COVID-19 fatality rate, calculated from S. Korean data

by Dave Burton

April 16, 2020, 11pm  (updated April 27, 2020)

To calculate the true COVID-19 fatality rate for a country like the United States, where many cases are undiagnosed, and where many of the infections have not run their course (so we don't know how many will recover, and how many will die), is challenging.

But the calculation is not challenging for South Korea. The South Koreans have identified nearly all of their COVID-19 infections (including the asymptomatic cases), and over 70% of their cases are resolved (either recovered or died), so we can easily calculate the approximate true fatality rate.

(click graph to enlarge)

South Korea uses aggressive testing and contact tracing to find nearly everyone who is infected, regardless of whether they have symptoms. We know from the dwindling infection and death rates there that they have been almost 100% successful. Over the last week they've averaged just 30 new cases and 3.6 deaths per day, in a country of 52 million people. That means there cannot be more than a few dozen infected people in South Korea who have escaped detection, and so have not been quarantined.

(For a good description of how South Korea has managed that feat, see this valuable article, starting around the 15th paragraph.
April 19, 2020 UPDATE: Here's another article about how South Korea did it.
April 20, 2020 UPDATE: Taiwan has also been very successful at combatting the epidemic; here's an article about how they did it.)

Numbers updated April 27, 2020:
As of today (April 16  27, 2020), South Korea had reported:

10,613  10,738 known cases
7,757  8,764 recoveries
229  243 deaths
2,627  1,731 active, unresolved cases

(As a result of contact-tracing, they also have about 20,000 people in isolation, with no symptoms, but who are suspected of having possibly been exposed, and are awaiting test results; the vast majority of them will test negative.)

From those statistics we can calculate:

● 243 deaths / 10,738 cases = 2.16  2.26% naively-calculated fatality rate. That's the lower bound on the true fatality rate (because the denominator includes 2,627 1,731 active cases, some of whom will probably die).

● 243 deaths / (8,764 recoveries + 243 deaths) = 2.87%  2.70% fatality rate. That's the upper bound on the true fatality rate (because, on average, recoveries take longer than deaths).

South Korea is a first world country, with excellent healthcare and communication infrastructures, and a healthcare system that is not overloaded, and they identify COVID-19 patients very early, enabling early treatment, and their doctors have plenty of experience treating those patients. In other words, for a patient, it's one of the best places in the world to contract COVID-19. But even there, with all those advantages, the true COVID-19 fatality rate in South Korea is between 2.26% and 2.70% (2.48±0.22 %).

In most other places, including the United States, it is presumably worse.

That fatality rate compares to a U.S. typical seasonal flu fatality rate of 0.13%, based on these figures:

So, even with the very best of care, COVID-19 is about twenty times as deadly as a typical seasonal flu.

With less than the best care, i.e., in most places, it's presumably even worse than that. So unless an effective treatment is found, most countries, including the United States, can expect a COVID-19 fatality rate of at least 2.5%. (In the United States, the current naively-calculated case fatality rate is 34,619 / 677,570 = 5.11%, but that's presumably inflated by the fact that 50% to 80% of our COVID-19 infections are probably undetected.)

April 20, 2020 FOOTNOTE

“Expert” estimates of the true number of cases of COVID-19 in the United States vary by more than a factor of ten. However, we can work backward from the number of fatalities, and the best-case (South Korean!) fatality rate, to estimate the true number of cases.

When we do that arithmetic, we find that for every known COVID-19 case in the U.S. there have probably been between 1.1 and 4.6 undetected cases (2.85 ±1.75), which were (or are so far) either completely asymptomatic, or had (or have) symptoms too mild to require hospitalization.

Based on the current (April 20, 2020, AM) number of known U.S. cases (764,265), that calculation indicates that the actual number of COVID-19 cases in the U.S. is between 1.6 & 4.3 million (i.e., 3.0 ±1.4 million), which is 0.9±0.4 percent of the U.S. population.