Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Covid Deaths Are Real: Rebutting Dr. Briand
I write to rebut the claims of Dr. Genevieve Briand, a senior lecturer at Johns Hopkins who holds a PhD in Economics and recently released a study questioning the Covid death statistics. The paper was subsequently withdrawn by Johns Hopkins, quite properly in my view. Dr. Briand’s analysis is deeply flawed.
Misleading and erroneous analyses like these have serious effects. It led our friend iWe to author a post yesterday titled Covid – Just A Big Hoax?, which cited Dr. Briand’s study as supporting the assertion “that the total death rates HAVE NOT CHANGED.”
This is false information. This is not entirely iWe’s fault; though everyone should be very careful about information sources at this time.
I have decided to rebut Dr. Briand’s erroneous analysis. The article summarizing her analysis is here; the Johns Hopkins explanation of its withdrawal is here, and an hour-long video explaining her results is here. My data sources and methods will be set forth in the technical note at the bottom of this post.
I. Age Distribution Analysis
The lead graph in Dr. Briand’s analysis relies upon this chart, showing weekly deaths in the US from February to September 2020:
This graph is color-coded by age category, with reported deaths of older people at the top, showing the percentage of total deaths in each age category — for example, the light blue bar at the top shows that about 30% of all reported deaths in the US have occurred among people aged 85 and up. This graph includes all reported deaths, from all causes, not just Covid.
This does not show the absence of an increase in the number of deaths. It shows that the proportion of deaths by age category was not noticeably changed by Covid. The article setting forth Dr. Briand’s analysis claims:
Surprisingly, the deaths of older people stayed the same before and after COVID-19. Since COVID-19 mainly affects the elderly, experts expected an increase in the percentage of deaths in older age groups. However, this increase is not seen from the CDC data. In fact, the percentages of deaths among all age groups remain relatively the same.
Who are these “experts”? Why would this be at all surprising?
We know that Covid mostly kills old people. What about other causes of death? Do you think that other causes of death affect mostly young people? Of course not. Most people who die are old, thank God. The alternative is for the young to die in large numbers, which would be tragic. Something’s gotta get us all in the end.
But to rebut Dr. Briand’s claim requires empirical evidence, so I consulted two sources: (1) the CDC page for Covid deaths, which reports deaths by age category for (approximately) February-November 2020, and which lists both Covid-involved deaths and total deaths, and (2) the CDC report on final death figures for 2017, which reports all 2017 deaths by age category. This allowed me to calculate the total percentage of reported deaths, for each age bracket, for the following 3 periods:
- All deaths in 2017
- Covid deaths in 2020 (approx. February-November)
- Non-Covid deaths in 2020 (approx. February-November)
Here is the result:
It’s hard to tell the difference between the periods, isn’t it? Unsurprisingly, Covid deaths and non-Covid deaths occur mostly among the old — roughly 30% among people aged 85 and over, another 24-27% among people aged 75-84, and another 19-22% among people aged 65-74.
Thus, we would not expect Covid to significantly change the distribution of deaths by age category.
Remember the quote from the article presenting Dr. Briand’s analysis: “[E]xperts expected an increase in the percentage of deaths in older age groups.” Again, what experts? None are cited. Their hypothesis is just plain silly. We all know that deaths occurred overwhelmingly among the old, even before Covid.
I find it both surprising and disturbing that people would be misled by such nonsense. This is the sort of ridiculous claim that a thoughtful, critical reader should immediately recognize as implausible. Proving it to be incorrect takes additional work — for example, you need to find the data on the CDC website, and you need to know how to use Excel or a similar program to analyze and graph the data.
Fortunately for you, dear reader, I happen to know how to do both of these things. :)
But seriously, the real danger is the uncritical acceptance of implausible and unsubstantiated claims. This seems particularly common when the erroneous conclusion is congenial to one’s political or moral position. Be careful about this. There is very, very bad information coming from the Leftist media, and there is very, very bad information promoted by alternative conservative sites.
II. The “Excess Deaths” Analysis
Dr. Briand’s analysis does not specifically analyze the “excess deaths” information, but the article reporting her findings claims:
Briand also noted that 50,000 to 70,000 deaths are seen both before and after COVID-19, indicating that this number of deaths was normal long before COVID-19 emerged. Therefore, according to Briand, not only has COVID-19 had no effect on the percentage of deaths of older people, but it has also not increased the total number of deaths.
This conclusion is absolute nonsense, though the first part of the first sentence is correct. The number of weekly deaths generally ranged from 50,000 to 70,000 throughout 2017, 2018, 2019, and 2020. This is immaterial. It is the sort of thing that is meant to mislead.
There is a good graph from the CDC that rebuts this claim (link here):
But perhaps you don’t trust this one. I haven’t personally verified the information, and when you look closely, it’s not actually reporting death counts. The blue bars are “Predicted number of deaths from all causes,” and if you follow the link and point to an individual data bar, it reports three numbers: “Average expected number of deaths,” “Upper bound threshold for excess deaths,” and “Predicted (weighted) number of deaths.”
I don’t think that the CDC is playing games with us. The notes explain that there is a lag in the reporting of deaths. I deduce that this particular graph uses averages to predict what the total number of reported deaths will be for each week once reporting is complete.
Also, the orange line is the “upper bound threshold for excess deaths,” which appears to be a sort of “trip-wire” for determining whether there is a serious problem. As you can see, the orange line runs quite a bit above the actual reported deaths.
I decided to so my own analysis. I was able to download data from the CDC for actual reported weekly deaths, from 2017 to 2020. The data file included both the “upper bound threshold for excess deaths” and the “average expected count.” In order to check the plausibility of the CDC’s “average expected count,” I separately calculated the weekly average deaths for 2017, 2018, and 2019, on a week-by-week basis — that is, I averaged the reported deaths for the first week in January 2017, 2018, and 2019; then for the second week in January for each such year, and so on. (One caveat — the data set did not include the first week in January 2017, for that one week, my calculated average includes only 2018 and 2019 data). Here is the result:
The red line is the actual reported deaths for each week in 2020. Note the spike starting around Week 13 — mid-March. That’s Covid. Important caveat: Recent deaths are significantly under-reported, as the data is not yet in. Do not interpret the decline at the right side of this graph as an actual decline in deaths. It is almost certainly a result of the lag in reporting.
I have two lines for the “normal death” threshold. The blue line is the CDC’s “average expected count.” The orange line is the actual weekly average for 2017, 2018, and 2019, calculated by me. As you can see, the blue line and the orange line are virtually identical, giving me high confidence in the CDC’s calculation of the “average expected count.”
The green and yellow lines are “excess deaths.” The yellow line shows the excess of actual reported deaths in 2020 compared to the CDC’s “average expected count” for each week, while the green line shows the excess of actual reported deaths for 2020 compared to the weekly average from 2017, 2018, and 2019.
This graph demonstrates that Covid is real. Its effect shows up in a sharp spike in reported deaths, starting in mid-March 2020. This is entirely in accord with Covid death figures reported elsewhere.
I performed one more calculation — my own estimate of “excess deaths,” using the green line — i.e. the increase in weekly deaths reported in 2020, compared to the average weekly death figure for the corresponding week in 2017, 2018, and 2019.
For the period from Week 12 to Week 44 — i.e. the week ending March 21, 2020 through the week ending October 31, 2020 — my calculation indicates 316,800 excess deaths. I did not include reported deaths in November because the data is evidently incomplete.
This rebuts Dr. Briand’s claim that Covid has not increased the total number of deaths in the US. That claim is demonstrably incorrect.
III. My comments
After initially taking down the article without explanation, Johns Hopkins posted an explanation, and re-posted the original article. My initial impression is to think that this response by Johns Hopkins is admirable, though on reflection, this is a sad commentary on the state of our academic and public discourse. Johns Hopkins acted properly and honorably, but this should not be surprising. This is how everyone should behave, all of the time. Still, they deserve kudos for doing the right thing.
I will anticipate an objection (which was made to my initial rebuttal, in the comments to iWe’s post). Some may claim that they do not trust the CDC figures. I see no basis for such suspicion. This is a particularly troubling response by anyone who found Dr. Briand’s flawed analysis to be convincing, as she relied on CDC data.
I am critical of one thing that Johns Hopkins stated in its explanation of its withdrawal of Dr. Briand’s study. “As assistant director for the Master’s in Applied Economics program at Hopkins, Briand is neither a medical professional nor a disease researcher.” Fair enough, but Dr. Briand (here) holds a PhD in Economics and, for years, has taught econometrics and statistics. She doubtless knows far more about statistical analysis and mathematical modeling than 99% of medical professionals. So do I, it turns out, which makes me pretty weird (even among lawyers).
A final note. There is clear empirical evidence of a significant spike in total deaths in the US, precisely corresponding to the Covid pandemic. However, it is not necessarily the case that all of those deaths were the direct result of Covid. Some may have been the result of the response to Covid, ranging from deaths of despair (such as suicide) to deaths from other causes due to failure to seek medical care. Sorting out the precise impact of Covid itself, as distinguished from its secondary effects, with require further work. Actually, I think that this is the sort of thing that Dr. Briand was trying to do, but the details got lost due to her top-line errors.
I hope that this analysis proves helpful.
IV. Technical Notes
My data source for 2020 reported deaths by age group, both Covid and total, is the CDC (here). I calculated the non-Covid total as the difference between all deaths and Covid-involved deaths. This page will be updated periodically, so here is the screenshot:
My data source for 2017 deaths by age category is the National Vital Statistics Reports, Vol. 68, No. 9, “Deaths: Final Data for 2017” (here). The relevant data is in Table 2 on page 23.
Note that in both cases, I had to combine certain age categories to match the 10-year increments reported in Dr. Briand’s analysis.
My data source for reported deaths and excess death calculations, 2017-2020, is also the CDC (here). This is the same page as the blue-bar excess-deaths graph reproduced above. To access the data, scroll down to the “Options” section, “Download Data” subsection, “CSV Format” column, and click on “National and State Estimates of Excess Deaths.” This will allow you to download a .csv file that can be opened by Microsoft Excel.
Published in Healthcare
The Briand article was withdrawn but now it’s back up?
George Floyd tested positive for Covid
but why are people not getting the flu?
Flu and covid symptoms are similar?
nursing home and long term care residents
Tracking cases for the first time to justify ’emergency lockdowns’ for “15 days”
Have any maladies increased this year?
My guess is it’s few or none to over report covid deaths
remember that statement from the CDC a few weeks ago, only 6% die “of covid” vs 94% die “with covid”
Jerry exaggerates when he says Briand is “deeply flawed”.
More than half covid deaths have occurred in nursing homes where the average expectancy is 14 months.
CDC said 94% die “with covid” vs 6% die “because of covid”
Average life expectancy 78.8 < average covid death age (79 or 80)
Has US life expectancy decreased because of covid?
Have life insurance premiums increased because of covid?
If there are “excess deaths” in 2020 will there be fewer deaths in 2021?
Covid deaths are real but excess deaths have been exaggerated
“If [the COVID-19 death toll] was not misleading at all, what we should have observed is an increased number of heart attacks and increased COVID-19 numbers. But a decreased number of heart attacks and all the other death causes doesn’t give us a choice but to point to some misclassification,” Briand replied.
The death of a loved one, from COVID-19 or from other causes, is always tragic, Briand explained. Each life is equally important and we should be reminded that even during a global pandemic we should not forget about the tragic loss of lives from other causes.
According to Briand, the over-exaggeration of the COVID-19 death number may be due to the constant emphasis on COVID-19-related deaths and the habitual overlooking of deaths by other natural causes in society.
Look at this graph closely… covid cancels out the other causes of death and vice versa
Isn’t heart disease a major comorbidity for covid mortality?
The correct answer is yes
“I write to rebut the claims of Dr. Genevieve Briand, a senior lecturer at Johns Hopkins who holds a PhD in Economics and recently released a study questioning the Covid death statistics. The paper was subsequently withdrawn by Johns Hopkins, quite properly in my view. Dr. Briand’s analysis is deeply flawed.”
@arizonapatriot — Hos is this different from cancel culture? Even if her analysis is deeply flawed (which it isn’t) why should her paper be withdrawn? How is this different with the media supressing the Hunter Biden story?
I’m beginning to think that her problem was incomplete data.
I was avoiding other work, so this evening I took the data Jerry pointed to on this CDC page, both the provisional weekly cause-of-death numbers (2019 and 2020) and the final numbers from 2014-2018. I summed the data by cause of death for each year between weeks 16 and 41 (to match the start of COVID, and where the data is a little more free of revisions)–roughly mid-March to mid-October. Then I compared total deaths over the number of weeks. For all causes, the total over the 25-week period is 1,611,050 in 2020, compared with 1,366,918 in 2019, and averaged 1,313,803 between 2014 and 2019. The differences are 244,132 and 297,247, respectively, much larger than the COVID-19 (Multiple Cause of Death) at 182,344 between weeks 16 and 41. This data is not broken down by age, and it is slightly different that the provisional data in the Deaths by Week by State and by Age series, which does not list cause of death.
I will be interested in what the final data show, but this appears to contradict Dr. Briand’s finding that other causes of death are lower, at least to this point in time. But to be fair she was talking about the oldest portion of the population. However, that may be due to the time periods she chose to analyze, and the complete final data may show something else. One interesting thing, subtracting total deaths from total natural causes, which would include accidents, homicides, suicides, overdoses, etc., in 2020 are up 4.1% over 2019. Again, my statistical work is subject to review and revision, and as always subject to change if I’ve misunderstood the data.
Tables to follow:
Here are tables from that data:
Second part of table:
“Fail to Reject” means that we cannot reject the hypothesis that the 2020 deaths by cause are equal to the mean of the 2014-19 deaths.
Total deaths appear to be much larger in 2020 than in the 2014-19 period. Not very much seems to be smaller, as was claimed. It may have been true for some weeks in the period, but not the whole period. Of course, these data will change when the full data is complied.
BC, not exactly. The article was withdrawn without adequate explanation, but then Johns Hopkins posted an explanation that included a link to a PDF of the article, with the PDF prominently marked “Retracted.” In the circumstances, this was an improvement, but still not as good as encouraging a discussion by the author and her critics.
Mark, thanks. I thought that you were citing a source other than Dr. Briand (or, more specifically, the article by Yanni Gu summarizing Dr. Briand’s data). Dr. Briand’s assertions and conclusions are simply wrong (about the two issues addressed in the OP; she may be correct about the change in death figures in specific categories in April 2020). But this clears up our misunderstanding.
Dr. Briand’s analysis is deeply flawed, as I demonstrated. Look again at my Excess Deaths graph in the OP. Dr. Briand reportedly said that Covid “has also not increased the total number of deaths.” That is not true.
I don’t think that this is a “cancel culture” problem, because Dr. Briand’s analysis was factually incorrect. I can give you an example.
I just calculated US homicide rates by race of offender, for 2019. If I’ve done the calculation right, the rate is about 2.7 per 100,000 for whites, and about 20.6 per 100,000 for blacks. The black homicide rate is about 7.7 times the white homicide rate, by my calculation.
If I reported this in a paper, and my institution withdrew the paper even though my calculations were correct, but because they were politically unpopular, that would be an example of “cancel culture.” It would be suppression of the truth.
On the other hand, what if I did the calculation wrong, and erroneously reported that the black homicide rate was 20 times higher than the white rate? Then, the hypothetical withdrawal of my paper by my institution would be perfectly proper, because the paper would be wrong. Leaving it out there would be spreading misinformation.
Dr. Briand’s claims were factually wrong, as I demonstrated in the OP. It is perfectly appropriate for her institution — Johns Hopkins University — to withdraw that article. Johns Hopkins went one better, by making the erroneous article available with a note that it has been withdrawn. This allows the rest of us to confirm that Dr. Briand’s analysis was incorrect, and that the withdrawal of the article was proper.
Which it was.
I believe Dr Briand was look at 6 month period between mid March and mid September, 26 weeks
Fail to reject null hypothesis = not significant
No that is a really good point. COVID is the same. I saw CDC recently update is that only about one in 8 who get COVID are tested or test positive. If I remeber correctly that is in the same range as the flu..
Has Briand been given the opportunity to counter her critics?
Mr. Crypto$, I don’t know what opportunities have been extended to her, nor what she may have said/published subsequently on other platforms.
The Johns Hopkins explanation of the retraction, along with a link to the original (retracted) document, is here:
A closer look at U.S. deaths due to COVID-19
I see no changes there, and nothing to indicate that Ms. Briand has responded.
Incidentally, some of the thousand or so comments on that page are interesting.
1/8 = 12.5%
On the Diamond Princess cruise in Feb or Mar, 20% of crew and passengers tested positive. The rest did not despite exposure in close quarters.
An epidemiologist at Oxford ( I forget her name, she signed the Barrington Declaration, one of the 3 originals) estimated 80% of population has natural T cell immunity against covid