Post #793: Death certificate data and unnecessary confusion out of the CDC

Few people are familiar with the mechanics of death certificates and death certificate data.  I happen to be one of them, having studied and written about end-of-life care in the Medicare program. I’m going to take a few minutes to try to clear up some absolutely unnecessary confusion regarding counts of COVID-19 deaths.

The CDC has been putting out statistics on COVID-19 deaths for months now.  For some reason, the most recent weekly release has caught the eye of the right-wing nuts, starting with Fox News and working its way down the food chain.  The statistic of interest is that only 6 percent of deaths with COVID-19 have COVID-19 as the only condition listed.  And that 94% of deaths with COVID-19 have at least one other condition listed.

The implication that all of them seem to be trying to make is that COVID-19 itself isn’t that dangerous.  It’s all that other stuff.  And that’s just plain wrong, and comes from ignorance about how death certificates are filled out and how the resulting information is processed.

Part of the problem is that the CDC writeup is geared toward professionals.  It assumes that you understand death certificate data.  It skips over the basics of death reporting, and goes straight into talking about the minutia.  Unfortunately, none of the people blathering about this seems to understand anything about death certificate data.  Which then allows all those people to start making ridiculous claims.  Which could have been quashed, by the CDC, from the start, simply by providing key basic information in the writeup.  Instead of burying it in the technical notes, which nobody reads.

The short answer here is that almost nothing about that 94% figure means what you naively think it means.*  So let me start with the craziest and most ill-informed assertions, and then work my way down to things that a reasonable person might think, but that just aren’t true.

* Hint:  If you die, you pretty much by definition have gone into cardiac and respiratory arrest.  The institution filling out the death certificate may, if it wishes, include those as conditions contributing to death.  And death certificates with those coded, in addition to COVID-19, will be counted among the 94% of cases that had “something else” in addition to COVID-19 on the death certificate.  But for the typical decedent, they aren’t meaningful diagnoses.  They’re just a description of what happened as they died.  This is so well-known that the Medicare Diagnosis-Related Group (DRG) definitions that determine Medicare hospital payment ignore these diagnoses when reported for decedents.  That is, cardiac and respiratory arrest normally count as major complications that increase the cost and payment for a case — except for people who died, in which case, Medicare just counts these as business-as-usual and does not offer extra payment based on the presence of such diagnoses on the claim.  It’s also fair to say that if you’re hospitalized for COVID-19, you’re pretty much guaranteed to have a diagnosis of respiratory distress And so, once again, if a death certificate has diagnoses of COVID-19 and respiratory distress, it’s not like the distress was somehow unrelated to the COVID-19 infection.  But it will kick that death certificate into the 94% that have some other condition coded, in addition to COVID-19.   


Incorrect conclusion #1:  Only 6 percent of  people who died with COVID-19 died from COVID-19.  This is one that the CDC could directly address, but they buried it in the technical notes.  To the contrary, of all the death certificates that mention COVID-19, 94% show COVID-19 as the underlying cause of death. 

Here’s what the CDC says, in the box below.  The universe of deaths here is all death certificates that make any mention of COVID-19. (And it’s just a coincidence that both the figure above and this figure are 94%.)

COVID-19 is listed as the underlying cause on the death certificate in 94% of deaths (see Table 1).

Source:  US CDC, technical notes for COVID-19 death data release.

So, to be clear, in the expert opinion of the physicians treating these patients, 94% of patients who died with COVID-19 died from COVID-19.  In the remaining 6% of cases, the physician judged that the underlying cause of death was something else, and COVID-19 just happened to be present at the time.

  People who will try to convince you otherwise are just dead wrong, and are directly contradicted by what the CDC wrote — in their technical appendix to that report.

To get from “death certificates with COVID-19” to “deaths from COVID-19”, you need to know just a little bit about death certificates.    Physicians sign death certificates (in almost all cases).  Typically, the death certificate will be signed by the physician most familiar with that patient’s case.  The physician signing the death certificate has to make an informed judgment about the underlying cause of death.  Everything you see, or have ever seen, about causes of death in the US is based on those physician judgments.

Really, all you need to do is glance a the US CDC instructions for filling in death certificates, and you will have a good idea of how additional diagnoses get onto the death certificate, in addition to the cause of death.  In addition to that underlying cause of death, death certificates contain fields for other conditions that were present at time of death, and most importantly, any conditions that lead up to the final condition that killed the patient.  In a sense, these are a) what lead to death and b) things that you might reasonably say contributed to the patient’s frail condition just prior to death.

Like so:  The first person died from a rupture of the heart muscle, but the underlying cause of death — the reason they were in that predicament in the first place — was clogged arteries of the heart (coronary artery disease).  The second person died from acute kidney failure, but the reason they ended up that was was diabetic coma as a complication of diabetes.  The underlying cause of death was diabetes.

And so, as you look at those examples, you also realize that if the physician could have attributed the underlying cause of death to some long-standing complication, he or she would have.  (Because that’s what they are supposed to do.)  By implication, the physician’s judgement was that those other conditions were present but did not lead up to the death.  I.e., that COVID-19 came in and killed that person.  And while other conditions were present, the death was not the result of some expected natural progression of some pre-existing condition.

One final note:  Why doesn’t the CDC just use the 94% of cases where COVID-19 is the underlying cause of death.  Why complicate things, at all, by bringing in that additional 6% of cases where COVID-19 is mentioned but not listed as underlying cause.  My guess, from reading the report, is that they think that a substantial fraction of cases with flu, pneumonia, and COVID-19 are actually people who died from the COVID-19, despite what the underlying cause data said.  So I think they did that as a matter of completeness.  Certainly there would be no material change to the overall picture if they dropped that marginal 6%.


Incorrect conclusion 2:  94% of COVID-19 decedents had some other pre-existing condition on the death certificate.  While it is technically correct to say that only 6% had just one diagnosis listed — COVID-19 — it’s that “pre-existing” phrase that I’m focused on here.  Because that’s not right either.  Much of the diagnosis information on death certificate consists of things that only occurred during the hospitalization for COVID-19.  Many of them are things that more-or-less always occur when people die, and it’s just up to hospital’s coders as to whether they get put in the death certificate or not.

Technically speaking, these conditions arising during the stay are “complications” and not “comorbidities”, something that death certificates distinguish poorly if at all.  That said, Medicare’s hospital discharge data now routinely separate those cleanly, because hospitals must flag whether or not a diagnosis was “present on admission”(POA)  in order to get paid under the traditional Medicare program.  Those POA flags could, I think, be used to separate out pre-existing comorbidities from complications occurring during the hospital stay.

Bearing in mind that most of the decedents were elderly, the five top identified diagnoses (other than COVID-19) were:

  1. Influenza and pneumonia.
  2. Respiratory failure
  3. High blood pressure
  4. Diabetes
  5. Respiratory distress

Items 1, 2, and 5 are almost certainly consequences of the COVID-19.  These really are complications.  It’s tough for me to imaging dying of COVID-19 without going into respiratory distress and then, eventually, respiratory arrest.  And “pneumonia” can simply mean inflammation of the air sacs of the lungs, which, if you’re dying from COVID-19, you probably have.  And if you don’t, an extended period on a respirator, as an elderly person,  is a pretty good way to guarantee that you have some sort of micro-organism-caused pneumonia.

Items 2 and 3 are just about the two most common chronic conditions in the elderly.  Sure, about one-fifth of the (mostly elderly) decedents in the analysis had high blood pressure listed as a factor contributing to their deaths.  But guess what?  About 60% of the elderly have high blood pressure, in life.  Diabetes?  I believe the current figure is that 23% of the elderly have diabetes.

This is something that I had seen in my own work on Medicare end-of-life care.  What you see as the secondary or contributing diagnoses on death certificates is almost an exact mirror for the prevalence of chronic conditions in the population.  Which, when you think about it, is just common sense.  The disease burden that they carry in life is present at time of death.

The upshot is that death certificate data aren’t even very helpful for determining the full extent of chronic conditions present at time of death. Conditions are only listed if the physician of record believed that they contributed to likelihood of death.  And so, while about 60% of the elderly have high blood pressure, only about 20% of these COVID-19 death certificates listed high blood pressure.  If you relied in death certificates, you’d think that high blood pressure protected you from COVID-19.  Which is not right.


If you actually want to know something about comorbidities and COVID-19 deaths you have to look at hospital discharge abstracts.  There, the goal really is to list everything that’s wrong with the patient.  And there, I am completely sure that the vast majority of COVID-19 decedents had other pre-existing conditions.  Why? Because most COVID-19 decedents are old, and easily 80% of the elderly have at least one chronic condition present.  If COVID-19 merely struck down a random sample of the elderly, 80% would have some other condition present.

The scholarly research does seem to say that, of hospitalized COVID-19 patients, certain pre-existing conditions make it more likely that they will die.  And, most of these conditions made sense (to me) as they indicated a generally more frail older person.

But once again, in this regard, COVID-19 is absolutely no different from any other cause of death.  For every type of hospital-based death in the elderly,  the higher the underlying disease burden, the more likely they are to die.  There’s an informal rule of three organ system failure:  By the time a Medicare beneficiary has three failing organ systems, likelihood of death increases greatly.

So, if you were to look at hospitalizations for anything in the elderly — heart disease, hip fracture, cancer — you name it, you will ALWAYS find that presence of significant comorbidities increases chance of in-hospital death.  Why anyone was surprised that this was also true for COVID-19 was, well, kind of a surprise to me.  Just Google something like Impact of Comorbidities on _____ Mortality, filling the blank with any cause of death.  Guarantee you’ll get some hits.  And in most cases, the estimated magnitudes are as large as you see posted for COVID-19 deaths.

The more stuff you have wrong with you, the more likely you are to die from fill-in-the-blank.  And that includes COVID-19.  It’s not really a surprise.  And you really can’t infer that from death certificate data.  You can plausibly infer that you are less likely to die from COVID-19 if you are otherwise healthy.  That much is true.  You cannot infer that “its the other stuff that actually killed them”, for patients who died from COVID-19.

Let me put it this way.  If you’re fat, you probably can’t run very fast.  That said, if you’re a fat person, and you got hit by a train, it’s the collision that killed you.  The fact that you couldn’t dodge very well is really a secondary issue.