Mortality with Meep: Were COVID Deaths Grossly Overcounted?
No. They weren't.
The last post depressed me so much, I’d rather write about something happier: COVID mortality.
Latest COVID mortality myth: that COVID deaths were something else
The “Only 6% of COVID deaths were really COVID deaths!” myth has been going around, based on the fact that the vast majority of COVID-related death certificates had multiple causes listed on them, not just COVID.
Example from Politifact: Trump repeats false claim about COVID-19 deaths on Fox News
Before I get into all the details of how death certificates list causes of death and the average number of causes listed, I have a mini-rant.
Be aware of what you don’t know and learn from mistakes
The major thing that has been driving me nuts since February are people taking on the pose of expertise or certain knowledge when they don’t actually have that knowledge.
I’m not going to get Rumsfeldian on everybody, because the following isn’t pithy. Here are the pertinent categories of people:
People who know they don’t know, and more importantly know that nobody knows, but exude authority and knowledge because they think laypeople can’t handle the fact that nobody knows. [I despise these people, but I think they’re fewer than you would think]
People who know they don’t know, and more importantly know that nobody knows, but are dealing with people who really don’t deal well with uncertainty. They try their best to communicate the uncertainties but are inevitably misunderstood, whether due to ignorance or deliberate choice. [I don’t mind these folks…. mainly because this is often the category I’m in.]
People who really don’t know, but assume that due to their facility to numbers, they understand what’s going on with the data, and don’t realize all the assumptions they’re making in interpretation. [This group is very active on twitter]
I can keep going on in this vein, but I won’t. Most of the ignorant people driving me nuts, I don’t even know and interact with only once, if that. I have enough pain in my life.
But it is really sticking in my craw how many people seem blissfully unaware of their own level of ignorance, and, more specifically, all the things they’re assuming about data, epidemic dynamics, and more.
My main point is that so many people are not learning from their errors, pinballing from one reaction to another, and never revisiting their earlier, erroneous analyses and predictions. It doesn’t matter much, on a global scale, if it’s just random people on twitter, yelling into the void. It does matter when it’s policy-makers who never course-correct because they’d rather not admit error or take responsibility for it.
Okay, that’s enough of that. Time to get into death data!
Yes, really, most official COVID deaths are COVID deaths
And by “most”, I mean over 90%.
No, the CDC did not ‘quietly adjust’ US coronavirus deaths
As of Aug. 31, the CDC reported that 182,622 Americans have died since the start of the pandemic — and some estimates put the death toll even higher. The agency told us that the vast majority of deaths involving COVID-19 can be attributed to the virus.
First, this is poorly written. Much more than 182k Americans have died since March. Given that almost 3 million Americans die in a “normal” year, one expects about 1.5 million people dying over 6 months (March-August).
They mean 182k “excess” deaths.
The social media posts trace back to an Aug. 26 update from the National Center for Health Statistics (NCHS), which is part of the CDC. The report is updated every Wednesday and outlines provisional death counts for COVID-19.
In his post, King cited an older version of a data table about deaths from COVID-19 in combination with other conditions, such as pneumonia and influenza.
In a section titled “Comorbidities,” the NCHS wrote: “For 6% of the deaths, COVID-19 was the only cause mentioned. For deaths with conditions or causes in addition to COVID-19, on average, there were 2.6 additional conditions or causes per death.”
This is the point at which the ignorant screeched OMG ONLY 6% OF COVID DEATHS ARE COVID DEATHS.
Do you know how many death certificates normally have only one condition/cause of death on them?
Do you know the average number of causes of death listed on a death certificate in the U.S.?
Let’s find out.
Standards for listing causes of death
Principles and Pitfalls: a Guide to Death Certification
ABSTRACT
Death certificates serve the critical functions of providing documentation for legal/administrative purposes and vital statistics for epidemiologic/health policy purposes. In order to satisfy these functions, it is important that death certificates be filled out completely, accurately, and promptly. The high error rate in death certification has been documented in multiple prior studies, as has the effectiveness of educational training interventions at mitigating errors. The following guide to death certification is intended to illustrate some basic principles and common pitfalls in electronic death registration with the goal of improving death certification accuracy.
……
For mortality research purposes, it is important that the cause-of-death section be reported as specifically as possible. The cause-of-death section is divided into two parts. Part I is used to report the causal chain of events leading to death. The most recent or “immediate” condition that led to death is placed on Line a; other conditions (if any) that gave rise to the immediate cause of death are listed sequentially on Lines b–d.1,2 The last and most remote condition in the chain of events leading to death is known as the “underlying” cause of death, and should be etiologically specific.
……
The bottom lineThe condition listed on the bottom line of Part I (ie, the underlying cause of death) is arguably the most important in that this is generally what will be coded as the cause of death. Mortality data worldwide are coded according to the current International Statistical Classification of Disease and Related Health Problems (ICD-10) system that is published by the World Health Organization (WHO).16 The system facilitates interpretation and comparison of mortality data by translating the cause of death into an alphanumeric code that corresponds to a particular disease or injury. *From a public health perspective, the most effective strategy is to prevent the initiating disease or injury that precipitated the chain of events leading to death*.16 For this reason, it is important to carefully consider underlying causes.
So first, those filling out the death certificate are exhorted to fill out the whole chain of events leading to death. The piece I linked goes through multiple examples, and as I’m not a medical person, I don’t know common cause-and-effect chains. But, in general, death occurs because your heart stopped, permanently. But what caused the heart to stop, can have many different pathways.
Think about what it means to die of diabetes, Alzheimer’s, or even cancer. How do they actually kill you? What’s the path? Should Stu’s chemo kill him before cancer spreads, I wouldn’t be surprised if the Underlying Cause of Death is considered cancer (after all, he wouldn’t be on the heart-destroying chemo if it weren’t for cancer.)
WONDER at CDC, multiple causes of death: Multiple Cause of Death 1999 – 2018
About cause of death classification:
The underlying cause-of-death is defined by the World Health Organization (WHO) as “the disease or injury which initiated the train of events leading directly to death, or the circumstances of the accident or violence which produced the fatal injury.” Underlying cause-of-death is selected from the conditions entered by the physician on the cause of death section of the death certificate. When more than one cause or condition is entered by the physician, the underlying cause is determined by the sequence of conditions on the certificate, provisions of the ICD, and associated selection rules and modifications.
……
About Multiple Cause of Death:
Each death certificate contains up to twenty multiple causes of death.
….
When data are grouped by Multiple Cause of Death, the numbers of deaths are shown attributed to each cause. However, a single person’s death may have up to 20 causes on the death certificate. Thus the total number of deaths per cause may exceed the actual number of dead persons in the selected population.
So, the minimum number of causes of death on the certificate is 1, the maximum in the current system is 20.
Let us see if we can figure out the average number of causes of death on a single death certificate.
Method for getting average number of causes of death on the certificate
When I go to WONDER at the CDC, I can query their death database in various ways, and, sorry, some of the queries take too long to process, so the only one I can get well in a reasonable amount of time is the average number of causes of death on a certificate in any given year.
Here is how I do it:
First, I break out the total deaths per year. I get the death totals from the Underlying Cause of Death database. Remember you get only one underlying cause per death. It runs from 2009-2018. and I’m doing all the years, just because. I’m going to graph this, because I find it interesting: the total number of deaths & the crude death rate (# of deaths per 100,000 population).
Unsurprisingly, the number of deaths increases each year. Don’t get fooled by my choice of vertical scales for the two series; I chose them so you could see the patterns. The increase in the number of deaths per year has been about 0.9% for the 19-year period.
The raw death rate trend is interesting, and I’d like to graph it for a longer-than-19-year period, but that’s for a different time. It’s not just a case of changing mortality rates and changing population size, but also the age composition of that population. The oldest Boomers hit age 63 (right after Social Security eligibility) in 2009.
The next step is to use the multiple causes of death database, which will do stuff like tell you how many death certificates had “T40.5 (Cocaine)” on them as one of the many causes of death, but not necessarily the underlying cause of death. I believe this is counting all of them.
Then I divide the total number of deaths by causes (144,075,392 for 1999-2018) and divide by total number of deaths (50,567,774) to get the average number of causes of death per death certificate: 2.8 for the period 1999-2018.
Rising trend in number of causes of death on certificates
Here is the same, done year-by-year:
So you can see that the average number of causes of death is rising. In recent years, the average is over 3.
This increasing trend could be driven by changing standards for death certificates or it could be driven by more and more deaths being that of the elderly, who have multiple co-morbidities. It can be shifts in causes of death — some causes are more likely to have multiple things going on (such as an elderly person with diabetes, heart disease, and Alzheimer’s getting the flu and dying of pneumonia), versus dying in a car accident (which may have drug/alcohol use also listed, but probably has fewer conditions listed on the death certificate).
[That’s another thing driving me nuts about the COVID death data: you are given how many had co-morbidities (diabetes, heart disease, obesity, high blood pressure, Alzheimer’s, etc.), but the vast majority of deaths are of elderly people. How many of those don’t have any co-morbidities? ]
I may revisit this “multiple causes of death” issue another time, digging into flu/pneumonia deaths and how many causes (and what are the most common ones) we see. The issue I have is that the database calls are very slow, even for some of the simplest queries, and I don’t have time to waste on such a slow system. [I would looooove to just download the whole data set… wink wink.]