Why We Don’t Know How Many Americans Are Infected With Coronavirus—and Might Never Know
Sparse testing is just one reason the official tally is far too low, but the numbers will get more reliable over time
The number of known infections with the new coronavirus has grown rapidly in the U.S., with more than 270,000 cases as of Friday. More than 7,000 have died. But there are all kinds of holes in those numbers.
People who got sick—or died—before testing was widely available don’t show up in the tallies. Even now, many people who want to get tested can’t, and some who could get tested choose not to. Some people are tested more than once. That undercount is problematic because it slows policy makers and hampers the public from understanding the full portrait of the epidemic.
From the time of infection, a patient can spend days without symptoms, more time feeling early symptoms and even more time waiting to get tested and then finally getting results.
Positive test counts are always behind the number of actual cases of Covid-19, the disease caused by the virus—often by a week to a month. Severe cases are the most likely to get tested.
Without widespread testing, health officials don’t have enough information to fully answer key questions, such as how many people are infected or the typical transmission rate per person. Hospitals can be blind to the oncoming wave of patients because many people remain untested until they appear in an emergency room.
The picture becomes a little clearer when states report the number of tests that come back negative, in addition to the positive cases, revealing the total number of tests that have been analyzed.
Missing cases can lead to skewed death rates. When countries test only their sickest patients, the death rate can appear extraordinarily high. For instance, Italy, which has rationed tests, had a mortality rate above 12% as of April 3, far above the global average, according to data compiled by Johns Hopkins University.
Further compounding confusion: Italy hasn’t been counting people who died at home. As a result, the chart below, which reflects the official tally of Covid-19 fatalities in Italy, should come with an asterisk to reflect that many more people likely died from it.
Because deaths from Covid-19 are far more likely to be reported than the much greater number of those who got the disease, some academics have tried to work backward from deaths to estimate the volume of unreported cases.
Preliminary research by mathematical modelers in the Centre for the Mathematical Modelling of Infectious Diseases at the London School of Hygiene & Tropical Medicine found the U.S.’s case count likely represented just 14% to 19% of actual infections that produced symptoms. (By contrast, South Korea had identified between 53% and 90% of total cases in the country, the researchers said.)
Local governments vary widely in how they collect and disclose their testing results, which can affect statewide totals. For instance, most states report just positive or negative test results, while others also report tests for which results are pending.
There is also the issue of tests done by private laboratories, working under contracts with hospitals and local health departments. Some states don’t report them or do so inconsistently. Until that data is more complete, state comparisons will be impossible and national figures will be potentially misleading.
Experts say that the goal should be broad testing throughout the population to get an accurate picture of who is infected. Iceland has tested a broad cross-section of its population and found that middle-age people make up 40% of those testing positive. In the U.S., where early testing has focused on people who show symptoms, the age distribution has skewed toward older groups.
Even when tests are given, they may incorrectly identify infected patients as coronavirus-free. Experts have estimated that nearly one in three infected patients receives a false negative test result.
A study in China among patients with Covid-19 symptoms compared the results of lab tests with those of CT scans taken of their chests, which can detect signs of infection in the lungs.
The picture is expected to sharpen in the coming months. New types of antibody tests will allow researchers to determine how many people had Covid-19, even if they were asymptomatic or showed only mild symptoms.
In San Miguel County, Colo., best known for the resort town of Telluride, a local couple who own a medical-testing firm are paying for antibody blood tests of the entire population of 8,000. Such tests will give “us a way to look back and reconstruct the past outbreak,” said Natalie Dean, an assistant professor at the University of Florida’s Department of Biostatistics.
Having the total number of infections would allow officials to calculate the overall infection rate, and by subtraction, what share of the population remains uninfected and vulnerable. It would also allow a closer estimation of how many of those infected develop severe cases or die—information critical in planning for hospital capacity.
Still, clear answers will take years and require careful analysis of death certificates, hospital records and broad antibody testing of the entire population.
And even death certificates won’t reflect the larger toll of the pandemic, as some people who never contract the virus will still die from a failure to get adequate health care, such as those who couldn’t get lifesaving transplants.