is added a piece of evidence that suggests the growth rate of the virus is faster but that the number of serious cases is fewer. He hopes he can help guide future models if backed up by other evidence.