“There are three kinds of lies: lies, damned lies, and statistics,” quote popularized in the US by Mark Twain
This is an almost direct follow-up to my latest article for Timmerman Report (“Let it Rip or Shelter at Home?)
Usual caveats apply: I am not an epidemiologist, and not a virologist. This is still a new virus for us as a species so there is no natural immunity. There is still a ton we do not know about the virus and its biology, blah blah blah…
How bad is this virus, really? As we are spending endless days locked up in our apartments (yes, I am running low on pasta), what is the real risk for the individual? This is especially an issue in countries, like the US and some other Western democracies, where individual rights are prized above all, and where there is a natural tendency to be skeptical of government intervention.
Erring on the side of caution (precautionary principle) might sound good in theory, but how can we go on as a society for more months when entire swaths of the economy are basically veering towards bankruptcy as I write?
What is, then, the “real” fatality rate (number of people dying after being infected)? Should we expect a 0.04% fatality rate, or a horrific Northern Italian / spaghetti western-style situation with a crude case fatality rate (CFR) of almost 12%?? (More on these numbers below).
Obviously, there is potentially a very wide range of policy actions depending on this “real” fatality rate. For a great discussion of this (and of everything coronavirus), please follow Kai Kupferschmidt (@kakape, a German journalist who writes for Science and has been following this outbreak from the beginning).
Starting from the premise that every death (according to my Roman Catholic upbringing, at least) should be avoided, especially if in excess to what a “normal” fatality rate is, I would like to offer, first of all, some important clarifications.
The fatality rate for any given infectious pathogen is, as for everything, context-dependent: population demographics (age structure) and overall health conditions (rate of pre-existing conditions, including obesity, diabetes, smoking rates, etc), as well as status of healthcare infrastructure (ICU beds, ventilators, number of nurses, availability of drugs, etc) are all huge factors. This is not a good time to be infected with SARS-CoV-2 if you are an elderly obese male smoker with heart conditions, to be honest.
So, it is to be expected that different countries’ populations, especially since they might well be at different stages in their own outbreak, might see different fatality rates.
I think it might be useful to describe some basic terminology used in the field below.
IFR: Infection Fatality Rate, also referred to above as “real” fatality rate: this is a “simple” calculation (expressed as a percentage): It’s expressed as the number of fatalities divided by the number of infected individuals. NOTE that infected individuals = detected + undetected (asymptomatic / other infected but not tested) infections. The obvious problem in calculating this “real” fatality rate is the denominator: knowing how many people have been infected. When you don’t adequately test across the population, you don’t really know the denominator.
CFR: Case (or “Crude”) Fatality Rate: also a percentage. It is often calculated by simply dividing the number of deaths by the number of confirmed cases. This is obviously a much rougher (hence “Crude”) measure since not every infection leads to disease and not every infected individual is identified and counted (see above about the number of asymptomatic carriers).
Because a limited number of tests have been performed, it is presumable, and even very likely, that a large number of people have been infected with the virus but have not been tested. In this case, the denominator should be MUCH larger, and therefore this virus that you are all worried about is really. Not. That. Bad!!! The latest data published in The Lancet Infectious Diseases on March 30 pegs the death rate from confirmed COVID19 cases at 1.38 percent, and the overall death rate, including unconfirmed cases, at 0.66 percent.
Indeed, if we are overestimating the eventual IFR, (by relying on CFR, and correspondingly underestimating the number of infected patients, the denominator, due to the lack of testing) then CFR will appear much higher than the real one. There has indeed been a recent “surge” in pundits discussing various epidemiology models which suggest the overall infection rate is far in excess of the official numbers (everybody is an epidemiologist lately). If there were so many more people infected than have been detected, fatalities across the entire population would then appear less dramatic and might lead one to reconsider some of the current strict stay-at-home policies.
To be fair to this argument, there are a number of studies, including in the town of Vo’ in Italy as well as from many other studies, that establish that ~50-55% of people infected are asymptomatic (see Timmerman Report “8 Days Later”). It is also true that most countries, with the possible exception of Korea and Germany, have not yet implemented a broad testing program across their population (Germany just announced a study testing 100,000 individuals at random from the community to assess more concretely these very same prevalence statistics).
That said, current very crude CFR rates in Germany are close to 1.2% (and starting to really go up, as the virus starts reaching more of the elderly population). For South Korea the numbers are higher, ~1.7%. Assuming (big assumption here) ~50% of infected are asymptomatic, we might end up with an IFR here of between 0.7-0.9% or close to 1%.
There are a few sources of additional information: the Princess Diamond cruise ship had ~1% fatality rate (7 deaths / 619 infections: I have been REALLY trying to find out if they had tested everybody in the ship, but I could not ascertain that). These are very small numbers, obviously.
Let’s now look at the numerator, the number of deaths. Yes, that should be not much of a discussion, really. After all, “in this world nothing can be said to be certain, except death and taxes” (Benjamin Franklin, in a letter to Jean-Baptiste Le Roy, 1789).
Well, I beg to differ on the former (and yes, I do pay my taxes in the US, thank you very much).
I will not comment on the recent multiple media reports on China having possibly concealed the extent of the coronavirus outbreak, under-reporting both total cases and deaths suffered from the disease. However, there are other statistics that are worth quoting and discussing.
Starting in the UK: the government (finally…) announced today, April 1, the criteria for counting a fatality as related to COVID (2,352 as of 5pm GMT on March 31): only people who tested positive and died subsequently to hospitalization were counted.
Obviously, this undercounts (possibly substantially) the real number of fatalities attributable to the virus (even more so as testing is still not provided at scale in the UK). (source: www.gov.uk). As of 9am GMT on April 1, 152,979 people have been tested, of which 29,474 have tested positive (~19.2%). Using the government criteria, this is a CFR of ~8%. I would really like to know what the number of fatalities in assisted living facilities, as well as homes, has been during the same period versus, say, last year. I think it is very fair to say that the denominator should be higher, but then, perhaps, the numerator as well should be.
If only there were a way to compare overall fatality rates in a defined place affected by the virus now versus, say, a year ago when we were presumably dealing with seasonal flu and other pathogens we have tools to deal with. Interestingly enough, there is: Italy might provide once again some perspective and interesting statistics.
Bergamo (source: ecodibergamo.it) is one of the most severely hit municipalities in Northern Italy (Lombardia). In March 2020, more than 5,400 people have died in the province (6x more than in March 2019), 4,500 apparently due to the novel coronavirus, SARS-CoV-2. Of those, only 2,060 deaths have been officially certificated as caused by the respiratory illness we now call COVID-19 (the entire number of fatalities in Italy officially attributed to the virus is 13,155 as of right now, ~7 pm ET, Apr. 1).
Many of the additional fatalities are in the demographics most susceptible to the infection (elderly), who died at home, or in assisted living facilities. They were never tested for the virus, despite exhibiting the telltale symptoms. The official death cause is reported as interstitial pneumonia (the virus causes pneumonia by invading the lungs: the immune system response to the invasion then causes a severe inflammatory reaction leading to death).
Another source of information, from ISTAT (the Italian statistics agency), highlighted by Matteo Cavallaro (@matheusagaso), reports comparative fatalities data from 1,084 Italian municipalities in March 2020 vs 2019: March 2019: 8,054 deaths; March 2020: 16,216 (~2x).
Those excess fatalities were highly concentrated (4,079 of 8,162 excess deaths) in 4 provinces: Bergamo with 2,043, which roughly tallies with the Eco di Bergamo stats above; Brescia (+879), Milano (+639) and Cremona (+518), all of them in Lombardy, the epicenter of the Italian outbreak.
The increase in fatalities is also more frequent in men (+144% vs March 2019) than women (+79%), and in the elderly (in Bergamo, 1,949 of the 2,043 excess deaths are in the population >65 years old). These excess deaths are largely, and clearly, attributable to the new coronavirus.
Keep in mind: People in Italy, and especially in the provinces hardest hit, were in lockdown since March 9 to flatten the curve as much as possible. So, the overall number of car accidents, heart attacks, etc, which are other causes of mortality, are actually DOWN substantially in March 2020 vs 2019. I could not find the numbers for those municipalities, but I did find out that in Los Angeles (yes, they drive almost as badly as Italians) March car accidents were down more than 26% versus March 2019. For the week ending March 27, when people really started taking the stay-at-home instructions seriously, accidents were down 60%! Simple explanation: when people stay at home, they drive less, and get into fewer traffic accidents. However, at the same time, hospitals in the region were unable to withstand the onslaught of new patients: people were counseled to stay home, until the most severe symptoms emerged. As a result, most people dying in houses or nursing homes will not be tested for the virus.
Finally, there are factors to consider: the concept of “excess deaths” versus a period pre-pandemic. Excess deaths are caused not just by the virus itself, but also by the lack of normal standards of care which are suddenly not available any more to any patients suffering from other illnesses. If the hospital is completely full, and you have a stroke, you could be in a tough spot. There are indeed signs that excess overall deaths are climbing in Italy, especially (again) for demographics over 65.
In conclusion, many people want to know — exactly how bad is this outbreak? Is it worse or better than we thought a few weeks ago? (Note, the word “we” is doing a lot of work here, as there is a lot of variability in outbreak severity and response between countries and even different states / regions within countries).
The simple, terrible math is driven by the number of deaths. They are mounting and increasing the numerator. We are getting a clearer sense now that we know some deaths haven’t been officially attributed to COVID19, but would have been if proper testing had been in place. Now that we can begin to get at the true death toll, and we’re beginning to roll out large-scale testing, a clearer picture is emerging of the infection fatality rate and case fatality rate.
For the time being, I’m basing my answer on the following: 1 case of death not officially associated with COVID-19 “offsets” 50-100 non-tested / asymptomatic individuals (if we believe the real IFR is between 1-2%, which is probably not too far from the truth). So, the numerator and denominator effects described above are not equivalent. Each increase in the numerator offsets A LOT of increases in the denominator.
The realization that a large number of virus-associated deaths went un-reported in several countries like UK, Italy, now France and possibly China should strengthen even more our resolve. We need to stick to the physical distancing orders in place to limit the virus spreading as much as possible, at the same time we work at breakneck speed to increase capacity for testing and manufacturing of protective equipment for healthcare workers struggling with the surge of COVID-19 patients.
I would like to leave you with a quote (translated by GoodQuotes from the French):
“But again and again there comes a time in history when the man who dares to say that two and two make four is punished with death. The schoolteacher is well aware of this. And the question is not one of knowing what punishment or reward attends the making of this calculation. The question is one of knowing whether two and two do make four.” –Albert Camus, La Peste
Hope all of you are safe and sound.
Follow Otello Stampacchia on Twitter: @OtelloVC
This article expresses the personal views and perspectives of the author. The views and perspectives expressed here do not necessarily represent the views or perspectives of Omega Fund Management, LLC or any officer, director, partner, member, manager or employee of Omega Fund Management, LLC or any of its affiliated entities.