By Kevin Roche
Most experts assume that some multiple of the number of people who have tested positive for coronavirus disease have actually been infected. Testing has been episodic at best and generally limited to people with symptoms or who believe they were in contact with an infected person. Without testing, people have come up with some innovative ways to guess about true infection rates, like the studies we reported on looking for virus in sewage in Massachusetts and several studies that looked for it in blood donations.
A paper at Medrxiv used another technique to estimate potential infection rates. (Medrxiv Paper) The researchers used existing surveillance networks for influenza to identify cases of influenza-like illness (“ILI”). They quantified the background level of influenza-like illnesses and compared this with rates seen in the late winter and early spring of 2020.
They found that there was a surge in influenza-like illnesses. For example, in Oregon in the second week of March there were reports of 50% more ILI cases than in any previous year for the same period. Now it is possible that as people became aware of coronavirus they were going to see the doctor for illnesses they would normally just treat at home. But the increase in ILI cases tracked patterns of confirmed coronavirus infections and deaths. Based on the researchers’ analysis, they believe there were as many as 9 million excess cases of influenza-like illness, many or most of which could have been coronavirus infections. This would mean that only about .75% of all coronavirus cases are being detected in the United States. It also indicates that case numbers are doubling about every 2.5 days on average across the US (note that this would be based on data before most of the lockdowns had begun, which obviously would slow that doubling time).
If this research is accurate (and remember these are early release papers that have not been peer-reviewed and should be subjected to confirmatory research), it would obviously mean that we are making much faster progress toward building population immunity (assuming these infected individuals developed antibodies) and it would lead to much lower estimates of case fatality rates. I looked on the Centers for Disease Control website, which operates the influenza-like illness surveillance network and learned that, indeed, there have been elevated reports of ILI this year.
With 25,000 current deaths attributed to coronavirus, even if you assumed there were only 9 million infections (given that a month has passed, even with mitigation measures, if this data is accurate there would be many more infections by now, potentially in excess of 20 million), the case fatality rate would be .27%. There will be more deaths, but also many more infections, so expect that CFR to ultimately drop further. It also suggests that almost all infections are asymptomatic or mild illness. This is an interesting paper from a technique perspective and if accurate, very hopeful in terms of the seriousness of the epidemic.