Post #1307: More mortality rate nonsense, Part 2.

Posted on October 28, 2021

 

I need to finish the analysis started in my last post.

At the behest of a reader, I’m trying to track down the evidence behind a claim in this Fox interviewBroadly, the claim is that COVID-19 vaccines are causing an increase in non-COVID deaths.  Plus-or-minus some strategic weasel-wording.

In my last post, I went through all the real research showing that’s not true.  There is no evidence of increased non-COVID mortality among people receiving vaccines, based on:

  • the randomized clinical trials of the vaccines.
  • a study of 11 million people in seven large health plans.
  • a study of the residents of several hundred nursing homes.

Separately, in an earlier post I noted that the rate of death among the vaccinated, as reported to the U.S. Vaccine Adverse Events Report System (VAERS), is a tiny fraction of what would be expected, based on the background mortality rate for the elderly alone.  That’s Post #1208, A funny thing about deaths in the elderly.

At this point, as you can probably guess, I’m a bit skeptical of the claim that COVID vaccines are causing non-COVID deaths.  In terms of any real research that actually looks at vaccinated individuals, there’s nothing to suggest that there’s an excess of non-COVID deaths.

In theory, one specific part of the Fox interview claim can be checked empirically.  Part of the claim is that U.S mortality statistics show excess non-COVID mortality in the U.S. age 20- (or maybe 30-)to-50 population in 2021.

Plus-or-minus their loose wording, that’s what I’m trying to track down here.  After I find that — if I can find it — then I can asses whether there’s any plausible to link it to COVID-19 vaccination.


Don’t lose sight of the big, obvious linkage between COVID-19 vaccination and non-COVID deaths.

I’m going to spend a lot of time going through the U.S. mortality data in excruciating detail. We are definitely on track to end up in the weeds.

But before I do that, I want to point out the elephant in the room.  So let me take one minute to state the obvious, starting from the graphs in my last post.

Here’s a graph of non-COVID deaths in 2020 and 2021, from this scholarly source.  They clearly show some significant ups and downs.  You can’t see it here, but that’s far more variation than you see in a normal year.

Here’s the same graph, with COVID deaths included.  So now, the top line on the graph is total U.S. deaths including COVID-19 deaths.

Source: Mortality Tracker:  the COVID-19 case for real time web APIs as epidemiology commons.

It’s completely clear that when COVID-19 deaths peak, there’s a small spillover to reported non-COVID deaths.  In other words, spikes in COVID deaths are driving some modest increases in non-COVID deaths. 

There’s nothing subtle about this.  Not only are the extra deaths clearly visible, but the timing tells you this is cause-and-effect.  Don’t have to appeal to any mysterious lags.  When COVID-19 deaths go up, non-COVID deaths rise to a lesser degree.

I went through three plausible mechanisms for this in my last post.  Might be simple under-coding of COVID-19, so the increases in non-COVID deaths are merely mis-classified COVID-19 deaths.  Might be that that COVID-19 is contributing to the impending deaths of otherwise frail individuals.  And it might be that the crush of COVID-19 cases is crowding other ill individuals out of hospitals, ERs, and other parts of the health care system.

In any case, the plain reading of the data is that, for whatever reason, the peaks in COVID-19 deaths are causing modest increases in reported non-COVID deaths.

OK, then:  Who is to blame for those additional non-COVID deaths, occurring underneath the COVID-19 death peaks?

Answer:  The un-vaccinated.

Why?  Because they account for the overwhelming majority of the COVID-19 deaths.  To a close approximation, the peaks in the COVID-19 deaths are peaks in the deaths of un-vaccinated individuals.

Here’s how the death count splits out in Virginia, as of the last available week:

Source:  Calculated from Virginia Department of Health COVID-19 dashboard.

And here’s how it splits out in the U.S. as a whole, for the last available month:

Source:  Calculated from CDC COVID data tracker.

No matter how you slice the pie, about 90% of all COVID-19 deaths now are occurring among the un-vaccinated.  (And that’s despite the fact that the un-vaccinated are a minority of U.S. residents, and that they are concentrated in younger age groups that have a lower risk of death.)

And so, if you want the straight story on this, connect the dots.

  • The un-vaccinated are largely to blame for COVID-19 deaths.
  • Those COVID-19 deaths appear to spill over into additional non-COVID deaths.

Therefore, to a first approximation, who is to blame for those additional non-COVID-19 deaths?  The deaths that you can see, right in front of your face, visible to the naked eye, on that chart above?  It’s the same people who are to blame for the COVID-19 deaths.  Blame un-vaccinated adults.

The obvious conclusion?  If you are genuinely worried about the potential for excess non-COVID-19 deaths, attributable to some actual cause, the first step to take is to get vaccinated and encourage others to do the same.


A more formal data analysis.

So, what was that Fox guy talking about?  Excess non-COVID-19 mortality in 2021?  And then, what’s the evidence (if any) that links that to vaccination?

First, define excess mortality.

Let me start by making sure were on the same page regarding the phrase “excess mortality.  In this context, it has a very precise meaning.

First, by “excess”, we mean “more than predicted”.   Literally, more deaths than were predicted by a sophisticated statistical model.

In theory, anybody could make a prediction.  It’s not hard.  Take average actual deaths, week-by-week, for the past decade.  Inflate for overall population growth.  And that’s a prediction.  It’s actually a fairly good prediction, as these things go.

In practice, the only prediction that matters is the prediction generated by (broadly speaking) the U.S. CDC.  They account for (e.g.) the aging of the population, the seasonality of deaths, population growth, and so on, to come up with their week-by-week prediction of total deaths.

Second, in practice, that’s often taken to mean “more than predicted, by a statistically significant amount”.  The whole point of “excess mortality” is to flag events that are unusual.  For example, to identify some year’s flu season as being a particularly deadly one.  CDC doesn’t want to flag normal year-to-year variation in mortality.  Instead, it only flags mortality as “excess” if there’s less than a 5 percent probability that such a mortality rate would be observed purely by chance, just from normal year-to-year variation.

In short, “excess mortality” has a very specific meaning.  It means that somebody (the CDC) used historical data, and a statistical model, to project how many deaths would would be expected to occur in a typical year.  And then, in practice, you only flag the mortality rate in some period as unusual if the actual, observed death rate exceeds the predicted value by some significant margin.

Because the CDC is more-or-less the arbiter of what constitutes “excess”, you really can only talk about excess mortality if the CDC identifies it for you.  In practice, it would be next-to-impossible to replicate the CDCs statistical approach to come up with a CDC-analog for (say) expected deaths by cause of death.


CDC mortality data

Near as I can tell, all of the 2021 CDC mortality data available to the public can be found from the links on the CDC web pageFor years prior to 2020, CDC published detailed (de-identified person-level) files of deaths by year.  For those, the detailed data would allow the user to tabulate the data as needed.

For current (2020, 2021) data, by contrast, CDC has produced a limited set of “provisional” data files, aggregated in various ways.  Those are the files available from links on that page.  If you want 2021 US mortality data, this is what you have to work with.  Any U.S. 2021 mortality data that you see published — including the highly-colored chart above — almost surely derives ultimately from these CDC files, or from tabulations directly provided by the U.S. CDC.


COVID-19 in a five-year perspective, from CDC data.

Source:  Calculated from CDC mortality files downloaded via this web link on  10/28/2021.

The first graph shows actual and expected (predicted) deaths.  This is, needless to say, the CDC’s prediction, not mine.  As you can see, the expected (predicted) deaths number just perks along smoothly, with modest peak each winter.  Actual deaths has a few more ups and downs, in the period prior to COVID.  And then COVID comes along in 2020.

If you ever hear people claim that COVID is really no worse than the flu, maybe show them this chart.  On the left, I’ve circled a flu season that was “the worst season we’ve had in the last decade, according to no less an authority than Dr. Anthony Fauci.  Keeping in mind that total deaths is the area under the curve, the worst U.S. flu season in the last decade isn’t even in the same ballpark as the 2020-2021 winter COVID-19 season.


Finding 2021 non-COVID excess mortality, and putting that into perspective.

The claim of interest isn’t about COVID-19 mortality, or even the flu.  It’s about an excess of non-COVID-19 deaths in 2021, following the mass COVID-19 vaccination of the U.S. public.  So now let’s put a finger on that.

Here’s the past two years, showing total deaths, predicted deaths, and COVID-19 deaths.  And then, if I merely subtract out the COVID-19 deaths, we’re looking at a comparison between non-COVID mortality and predicted mortality.  I’m going to take the difference between those two lines as “excess non-COVID mortality”.

Here’s the data as provided by CDC:

Source:  Calculated from CDC mortality files downloaded via this web link on  10/28/2021.

And here it is after subtracting the bottom line (COVID-19 deaths) from the top line (all deaths).  So the graph below is all non-COVID deaths, compared to predicted deaths.  And the difference between those lines would be unexpected non-COVID-19 deaths.

Source:  Calculated from CDC mortality files downloaded via this web link on  10/28/2021.

And there it is, circled above.  That gap, between those lines, is 2021 excess non-COVID deaths.   That’s the thing we’ve been looking for.  That’s what the Fox interview was about.  The claim is, that was caused by COVID-19 vaccination.

Source:  Calculated from CDC mortality files downloaded via this web link on  10/28/2021.

But in terms of attributing this to vaccines, we hit an immediate and obvious snag.  There were plenty of gaps between those two lines, as larger or larger than the last one, occurring well before vaccines were even available.  In fact, in the past two years, for the majority of the time, there have been big gaps between the curves for predicted deaths and total non-COVID deaths.

Finding a gap between actual and predicted deaths isn’t unusual.  Ever since COVID-19 hit, that’s been the norm.  There’s been a gap that size more often than not.

Source:  Calculated from CDC mortality files downloaded via this web link on  10/28/2021.

In fact, if you now pull back to the five-year perspective above, you see that differences between those two curves occur all the time.  Not just prior to COVID-19 vaccines, but prior to COVID-19.  Differences that are every bit as large as the late-2021 excess non-COVID-19 deaths.

To be clear:  If that late-2021 event had been something new and unique, then there would be reason to look further.  If a gap appeared between those lines, where none had ever been before, then that would surely be something worth looking at.  But the reality is that gaps appear between those two lines all the time.  It’s just business as usual.  To take that last gap and say, well, that must be due to vaccines, that’s clearly nonsense.  That’s no evidence at all that the last (of very many) differences between actual and predicted U.S. deaths is somehow linked to vaccines.

At some point, I may want to return to this and drill down into that late 2021 excess mortality.  Possibly there is some useful lesson there.  I’ve heard a lot of chatter are about drug deaths being up and such, it might bear looking into.

But given how frequently actual and predicted mortality diverge, it doesn’t seem like there’s any pressing need to look.  Mainly, nobody has any clue what’s driving that excess mortality in that particular instance.  Because it’s just one of a long string of such gaps that have been occurring for at least the past half-decade.  Given that, taking time to lay out the details of it seems like wasted effort.

When you pull back and look at it, this is just a particular variation of the semi-attached figure.  That’s a standard trick explained in How to Lie With StatisticsYep, there is excess non-COVID mortality in late 2021.  Yes, there were vaccines given earlier in that year.  But nope, there’s no reason to think that one of them has anything to do with the other.  Episodes of such “excess mortality” occur all the time, vaccine or no vaccine, COVID or no COVID.

In fact, this one is a subset of the semi-attached figure that I refer to as a “stopped-clock methodology”.  So let me go on to explain that, as a sort of extras for experts.


Extras for Experts:  The uses of the stopped-clock methodology

“When your client is the target, your job is to fuzzy up the bullseye”.  I got that advice from a co-worker, back when I worked as a consultant in the area of Federal health care policy.

In that sort of environment — where people will purposefully offer misleading information to achieve their goals — if you want to separate fact from fiction, you have to get good at spotting the fuzzy.  By that I mean all the pseudo-science, “How to Lie With Statistics“, truthy-sounding stuff that turns out to be nonsense if you look at it in detail.

It’s not as easy as you think.  There’s big money in the artful lie.  In any given situation, it was a pretty good guess that the guys paid to fuzzy up the picture were making a lot more dough than I was.

But after a while, I came to realize that the Industrial-Fuzzy Complex uses a more-or-less standard set of tools.  Foremost of which is the stopped-clock methodology.

Define a “stopped clock methodology” as a statistical analysis that always gives you the same answer, no matter what you apply it to.  If the question is something like “is there a problem here?”, it’s a method that is guaranteed to give you the answer “yes, there is”.  All the time.  (Or, maybe, “no there isn’t”, all the time).  Either way, it’s a sciencey-looking method that always returns the same answer, no matter what the particular situation may be.

Useless, you say?  Then you’ve never worked with lobbyists.  Because if it’s your job to defend the indefensible, a stopped clock methodology is your friend.  It’s your sure-fire go-to defense.  It doesn’t have to be right, it doesn’t have to be good.  It just has to have enough face validity to provide cover for your duly-purchased elected members of the Congress to vote your way.

So let me begin this analysis by praising the logic behind this latest nonsense about the COVID-19 vaccines.   

Put yourself in the shoes of some foreign power, say, who wanted to discourage Americans from getting vaccinated, and so prolong the COVID-19 drag on American society and the American economy.  How would you scare them away from getting vaccinated?

Well, you can’t talk about vaccines and COVID-19 deaths.  At this point, unless you’ve been living in a cave somewhere, you realize that the cat’s out of the bag on that one.  Everybody who looks at the data concludes that the vaccines are quite effective at preventing COVID-19 deaths.

Here, for example, are the last four weeks of data from Virginia, showing how much more likely you are to die from COVID-19 if you are un-vaccinated.

Source:  Virginia Department of Health COVID-19 dashboard accessed 10/27/2021.

That’s just the raw observational data.  It doesn’t even account for the much younger age (and far lower risk of death) of the average un-vaccinated individual remaining in Virginia.  The true apples-to-apples impact of vaccines on COVID-19 deaths is almost certainly larger than what is shown above.  But this is good enough to get the point across with no statistical hocus-pocus.

Not to worry.  You’ll just have to focus on non-COVID deaths instead.

But you can’t directly compare non-COVID deaths among the vaccinated and un-vaccinated populations, either.  As I showed in yesterday’s post, if you do the obvious thing — track individuals who were and were not vaccinated, and compare the death rates — you end up with nothing.  No excess mortality in the randomized controlled trials, in a study of hundreds of nursing homes, or in a study of 11 million persons in seven large health care plans.

So your last chance of scaring people, on the risk-of-death front, is to fuzzy things up as much as possible.  The only thing left for you to do is to talk about  non-COVID deaths, and avoid directly comparing the the vaccinated and un-vaccinated populations.

But that’s OK, because if there’s enough stuff to look at, you can always find something to point to.  Surely, somewhere, for some population, for some cause of death, for some time period, some mortality rate must have gone up after COVID-19 vaccination started.  All you have to do is point to that and call it your smoking gun.  It doesn’t matter if no serious analyst would accept that as evidence because, let’s face it, the people who you’re aiming this at aren’t exactly the sharpest knives in the drawer.  All you need is something that’s adequate for your target population.  Pretty much any claim will do when your audience is sufficiently credulous.

And so we arrive at the current bit of Fox-pushed nonsense.  This nonsense being that, somehow, indirectly, in some countries, for some age groups, for some time periods — but not for others, and with no direct link to the level of COVID-19 vaccinations — the COVID-19 vaccine has caused large increases in non-COVID deaths that are visible as excess non-COVID mortality in national mortality statistics.

Which I now hope you can recognize as a classic stopped-clock methodology.  Give me enough cells to look at — cause of death x age x time period x county — and I’m fairly sure I can find one that went up after vaccines were introduced.  Point to that one, ignore all the others, and you’re good to go.  There’s your answer.

So long as some estimate, somewhere, says that some population had excess non-COVID mortality in 2021, that’s all you need.  Find the clock that’s stopped at the right place, then point to it, wave your hands, and tell people that it means what you say it means.