More Thoughts on Computing the COVID-19 Fatality Rate By Andrew C. McCarthy
It’s based on decisions about whom to include or exclude, which are often conjecture.
On the Corner last week, I groused a bit about the difficulty of tracking the coronavirus fatality rate. It appeared to be hovering at a bit over 1 percent in the United States. But those appearances can be deceiving.
The elusiveness, I noted, was evident from an observation by Anthony Fauci, the esteemed immunologist of the National Institutes of Health and President Trump’s White House Coronavirus Task Force. Writing in the New England Journal of Medicine in late February, Dr. Fauci hypothesized that the fatality rate may be “considerably less than 1%” because many people who are infected experience either no symptoms or very mild symptoms and therefore do not report. The fatality-rate statistics are skewed toward the people who do report.
The question naturally arises: How much less than 1 percent could the fatality rate be?
More specifically, could the fatality rate for the coronavirus disease that sprang from China late last year (as our Jim Geraghty has comprehensively documented) approach a figure as low as the fatality rate for influenza? The question is important. President Trump frequently touts a comparison of the new coronavirus to flu. Americans longing to return to a semblance of normalcy — understandably so, given the gargantuan ruin the lockdown is causing — complain that closing the country due to coronavirus is overkill, since we don’t do it for flu.
Regrettably, I reckon the answer must be that even if the coronavirus dipped perceptibly below 1 percent, it would still be much worse than flu. Why? Because none less than Dr. Fauci (among others) says so. Though he recently wrote that the rate could be “considerably less than 1%,” he has also recently testified, in a House hearing, that the novel virus from China has a “mortality rate of ten times” that of seasonal flu. He put the latter at 0.1 percent, which would rate the new coronavirus at 1 percent.
Confused? Me too.
Don’t take that as a jab at Dr. Fauci. To the contrary, I am impressed by his paradoxically confident humility. Fauci is first and foremost a scientist. When he doesn’t know, he says he doesn’t know. When he does “know,” he allows that this knowledge reflects nothing more or less than our best current information; more data is always welcome, and we must be willing to check our premises.
My point is that Fauci is a scientist, and I am not. If he is humble about our state of knowledge, I need to be humility on steroids. That said, my training is in proving things, or disproving them, as the case may be. I’m supposed to be able to do that across disciplines. At a certain basic level, facts are facts, even if there is more complexity in epidemiology than in, say, securities fraud. Regardless of the subject matter, the proof business calls for a healthy dose of skepticism about claims of certainty and a jaded ear attuned to bias in such claims — does the proponent have a motive to make things seem better or worse than they are? To weigh competing claims, one must also insist on apples-to-apples comparisons. That is not always easy because claimants tend to talk past each other, never agreeing on the meaning of the terms they spout.
Thus has my amateur obsession with epidemic fatality rates led to questions that, more often than not, lead to still more questions, not confident answers. It is a healthy reminder of how much I don’t know, and of how perilous it can be for analysts to step outside their lanes. Let me, in any event, share some of what I’ve learned — or, at least, think I’ve learned.
There are various kinds of coronaviruses. Specialists in this area take pains to distinguish a virus from a disease caused by the virus. The distinction is salient, but that does not mean it is well accounted for statistically — at least the way statistics are commonly kept. That is, not everyone who gets the virus gets the disease; but at least with the new coronavirus, the federal government (which is leading the response) assumes that if you’ve got the virus, you’ll get the disease.
On the distinction between the two, HIV/AIDS is a good analogy. Acquired Immune Deficiency Syndrome is the disease caused by Human Immunodeficiency Virus; but not everyone who tests HIV-positive will develop AIDS. With respect to our current coronavirus pandemic, the disease is labeled COVID-19 (shorthand for corona virus disease, discovered in 2019); it results from exposure to the virus called SARS-CoV-2. The latter label, the World Health Organization explains, stands for “severe acute respiratory syndrome coronavirus-2.” The “2,” as you might suspect, reflects genetic linkage between the novel Chinese coronavirus and SARS, which caused a frightening outbreak in 2002–03.
Not everyone who contracts the SARS-CoV-2 virus will develop the COVID-19 disease. This is where things get murky in the public debate, and why I say the difference between virus and disease is not necessarily discernible in the statistics washing over us.
When Dr. Fauci wrote that the COVID-19 fatality rate may be well under 1 percent when one factors in “cases” involving asymptomatic or only mildly symptomatic people, I assumed that he was talking about people who get the virus but do not report. But beware of that promiscuous word cases.
A person who does not report is not a case in the familiar sense — a case is a person who reports, is tested, and is treated if necessary. Therefore, while we assume there is a group of non-reporting people out there — perhaps a very large group — who have SARS-CoV-2 (and perhaps even have mild cases of COVID-19), there are also non-reporting people who do not have SARS-CoV-2 — they just have analogous symptoms (the kind common to cold, flu, other viruses, etc.). Adding non-reporting people for purposes of computing the COVID-19 fatality rate could thus unduly inflate the denominator and understate the danger. On the other hand, limiting ourselves to only reported cases in which patients test positive for the virus would miss people — probably a lot of people — who have the virus but do not report and quickly recover (although maybe not before they’ve passed it along to others). This would wrongly inflate the fatality rate higher (by depressing the denominator).
And then there is the matter of how stats are kept. According to the Centers for Disease Control, the test that is being administered “is designed to detect the virus that causes COVID-19” — viz., SARS-CoV-2. Does a positive test indicate that the person has the virus but may not have the disease? Not clear. The CDC elaborates, “If you have a positive test result, it is very likely that you have COVID-19” (emphasis added). Meaning: The CDC (at least in the statistics that are being shared publicly) assumes that if you have the virus, you have the disease.
Mind you: I am not criticizing the CDC. Given that a person who has the virus can pass it along to someone else, who could then develop the disease and potentially die, it makes perfect sense to treat these two discrete things as if they were the same — even though the virus could be so mild as to cause no symptoms, and the disease could be so severe as to cause death. All I am saying is that this mucks up the mortality math.
To summarize, we have (a) people who get COVID-19 and have reported; (b) people who get the SARS-CoV-2 virus and have reported, and who may not necessarily develop COVID-19 (although they are counted as if they have done so); (c) people who report because they fear they have the virus, but who test negative (so we assume they have not had the virus); (d) people who have the virus but do not report (meaning we can only estimate how many such people there are); and (e) people who have other conditions that could be mistaken for the virus and who do not report (meaning that counting them would skew our already uncertain estimate of virus cases that have gone unreported).
Trying to figure out an accurate fatality rate depends on how we count these different categories. It is a tricky business, rife with uncertainty. Yet it’s what must be done in order to grasp whether we are dealing with something that is just marginally worse than flu, or rather is significantly worse, even exponentially worse. Things appear deceptively dire if we calculate death rate solely by reference to reported COVID-19 cases; but the picture is deceptively benign if we measure deaths against an inflated conjecture about the non-reporting population.
I suspect that this explains the ostensible contradiction between Dr. Fauci’s two comments on the fatality rate. Context is everything. In his House testimony, Fauci was confronting a political narrative that COVID-19 is really no worse than flu, and reliance on reported cases showing a 10:1 difference was a useful rebuttal. In his essay, he was trying to demonstrate to other scientists (and whatever sliver of the public reads medical journals) the imprecision of our knowledge about exactly how much worse COVID-19 is than flu.
So, let’s pick up where I left off last week, with the day-to-day death rate, computed by comparing each day’s new deaths with new reported cases — which, as I acknowledged, is itself a rough and imperfect comparison.
In the six days between March 20 and March 25, new deaths continued hovering around 1 percent of newly reported cases. But as reported cases rose dramatically from 5,588 to 13,355 (as expected, due to the spread of the virus and much more widely available testing), there was a slight uptick in the death rate. On a day-to-day basis, it varied from a low of 0.9 percent on March 20 (49 out of 5,588) to a high of 2.0 percent on March 24 (225 out of 11,075). Meanwhile, looking at the national numbers since the outbreak started, Worldometer reports (as of midafternoon Thursday) that the fatality rate remains 1.4 percent (1,136 out of 78,796).
To repeat, that is based on reported positive cases. That is, if we know you have the virus, there is a 98.6 percent chance of survival (good odds, but not nearly as good as flu, with its 99.9 percent survival rate).
This does not account for negative tests — i.e., people who feared they might have COVID-19, got tested, and were negative. According to the COVID Tracking Project, there have been 367,143 negative tests (as of midday Thursday). As related above, the CDC says a positive test means one has the virus and, we should assume, is likely to have COVID-19. (Again, the negative tests do not account for people who have had the virus but did not develop COVID-19.) The total number of tests administered (positive and negative) is 432,655. That is a bare fraction of our total U.S. population.
Some perspective, then: Out of 330 million Americans, fewer than 1 percent have been tested. While testing should have been more widely available early on, it is increasingly available now, and of those who have been tested, the vast majority (85 percent) have tested negative. COVID-19 is infectious and can be deadly, so it cannot responsibly be sloughed off. That said, the percentage of Americans who’ve gotten it is tiny, and it has been fatal for just a thin fraction of those.
The biggest problem may not be the fatality rate, however marginally over or under 1 percent it turns out to be. It is that about 10 percent of those who have tested positive for the virus have required hospitalization — 6,322 out of 65,512, by my last count of the COVID Tracking Project. If this percentage holds steady, and the number of positive reported cases were to rise into the hundreds of thousands and then the millions, our medical system would be overwhelmed. That could result in a higher fatality rate not just for COVID-19, but also for other medical conditions that will exacerbate without treatment. That is why the stress on social distancing and “bending the curve.” What we’ve really got is less a disease challenge than a health-care management challenge.
Comments are closed.
