Recently, people have been circulating a statistic from the CDC that says 94% of death certificates listing COVID-19 as a cause of death also list at least one other cause of death. For instance, if someone catches COVID, can’t breathe anymore and dies, perhaps the doctors would also list “Respiratory failure” as one of the causes of death, in addition to COVID. Come to think of it, why do only a third of COVID deaths include respiratory failure as a cause, how exactly is COVID killing people if not by causing respiratory failure? Before parading around this statistic, I have to ask, do we really understand what it’s even saying?
That misleading statistic came to my attention because a friend wrote a Vox article about it. He brings not a medical perspective, but a psychology perspective, discussing the cognitive biases that make people bad at understanding causation.
Causation is also a favorite topic of mine as well, although I come at it from a different set of perspectives: philosophy, physics, and law. And although I don’t have medical expertise, it’s not hard to find the medical standard of causation from google, so I include that at the end.
In philosophy, there is a distinction between the “background conditions” of an event, and the “causes”. For instance, if someone dies from COVID, that person’s birth was a necessary precondition to dying, but we would generally consider it a background condition rather than a cause. The problem of distinguishing causes from background conditions is known as “causal selection”. The question philosophers ask is, what is the basis for causal selection? According to the Stanford Encyclopedia of Philosophy,
The standard view on selection is that there is no objective basis for selection: selection is interest-driven, pragmatic, and unsystematic (Mill 1846, Lewis 1986a, Mackie 1974).
While most people would agree that a person’s birth is not a cause of that person’s death, many philosophers do not think there is any objective basis for this. When people select causes, they’re using some subjective standard. The question is, which subjective standards are most appropriate for which situations? But at this point we’ve moved away from philosophy.
In my many years as a physicist, I’ve encountered many people people who seem to believe that physics is the basis of causation. I always have to tell them that we do not in fact, talk about causation in physics. We just, you know, solve equations to determine the trajectories of objects. You can solve those equations forwards through time or backwards through time, it’s all the same. Atemporal methods are also very common.
Furthermore, to the extent that physics does talk about causation, it does absolutely nothing to address causal selection. There is literally no physics distinction between saying that a person’s death was caused by a disease, and saying that their death was caused by the formation of Earth. As far as physics is concerned, all things in one’s past light cone (i.e. the entire observable universe) are viable causes.
For more practical concepts of causation, one place I would suggest looking is the law (and I’m thinking of US law here). Legal standards of causation are necessary to making decisions in real world cases, and must also be intuitive enough that they can be explained to the typical juror.
One important concept is “cause-in-fact”, or “actual causation”. Generally, cause-in-fact is determined by the “but-for” test–that is, if a person would have survived but for COVID, then we can say that COVID was a cause-in-fact of their death. The but-for test is very intuitive, but some of the consequences are counterintuitive. For instance, if a person dies because COVID aggravated their heart condition, then both COVID and the heart condition pass the but-for test. On the other hand, if a person suffers from two conditions, both of which are deadly, then neither of them pass the but-for test. In the latter case, the law would consider both diseases to be causes-in-fact, even though they don’t pass the but-for test:
The courts have generally accepted the but for test notwithstanding these weaknesses, qualifying it by saying that causation is to be understood “as the man in the street” would, or by supplementing it with “common sense”.
As you can see, legal standards tend to gloss over philosophical minutiae, for better or for worse. It’s still a significant step up from how most people think about causation.
Readers are also encouraged to look up other legal standards of causation such as proximate causation or substantial factor causation.
When people see this statistic that says 94% of COVID death certificates also include other causes of death, they attempt to apply a but-for test, believing that most of these people would have died regardless of COVID. The but-for test seems reasonable to me, but they’re applying it incorrectly. I mean, let’s supplement some common sense here. If a person’s cause of death is listed as both COVID-19 and pneumonia, it seems pretty clear that COVID was a likely cause of pneumonia in the first place, as that is the most common mechanism by which COVID kills people. They would not have had pneumonia but for COVID.
Furthermore, if you were to really take the but-for test seriously, you would just look at overall death counts, and compare them to previous years. You can find that data by looking up “excess deaths”–here’s an article from a couple weeks ago. The but-for test, rather than reducing the number of deaths attributed to COVID-19, actually increases the number. I imagine that at least some of these deaths are caused by the additional burden placed on the medical system, rather than directly by COVID.
Finally, let’s look at the problem from a medical perspective. I did the bare minimum and googled the standards of causation used in death certificates.
Part I is for reporting a chain of events leading directly to death, with the immediate cause of death (the final disease, injury, or complication directly causing death) on Line a and the underlying cause of death (the disease or injury that initiated the chain of morbid events that led directly and inevitably to death) on the lowest used line. Part II is for reporting all other significant diseases, conditions, or injuries that contributed to death but which did not result in the underlying cause of death given in Part I.
If COVID is only listed in part II, it seems like maybe the death shouldn’t attribute the death to COVID. So how does the CDC count it? I couldn’t find a definitive answer, but learned that the WHO does not count such cases (see page 7), and I don’t see why the CDC would behave differently.
In other words, 100% of the deaths attributed to COVID-19 are cases where COVID-19 is part of the chain of events leading directly to death. It’s almost as if medical experts might know what they’re talking about.
There is an excellent video on youtube about this problematic: https://www.youtube.com/watch?v=_TECf3xSFbU
Apparently, death certificates that list Covid as the only cause of death are probably wrongly filled out.
I think the most objective statistic is the number of excess deaths. While it’s true that someone with enough comorbidities to be highly vulnerable to SARS-CoV-2 has died “because of” a number of different factors, it’s also true that all other things equal, the COVID-19 epidemic played a significant role in their death.
However, even excess deaths is tricky because it could be a secondary result of the pandemic, particularly in the situation we saw in New York in March. If ERs are swamped with coronavirus patients and you die of a stroke because you are delayed in getting needed care (note: I am not a medical doctor, and I don’t take offense if someone corrects my scenarios) then do you count this as a coronavirus death? I wouldn’t, but it is a death attributable both to the pandemic and to the public health response.
The practical takeaway is that if you’re vulnerable, you should be very careful not to contract the virus, and even if you are one of those likely to have mild or no symptoms, there is still a public health benefit to not spreading it.
It’s like other multi-factor disasters. It takes a combination of factors for a restaurant to burn down from a grease fire. One that had much better fire safety in general might be able to confine the fire to the kitchen, while others might burn to the ground.
Here in Oregon, the state sends out a daily update with a paragraph for each newly deceased Oregonian, among other pandemic information. I began reading every update in April, and it was July before I saw a paragraph that did NOT end with “…had underlying conditions.” Since then I’ve seen a few more, but very few. I went looking for Oregon’s underlying conditions rates, and couldn’t find a public dashboard that shows it. I wouldn’t be surprised if 2% or fewer of Oregon’s COVID dead were devoid of underlying conditions.