The latest results of mass antibody testing in San Miguel County, Colorado, suggest that 1 to 2 percent of the local population has been infected by the virus that causes COVID-19.* The early results from the program, the first of its kind in the United States, underline the point that official counts of COVID-19 cases, which are limited to people who are sick enough to be tested for the virus, greatly understate the total number of infections.
United Biomedical, a company based in Hauppauge, New York, is collaborating with the San Miguel County Department of Health and Environment to test all 8,000 or so residents of the county. So far the company has drawn about 6,000 blood samples, although analysis has been delayed because of limited laboratory capacity.
As of today, the county had received 2,583 test results, of which 17 were positive and 43 were borderline.* Assuming that the initial samples are representative of the county's general population, that suggests between 0.7 percent and 2.3 percent of residents have been infected by the COVID-19 virus, including people who had mild or no symptoms.
Applying those rates to the U.S. population implies 2.3 million to 7.5 million infections nationwide, compared to about 700,000 confirmed cases so far.* There is reason to think the nationwide prevalence is higher than the rate in San Miguel County, a sparsely populated area of southwestern Colorado with 5.7 people per square mile, compared to the U.S. average of 94. Even in Telluride, the county's seat and biggest city (with about 2,500 residents), there are 1,120 people per square mile, compared to about 4,000 in Dallas, 4,700 in Denver, 12,000 in Chicago, 19,000 in San Francisco, and 28,000 in New York City.
Leaving aside differences in population density, the San Miguel County results suggest that the true number of infections in the United States may be three to 11 times as high as the number of confirmed cases.* Similarly, researchers at the University of Gottingen in Germany recently estimated that the official U.S. count as of March 30 excluded more than 90 percent of infections. Australia's chief medical officer likewise suggested this month that the global case tally was off by a factor of about 10. A study based on unusually high numbers of influenza-like illnesses, by contrast, estimated that official COVID-19 counts as of late March included just one out of 100 symptomatic cases in the United States.
Virus tests of volunteers from Iceland's general population have come up positive about 0.7 percent of the time so far. That result, which is similar to the low-end estimate for San Miguel County, is based on nearly 24,000 tests in a country with a population of about 340,000. It does not include people who were infected but no longer carry the virus, who would be identified by antibody tests.
Iceland reported its first confirmed COVID-19 case on February 28, more than a month after the United States did. And unlike the United States, Iceland has combined wide testing with aggressive contact tracing and targeted quarantines. It therefore seems likely that the prevalence of infection is higher in the United States.
The actual prevalence of the virus is crucial in estimating not just the risk of infection but the odds that it will kill people who catch it. Although the overall fatality rate is obviously relevant in deciding how to fight the epidemic, current estimates cover an absurdly broad range because we have no idea how many people have been infected. "Fatality rates based on comparing deaths, which are relatively easy to count, to infections, which are not, almost certainly overestimate the true lethality of the virus," The New York Times notes. "Health officials and epidemiologists have estimated there are five to 10 people with undetected infections for every confirmed case in some communities, and at least one estimate suggests there are far more."
Clarifying these issues will require mass testing that goes beyond a single rural county in Colorado. The Centers for Disease Control and Prevention, which helped engineer the testing fiasco that has left us ignorant of basic facts about the COVID-19 epidemic, recently began promising "wide studies of community transmission" based on antibody tests.
[*This post originally used test results that had been reported as of Tuesday; the numbers have been updated to reflect results reported as of April 17.]