Fact Check: Covid as a Leading Cause of Death in Children
CDC promoted a flawed pre-print from UK researchers
In the CDC ACIP meeting on June 17 to discuss childhood Covid vaccines, a table was presented showing Covid was a leading cause of death in US children as part of a slide deck on the epidemiology of Covid-19 in children and adolescents by Dr. Katherine Fleming-Dutra. The source was a pre-print written by a group of academics from the UK, including Dr. Seth Flaxman and Dr. Deepti Gurdasani, who is well-known on Twitter for her strong views on Covid. I later learned that a very similar slide was also presented at the beginning of the FDA VRBPAC meeting earlier in the week, as were other slides citing this "top 5 cause of death" claim.
The slide was shared on Twitter by Dr. Katelyn Jetelina ("Your Local Epidemiologist"), and retweeted by many influential people including Jerome Adams, Julia Raifman (tweet now deleted), Gregg Gonsalves, and Leana Wen. Only problem? It's completely and utterly false. The pre-print it's based on includes significant errors that invalidate the results. And the slide makes additional errors on top of the pre-print. It's really disturbing that data this poor made its way into the meetings to discuss childhood Covid, and that it took me less that a few minutes to find a major flaw (and then I found many more as I looked deeper).
On Friday, June 17, I contacted the study's corresponding author, Dr. Flaxman, who originally said he'd get back to me on Monday, but responded early Sunday morning to get more information about the source of the Underlying Cause of Death data I used for Covid (the CDC WONDER database, Provisional Mortality Statistics, 2018-present). He later posted on Twitter to say than an updated pre-print would be available soon.
UPDATE: A revised pre-print taking into account my critique was posted on June 28. The rankings are now for every age group, but the authors did some other questionable things to exaggerate Covid deaths in children, including changing the time period for the Covid deaths to the worst 12 month period in the pandemic. I posted a quick Twitter thread to review of the revised pre-print.
There are two major errors with this data, along with some other problems, so I'll go through them individually. The two major errors are of such consequence that I believe this study should be fully retracted.
Underlying Causes of Death
The first major issue I noticed was that they downloaded Covid numbers from NCHS. This source includes deaths where Covid is listed anywhere on the death certificate, not just the underlying cause of death. Death certificates include a single underlying cause of death, along with potentially several contributing causes of death. The NCHS data includes both, which overcounts Covid deaths because it includes death that had a different underlying cause.
The pre-print states "we only consider Covid-19 as an underlying (and not contributing) cause of death" but this is false. It also states "In children and young people (CYP) aged 0-19 years, data from the National Center for Health Statistics (NCHS) indicate there were 1,433 deaths for which Covid-19 was the underlying cause of death (March 1, 2020-April 30, 2022)." However, this includes deaths where Covid was NOT, in fact, the underlying cause of death. According to CDC WONDER, underlying deaths from Covid for this age group and time period actually totaled 1,088 deaths.
This is a major problem, because the pre-print adds Covid to a ranking of the top underlying causes of death from 2019 from CDC WONDER. The WONDER data they used is specifically Underlying Cause of Death data. This significantly overcounts Covid deaths compared to the other causes of death listed.
This is basically the same error that I addressed recently with Jeremy Faust's blog post comparing pediatric deaths from Covid and the flu. After I called out this error, Dr. Faust posted a follow-up blog post where he admitted it was wrong to compare multiple CoD data with underlying CoD data.
Cumulative vs Annualized Covid Deaths
The second major issue with the pre-print are the time periods for the deaths. The underlying cause of death data is for a single year - 2019 (more on that later). However, the rankings of Covid deaths by age group in the pre-print include both cumulative (over 26 months) AND annualized deaths for some strange reason. That means Covid is inexplicably ranked twice for each age group.
Below is the table ranking leading causes of death for 15-19 year olds. Notice that Covid is listed both as the 4th AND the 6th leading cause of death. This is non-sensical and extremely misleading. It is completely inappropriate to compare the cumulative number of Covid deaths over 26 months to deaths from other causes over a one year period. The only way to make a fair comparison is to use an annualized number. There's no good reason the cumulative number of Covid deaths over 26 months should be included on this list at all.
To make matters worse, CDC used the rankings for the cumulative Covid-19 deaths on their slide, further overcounting Covid deaths compared to the other causes of death listed.
I recreated the results from the pre-print using CDC WONDER to obtain deaths where Covid-19 is listed as the underlying cause of death in the time period listed, and annualized the results. The revised rankings are shown below, alongside the incorrect rankings used on the CDC slide. I also show the annualized number of deaths when using deaths with Covid as the underlying cause of death only, instead of the underlying or contributing cause of death.
Issues using Rankings for Pediatric Covid Deaths
Even given the corrected rankings above, there are also issues with the entire concept of showing the impact of Covid deaths in children using rankings. Rankings overstate the impact of Covid, because the top few causes of death far outweigh the causes further down the list. For example, in ages 1-4, accidents account for almost 25 times as many deaths as Covid-19 on an annualized basis. Furthermore, for each of the 4 age groups covered by the CDC slide, the very broad "accidents" is the leading cause of death. If we break that down further, causes of death like drownings, vehicle crashes, drug overdoses, would be individual causes of death greater than Covid in various age groups. Actuary Mary Pat Campbell explains this well in a couple of blog posts on pediatric Covid deaths:
The study authors incorrectly posit that the way the cause of death rankings are grouped, the ranking for Covid is "conservative" since it is a single cause rather than a group of causes (like accidents). However, this makes no sense when you consider that some of the subgroups under accidental deaths would alone be greater than Covid.
I have some additional thoughts on this study that I would like to delve into more at a later time. For example, why did they use data from 2019 and not 2020 or 2021, when aspects of our Covid response affected some of the leading causes of death? Also, why did a group of UK researchers analyze US deaths, for which they apparently did not understand the data well, instead of using deaths from their own country? Could it be because US counts Covid deaths very generously, so our data made it easier to present Covid as a leading cause of death in children? In addition, why did they inflate the counts by include 18 and 19 year olds in the data, when the pediatric population is generally accepted to be 0-17?
On the CDC side, how did Dr. Katherine E. Fleming-Dutra, MD at the CDC, a pediatrician and doctor of emergency medicine, not realize this data was seriously flawed and out of line with all other data about the impact of Covid on pediatric mortality? How did a pre-print get used in an ACIP and FDA presentation with such little oversight without the quality of the data being fully vetted? How did I uncover these issues, instead of them being identified by someone whose job it is to evaluate this kind of data? I am a mother who reviews this data on my personal time, and yet I seem to be much more knowledgeable about Covid deaths in children than most academics and public health officials working with Covid-19. We are forced to believe that the CDC researchers who put this data together are either incompetent or liars, and when all the mistakes go in the same direction, it certainly seems like the CDC uses whatever data they can find to push their agenda without any consideration to its veracity.
This is a national disgrace (and an international one, considering the paper came from the UK). I'm appalled at the low quality of the data presented and the fact that it was amplified in such an influential way when discussing the approval of vaccines in young children. The CDC and much of the academic and public health community have utterly failed the American public when it comes to accurately assessing the risk to children (among other things).