More than 11 million people in the United States have tested positive for the coronavirus, but the true number of people that have been infected is unknown.
For every infected person who gets a positive test result, it is believed that more go undetected because they don’t bother to get tested, have only mild symptoms, or never experience symptoms at all.
Yet, the true number of infections is a critically important figure.
Estimates of the total number of COVID-19 infections are used to gauge how far along states and the country are in the pandemic and how many people are vulnerable to the disease. It also indicates whether the population might be gaining protection from the virus in the form of herd immunity — that is, when enough people have gained immunity through exposure to the virus to slow its overall spread.
There is no official estimate of the total number of infections. The closest such figure is a remark made by Centers for Disease Control and Prevention Director Robert Redfield in a September congressional testimony that only about 10% of the country had been infected and that 90% of the population remained vulnerable. Redfield’s statement, the agency told the Washington Examiner, was based on testing performed at sites around the country through August in which the agency took blood samples and examined them for the presence of antibodies to the coronavirus.
The agency said that further such analyses would be released soon. In the meantime, outside analysts are left to guess at the total number of infections based on whatever data they are able to access.
One method: Using deaths
Some researchers have produced estimates of overall infections using a model based on the number of deaths. They use the death count because it is, in a sense, a “hard” number that is relatively easy to record — unlike infections, which depend on testing.
“We start with death estimates, then work backward, using infection fatality ratios to estimate infections based on deaths,” said Amelia Apfel, the media relations officer at the Institute for Health Metrics and Evaluation at the University of Washington.
In other words, the institute looked at specific areas hit by COVID-19 for which surveys of infections based on blood samples were available. It then compared the total number of deaths relative to the number of infections — not recorded cases, but total infections, including asymptomatic ones — suggested by the blood samples to determine infection fatality ratios.
Once the infection fatality ratio is established, researchers can plug in fatalities for other areas, such as other cities and states or the entire country, to arrive at a number for infections.
The IHME estimates that about 41 million people in the U.S. have been infected, or less than 15% of the population.
Such calculations, though, can vary significantly depending on the fatality rate assumed. For example, the Imperial College of London recently calculated that the infection fatality rate for the coronavirus was 1.15%, with a range of 0.78%-1.79%. Currently, the number of confirmed COVID-19 deaths in the U.S. is around 251,000. If the number of deaths is 1.15% of total infections, then the total infections in the U.S. would be just 22 million. However, if the number of deaths were 0.78% of infections, then total infections would be over 32 million. And other researchers have found that the infection fatality rate is even lower.
One factor adding to the complexity of determining true infections is that the fatality rate has dropped over time as hospitals have improved treatments.
Second method: ‘Capture-recapture’
Another way to calculate total infections is through a method that epidemiologists have taken from ecology, used to estimate animal populations that cannot be directly counted. It’s called the “capture-recapture” method.
Here’s how it works for ecologists: If a forest manager wants to know how many bears live in the woods, he would set out 10 bear traps. Once he caught 10 bears in the traps, he’d mark them and release them back into the wild. He’d then reset the traps. Once 10 bears were caught again, he’d check to see how many were marked. If four were marked, or recaptured, he could then estimate the total bear population in the forest based on the distribution of bears caught twice relative to bears caught twice (in this example, the estimate would be 25).
With a little more statistical workmanship, the same concept can be used to estimate the population of people infected with the coronavirus, substituting testing for trapping.
Dankmar Bohning, a professor of medical statistics at the University of Southampton, and his colleagues employed that method in a study of coronavirus infections in Europe for the International Journal of Infectious Diseases.
They estimated that the total number of cases was 2.3 times higher than the official number. In the U.S., that would mean that there are over 25 million cumulative infections in the U.S. However, Bohning said that new research he has conducted estimates actual cases as many as 4.5 times higher, which would mean there are almost 50 million cases in the U.S., not far off from the IMHE estimate.
Third method: Positivity rates
Still, a third method of gauging true infections is to compare cases to test positivity rates. In general, lower positivity rates indicate that testing is picking up most of the infections. Higher rates indicate that more infections are not being detected.
The site covid19-projections.com, run by data scientist Youyang Gu, developed a model for estimating true infections based on confirmed cases and positivity rates. A new model unveiled this week suggests that new case counts are picking up only about 1 in 3 infections. It also estimates that around 49 million people, or 15% of the U.S. population, had been infected as of Nov. 8 — but that the correct number could be as much as 22% or as little as 10%.
Some states have been particularly affected. North Dakota, for instance, had nearly 30% of its population infected by early November, according to the model. Tiny Vermont, on the other hand, had less than 2% infected.
It appears that in some places, such as Sun Belt states that saw huge spikes in the summer that then subsided, enough people gained immunity through infection that transmission of the virus has slowed down.
Although the CDC maintains that the “vast majority” of the population remains susceptible to COVID-19, some officials and epidemiologists disagree.
Dr. Scott Atlas, the neuroradiologist and critic of pandemic restrictions who has been advising President Trump on the pandemic, has said that Redfield is mistaken. Instead, Atlas has argued, people who have been infected with cold viruses similar to the coronavirus have developed protection in the form of T-cells that can fight the coronavirus. Those combined, with perhaps 20%-25% infected with the coronavirus, mean that far fewer people are susceptible to the virus.
That view is shared by, among others, Oxford epidemiologist Sunetra Gupta, one of the signers of the “Great Barrington Declaration,” which concluded that non-vulnerable people “should immediately be allowed to resume life as normal.” Gupta and a team of Oxford researchers calculated that enough people have preexisting protection against the coronavirus that the population herd immunity threshold might be as low as 10%.
However, many epidemiologists dispute that, arguing that it is not yet known whether T-cells related to cold viruses provide immunity from the coronavirus. Rather, they believe that 60%-70% will need to be infected with the coronavirus to achieve herd immunity. That is based on models and on places like Manaus, Brazil, where the spread of the disease did not stop when 20%-25% of the people were infected. Between 44%-66% of Manaus may have been infected.
