Slammin’ the JAMA -- studied mis-use of COVID data - Analysis
by D. V. Williamson
Reposted with permission
I am going to do some heavy editorializing. The reader, of course, does not have to agree with any of it.
Target-rich environment. That is what the Journal of the American Medical Association (JAMA) over-generously provides. I wonder that I could make a career of dismantling almost any randomly selected piece from the journal, for it has long become so invested in politically-correct “Science.”
I subscribe (for free) to the JAMA Online First distribution. The JAMA can make for depressing reading. I often pass it over. But here is a quick hit.
I get the sense that, over the last month or so, the self-anointed keepers of The Eternal COVID Flame have started to give total, all-cause excess mortality more attention. They used to concentrate on COVID case counts or on mortality attributed to COVID, but that just amounts to bean counting. Case counts and COVID fatalities ignore important aspects of the COVID phenomenon. For example, they ignore the fact that COVID fatalities do not map one-to-one into excess mortality. We all die from something, and most folks who have succumbed to COVID were likely to die of something else not too far into the future. It is hard not to conclude that COVID has contributed to mortality that would not have otherwise occurred—it did induce a non-trivial volume of “excess mortality”—but it did not contribute one-to-one. Moreover, there is plenty of evidence that excess mortality remains elevated in the United States and elsewhere, and the authorities have not attributed most of the persistent excess to COVID. There is something insidious going on. The COVID phenomenon is way more complex than COVID itself. It has been hard not to conclude that the public policy response to COVID has induced much mortality itself, especially on younger people.
So, the JAMA posted a piece on May 20 titled “Excess Mortality in Massachusetts During the Delta and Omicron Waves of COVID-19,” authored by John Hsu, MD, MBA, MSCE; Vicki Fung, PhD; Joseph P. Newhouse, PhD. For nearly two years now, I myself have been tracking excess mortality when and where publicly-available data have allowed that. I was curious to know whether or not the JAMA analysis would parallel my own. The CDC itself has recognized that there is no unique way of measuring excess mortality, but one would hope that all methods would yield similar results. But, do they? And why wouldn’t they?
The JAMA piece irritates me, because it passive-aggressively advances much editorializing of its own. It’s what JAMA does. Specifically, the very short JAMA piece opens:
The COVID-19 pandemic has produced excess deaths, the number of all-cause fatalities exceeding the expected number in any period. Given reports that the Omicron variant may confer less risk than prior variants, we compared excess mortality in Massachusetts, a highly vaccinated state, during the Delta and initial Omicron periods.
What, pray tell, is wrong with that, one might ask? I would suggest that the first sentence slyly invites the reader to conclude that COVID, and COVID alone, has driven excess mortality. The second sentence is just badly written. There is the oblique suggestion that the hazards associated with the Delta variant exceeded those of the Omicron variant; does excess mortality reveal anything about the differential hazards? There also seems to be some suggestion that vaccination rates shouldn’t frustrate the comparison, because Massachusetts had already been “a highly vaccinated state”. Or something. Finally, there is the sly suggestion that the ersatz “vaccines” actually work. They work by preventing infection. (?) They prevent COVID fatalities. (?) Again, something.
I would suggest that the opening should have read something to the effect of:
There are questions about how much the various strains of COVID-19 have contributed to excess mortality over the last two years. Among other things, there is some understanding that the Omicron variant has broadly imposed less risk than the Delta variant; it may have been more infectious, but it was also less fatal. Such is a progression generally observed with viruses as they evolve over time.
We concentrate on excess mortality starting in the middle of 2021 when the Delta variant was dominant and then examine late 2021 and early 2022 when the Omicron variant was becoming dominant. We do not control for changes in vaccination rates in the population. We nonetheless observe that the Omicron variant induced greater excess mortality than Delta. It is thus not obvious, absent further analysis, that Omicron did in fact pose less risk. Further analysis would fold in the effects of vaccination in that the vaccines may have been less effective in the face of Omicron or they may have even made some subjects more susceptible to infection from Omicron.
My wordy version sets up lots of questions and relegates the analysis, such as it is, to what it is: an observation. But I would suggest that that the observation—the idea that greater excess mortality attended the diffusion of the Omicron variant—yet remains problematic. My principal reason comes down to the question of how the authors measured excess mortality. Their scheme either obscures or ignores any evidence of non-COVID excess mortality when, I would suggest, there is abundant evidence across the entire United States of a great volume of excess fatalities that the authorities themselves have not attributed to COVID.
I will make my case with a series of graphs. I will start with the graph from the JAMA piece. Each quadrant exhibits excess mortality for different age cohorts for the interval 2020 through March 2022.
The graphs demonstrate virtually zero excess mortality for younger people aged 18-49 (the lower right graph). Most excess mortality is driven by fatalities among people aged 65 or older (the upper right graph). The graphical results seem consistent with the view that all excess mortality was driven by COVID fatalities. So, there was no excess mortality driven by causes other than COVID. I would strongly argue that there have been other causes, especially among younger people. Consider the following graph. I’ve featured it in earlier essays1:
This graph features total mortality, week to week, from the beginning of 2015 through the end of 2021. The smooth, wavey line is my benchmark for expected mortality. I generate it by mapping a sine wave and a linear time trend against mortality from 2015 through 2019. I then project that smooth, wavey line into 2020 and 2021. The difference between the benchmark and actual mortality indicates “excess mortality”. I feature excess mortality on the lower part of the graph.
The graph identifies four, maybe five waves of highly elevated excess mortality, and those waves were largely driven by COVID fatalities. But not entirely (according to the CDC data). Moreover, one might discern an elevated, baseline volume of excess mortality even after each wave recedes. What is going on, and why do we not see the same elevated, baseline of excess mortality in the JAMA graphs?
I go back to the CDC data and extract data specific to Massachusetts:
In this graph one can see that Massachusetts did appear to experience at least two episodes of negative excess mortality after each of two big COVID waves. But then, starting in the middle of 2021, baseline excess mortality seems to become elevated. Let’s zoom in on 2021:
In this graph I map benchmark and total mortality on the left axis, and I map excess mortality (the gray area) on the right axis. Note that excess mortality seems to start bouncing around 600-per-week starting in July 2021. In contrast, the JAMA graphs seem to suggest that excess mortality through most of late 2021 was zero.
What would account for the discrepancy between my computations (using CDC data) and the computations the authors applied in the JAMA piece? An answer might be buried in this cryptic passage: “We corrected expected deaths for the decreased population owing to cumulative pandemic-associated excess deaths, …” I have not yet sorted out the mechanics of the “correction,” but the idea seems to be that the age distribution of the COVD death toll may affect the age distribution of the population going forward. And, insofar as the COVID toll has been concentrated on the very elderly, then, yes, the age distribution of the population will start to skew younger … and healthier. Shouldn’t we make some allowance for that?
One can imagine reasons for making allowances for important changes to the age distribution of the population, but most analysis of excess mortality out there in the ether do what I do: compute benchmarks based on mortality data from 2015-2019, and project those benchmarks into 2021 and 2022. The assumption is that the populations in 2021 and 2022 would have looked a lot like the populations in 2015-2019. My concern is that “correcting” the benchmarks in 2021 and 2022 amounts to ratcheting the benchmarks upward to reflect a “new normal”—and thus ratcheting apparent excess mortality downward—when, in fact, the “new normal” reflects part of the deviation from actually normal, baseline rates. Ratcheting down excess mortality in 2021 and 2002 would diminish or even obscure evidence of persistent, elevated, non-COVID excess mortality. Worse, it could hide the fact that much of that excess mortality was concentrated on younger people. To see that, consider this graph:
The gray area indicates excess mortality for people aged 0 - 24 in the United States. The yellow line indicates fatalities “with COVID” or “by COVID”. The blue line indicates excess mortality after stripping out those fatalities “with COVID” or “by COVID”. The grey area north of the yellow indicate the volume of non-COVID excess mortality. Note that there is a lot of non-COVID excess mortality concentrated on people aged 0 – 24. What is that about? I could go on. There has also been much non-COVID excess mortality concentrated on people aged 25 - 44.
I’ve explained before that the CDC and NIH and any other health authorities have a lot ‘splaining to do, because they are the people sitting on top of data of sufficient granularity to figure out what is going on. But, they are not going to do that analysis, are they?
Meanwhile, what is this JAMA piece up to? The last sentence provides a clue:
[T]he present findings indicate that a highly contagious (although relatively milder) SARS-CoV-2 variant can quickly confer substantial excess mortality, even in a highly vaccinated and increasingly immune population.
We are thus to conclude … what? The vaccines confer immunity, but immunity does not actually make one immune. Moreover, the danger isn’t over. Omicron was bad. It is still bad. Danger!
This stuff is ridiculous.
