I wrote to David Leonhardt, May 31, when he wrote a recent misinformation column on masks and mandates not working: he infuriates me--he has a large audience and he "cherry picks" his experts and facts.
Here's what I wrote:
"I’m a physician. The pandemic drags on, with the US having far more deaths than other countries and misinformation contributes to our failure of public health.
Today’s newsletter is a classic of your brand of misinformation: citing experts who share your bias and presenting opinions as facts. Why don't you quote Bob Wachter's most recent Twitter post: If you're trying to stay well, time to up your game
Interview Gregg Gonsalves, Jeremy Faust, Esther Choo, Katelyn Jetelina— not just your small, curated group of proponents of failed public health policy. Your column is an excellent example of selection bias.
As this current surge is raging, quit pushing vaccination and Paxlovid as the only tools we have. Paxlovid was only tested on unvaccinated people during delta. Vaccination protects the individual and the current variants have immune escape. You know this.
Masks do work, and yet your column argues to drop them. Mandates work also--they're just not popular.
How about some balanced and accurate reporting?. We’re failing at containing this virus and more virulent variants are coming soon. Your brand of misinformation is particularly damaging.
I avoid Fox News, I never thought I'd have to avoid the NYT. "
I'm an NYT subscriber, and I have found The Morning's commentary on the pandemic to be VERY disappointing. My particular vantage point is that of someone with Long COVID. I think Leonhardt has routinely tried to show (peacock-style) that he has a new and interesting take on the pandemic. Unfortunately, his new/interesting takes are often misleading to the point of outright misinformation. In the Long COVID support community, we have given given everything he writes a standing trigger warning.
I subscribed just to make a comment. I saw an outraged thread from Yale about this. I want to say that, from the beginning, there was no representation at the table for most Americans, and especially PoC whose families tend to live in multigenerational houses. There was complete disregard for people who were unable to lockdown. So, if we want to talk about confounding factors, we need to look at all of this. I have worked for NIH, NSF, with some of the people who were part of this response and I found the thought in the NYT article that there is some change in the distribution of illness and death of great interest. We don't know if it is vaccines or something else, but it is interesting.
I worked with national labs and, in the beginning, they were interested in developing technology to increase safety and measurement of safety in indoor spaces. I could not get people in charge to hear this. At this point, humility could be very productive . I think it is a good time for representative, national groups to revisit models, approaches to decreasing transmission, etc. - including those that were suppressed and see which ones were best predictors of outcomes.
I am a liberal academic, but wish that all of us were better at these challenging conversations. Maybe this did confound factors -but people in biology, especially, do that all the time, in all honesty. I worked with NIH to increase statistical expertise in graduate training. So few understand the value of a technical replicate, for example. The politics of the time, with DJT president, threw everyone into System 1 thinking and alternate models seemed like threats. Some are awful, but maybe there are some that were more predictive of what happened than others.
I see so few authentic PoC at the table, I didn't see one PSA about best masks, best way to wear masks, where to put them, etc. We can do better and we need to have some insight so we serve all of our citizens better next time.
You are welcome. There is so much we need to know and learn. This public health crisis came at a bad time politically, so talk about confounding factors! I would love to see a deep review of our response to covid - because there were so many ideas out there that couldn't find any sunlight. It felt terrible that whole families were getting sick and dying because they lived and ate together and we didn't have any real prevention plans. Trying to figure out who needed food in rural areas and getting it to them was more of a challenge than it needed to be. I looked for programmers to "Amazonify" our towns, so small businesses could survive online together. This is why we needed a bigger table. Take care and stay healthy.
PLEASE send this to the NYT, they seem to be caught up in the current (not mindful) “breaking news” affecting so much speculation that passes for news currently. We deserve better from our “paper of record”!
Thanks for your deep dive into the data to provide a more accurate picture of racial disparities with Covid-19. I would encourage you to consider an Op-Ed piece (in your spare time - ha-ha!) for the NYT or a YLE-David Leonhardt point-counterpoint article. Thanks again for helping the NYT to keep its news fit to print.
I read that this morning and since it was in the NYT I thought it must have truth to it. I am so grateful you got this out and hope they recognize this publicly. We need the truth now and going forward. Instant news is not always the best news.
This is why I keep coming back to your analysis!!! You lay it all out in easy to understand language and help us get past the hyperbolic, misleading, and damaging politics of COVID. You are able to reach so many people with clear, understandable language. Please keep up your excellent work.
I read your newsletter to help me understand the sciences of public health. I am not a scientist but have a decent background in the basics including statistics. I want to always be aware of how mis- and disinformation is presented so maybe I can help others see the lies. Thank you for helping.
I was also puzzled by the NY Times coverage this morning. (I'm a biostatistician.) In addition to the age adjusted Simpson's paradox that you bring up, I was also thinking about the denominator--the population at risk. The NYTimes was highlighting vaccination (fair enough), but I thought there also should have been a discussion of the fact that so many people of color died early in the pandemic, that I suspect there are fewer people at high risk of death remaining in those communities.
Latinos, Blacks, and Native Hawaiians/Pacific Islanders represent over 32% of the US population, or about 106 million people. While the age distribution of PoC skews to younger ages, it is not so significant that all our older relatives died from Covid. This is the problem with percents vs numbers and using different groupings as the total, for example, covid illness, death, total US population, population of subgroups, etc. Yes, there is a concern about having lost so many elders and so much knowledge, but our cultural traditions are strong, especially in protecting elders.
Sorry, I did not mean to suggest that COVID significantly changed the age distribution. (That's a much longer standing issue.) A lot of my thinking revolved around unmeasured confounding--there's still a lot we don't know about what makes one person died from COVID and another survive. Age and comorbidities play a large role, but still, most 80 year olds survive COVID and some 40 year olds die, and we can't yet explain it. So in January 2020, there was a pool of people who were at very high risk of death (for biological reasons we can't yet entirely explain--and to be clear, biological reasons absolutely can be created by social determinants of health). Since then, that pool has been shrinking as it's members have succumbed. And they have succumbed at differential rates, so early on, the rate for people of color was much higher than for whites--so now, there are a lot of white people in that pool. Basically, I'm thinking about another Simpson's paradox, but with non-age related susceptibility to death.
I'm sure David Leonhardt is a nice guy and means well, but he has been committing this type of innumeracy for two years. This latest piece isn't much different.
I don't want to judge him too harshly because what he is doing is nothing unique in journalism. Highly recommend John Allen Paulos' "A Mathematician Reads the Newspaper" for an accessible collection of essays on this phenomenon. (Most examples come from early 90's, but still interesting).
This particular line of thought he has been building on since March [1], though his recent piece on May 31 was probably more controversial.
Katelyn - I write a Covid newsletter that focuses on my native state of Alabama. I've written over 450 letters, beginning in May 2020. In my state, the handful of Black Belt counties that have majority-black populations have generally fared better than white rural counties (located in northeast Alabama and the Wiregrass region of southeast Alabama) in terms of vaccination rates, infection rates and deaths. It makes me wonder if you might be making a similar mistake to the NYT. Adjusted for age is one thing, but what about if you adjust for geography. In the South, in particular, rural white citizens are overwhelmingly pro-Trump so they have lagged well behind in vaccination rates and other important metrics, such as mask-wearing and social distancing. Black citizens, on the other hand, have taken greater precautions, on the whole.
there are SO many confounders. this is why we build complex models (called multivariable regressions) to account for all of them to get closer to the "truth". I agree that geography is likely a confounder, same with religion, gender, marital status, zip code, and much more. Age, though, is a particularly strong confounder given it's the number one predictor of COVID19 death. like I said in the post, all of these questions are incredibly important. And they need to be answered but with the right models and peer reviewed process. Unfortunately that takes time and can't be answered in a hundred words and one univariate graph.
Multivariate analysis is the best we can do and I'm speaking as a person whose father was active in the MEPS group. But there is always an absolutely unavoidable reductionism in the analysis so all that can be hoped for is as tight an approximation as we have variables that don't descend to noise. Kudos to YLE for toiling to get us educated on this.
This discussion thread piques my interest because I am trying to build a better mental model of the statistical factors in general for problems like this. I have no formal stat training, but I definitely see the value in understanding better -- especially how phenomena such as ignoring confounding factors interplays with how misinformation originates and metastasizes.
On this particular topic, I'm wondering if you can clarify whether my current intuition is correct or not: While geographical-related variables do seem like very valid confounders, age-related variables appear to be much more dominant in the population overall?
Katelyn's specific statement in this post on the topic says "This is important because age is the strongest risk factor, by far, for dying of COVID." I suspect you've already established that position fairly firmly in previous posts, but I'm curious if you have any citations handy on the topic?
Age has always been the #1 driver of outcome, followed distantly by weight and the various comorbidities which go with obesity (hypertension, poor circulation, etc).
Consider there are roughly 50,000,000 people above 65 in the US, despite representing less than 17% of the total population that age group accounts for 75% of all covid deaths. If you extend that back to age 50+, you account for 94% of all Covid deaths.
By comparison under 30 deaths represents less than a single percent.
Here's a quick table showing covid deaths, total deaths, and %covid as cause of death per age group from CDC
I have a question. Why aren't Hispanics/Latinos in the second group of graphs? I didn't notice that at first, but that seems amazing, since we are the largest "minority" group - with 63% of Hispanics in the US having Mexican-ancestry. Did you find that odd? Thanks.
I wrote to David Leonhardt, May 31, when he wrote a recent misinformation column on masks and mandates not working: he infuriates me--he has a large audience and he "cherry picks" his experts and facts.
Here's what I wrote:
"I’m a physician. The pandemic drags on, with the US having far more deaths than other countries and misinformation contributes to our failure of public health.
Today’s newsletter is a classic of your brand of misinformation: citing experts who share your bias and presenting opinions as facts. Why don't you quote Bob Wachter's most recent Twitter post: If you're trying to stay well, time to up your game
Interview Gregg Gonsalves, Jeremy Faust, Esther Choo, Katelyn Jetelina— not just your small, curated group of proponents of failed public health policy. Your column is an excellent example of selection bias.
As this current surge is raging, quit pushing vaccination and Paxlovid as the only tools we have. Paxlovid was only tested on unvaccinated people during delta. Vaccination protects the individual and the current variants have immune escape. You know this.
Masks do work, and yet your column argues to drop them. Mandates work also--they're just not popular.
How about some balanced and accurate reporting?. We’re failing at containing this virus and more virulent variants are coming soon. Your brand of misinformation is particularly damaging.
I avoid Fox News, I never thought I'd have to avoid the NYT. "
The word "mask" is not in the article. "Paxlovid" is not in the article. The figures are from the CDC.
I wrote that May 31 to the misinformation he presented on masks and mandates in his column
She replied to this article from May 31 by David Leonhardt
https://www.nytimes.com/2022/05/31/briefing/masks-mandates-us-covid.html
This was in response to his column from last week
I'm an NYT subscriber, and I have found The Morning's commentary on the pandemic to be VERY disappointing. My particular vantage point is that of someone with Long COVID. I think Leonhardt has routinely tried to show (peacock-style) that he has a new and interesting take on the pandemic. Unfortunately, his new/interesting takes are often misleading to the point of outright misinformation. In the Long COVID support community, we have given given everything he writes a standing trigger warning.
I subscribed just to make a comment. I saw an outraged thread from Yale about this. I want to say that, from the beginning, there was no representation at the table for most Americans, and especially PoC whose families tend to live in multigenerational houses. There was complete disregard for people who were unable to lockdown. So, if we want to talk about confounding factors, we need to look at all of this. I have worked for NIH, NSF, with some of the people who were part of this response and I found the thought in the NYT article that there is some change in the distribution of illness and death of great interest. We don't know if it is vaccines or something else, but it is interesting.
I worked with national labs and, in the beginning, they were interested in developing technology to increase safety and measurement of safety in indoor spaces. I could not get people in charge to hear this. At this point, humility could be very productive . I think it is a good time for representative, national groups to revisit models, approaches to decreasing transmission, etc. - including those that were suppressed and see which ones were best predictors of outcomes.
I am a liberal academic, but wish that all of us were better at these challenging conversations. Maybe this did confound factors -but people in biology, especially, do that all the time, in all honesty. I worked with NIH to increase statistical expertise in graduate training. So few understand the value of a technical replicate, for example. The politics of the time, with DJT president, threw everyone into System 1 thinking and alternate models seemed like threats. Some are awful, but maybe there are some that were more predictive of what happened than others.
I see so few authentic PoC at the table, I didn't see one PSA about best masks, best way to wear masks, where to put them, etc. We can do better and we need to have some insight so we serve all of our citizens better next time.
Maggie, I'm so glad you're here and so thankful for what you shared! Welcome. There is a great community here.
You are welcome. There is so much we need to know and learn. This public health crisis came at a bad time politically, so talk about confounding factors! I would love to see a deep review of our response to covid - because there were so many ideas out there that couldn't find any sunlight. It felt terrible that whole families were getting sick and dying because they lived and ate together and we didn't have any real prevention plans. Trying to figure out who needed food in rural areas and getting it to them was more of a challenge than it needed to be. I looked for programmers to "Amazonify" our towns, so small businesses could survive online together. This is why we needed a bigger table. Take care and stay healthy.
PLEASE send this to the NYT, they seem to be caught up in the current (not mindful) “breaking news” affecting so much speculation that passes for news currently. We deserve better from our “paper of record”!
Thanks for your deep dive into the data to provide a more accurate picture of racial disparities with Covid-19. I would encourage you to consider an Op-Ed piece (in your spare time - ha-ha!) for the NYT or a YLE-David Leonhardt point-counterpoint article. Thanks again for helping the NYT to keep its news fit to print.
I read that this morning and since it was in the NYT I thought it must have truth to it. I am so grateful you got this out and hope they recognize this publicly. We need the truth now and going forward. Instant news is not always the best news.
Me too. Though I don’t actually understand this newsletter article either. I understand the premise but not how you can adjust for age.
I just forwarded your email to themorning@nytimes.com
I emailed David before this post went out (I know he follows this newsletter too). He quickly wrote back and we will be chatting by phone soon.
Here’s what I wrote: Hi David and team,
You may want to read this email from
Katelyn Jetelina. Looks like your story needs to be corrected. And fast.
I’ve noticed a definite bias in his Covid posts so I’m not surprised by this one. And I’m glad you are calling him out!
This is why I keep coming back to your analysis!!! You lay it all out in easy to understand language and help us get past the hyperbolic, misleading, and damaging politics of COVID. You are able to reach so many people with clear, understandable language. Please keep up your excellent work.
I read your newsletter to help me understand the sciences of public health. I am not a scientist but have a decent background in the basics including statistics. I want to always be aware of how mis- and disinformation is presented so maybe I can help others see the lies. Thank you for helping.
Amazing you got this wise analysis out so quickly!!!
Thanks for your quick, eloquent and informative correction.
I was also puzzled by the NY Times coverage this morning. (I'm a biostatistician.) In addition to the age adjusted Simpson's paradox that you bring up, I was also thinking about the denominator--the population at risk. The NYTimes was highlighting vaccination (fair enough), but I thought there also should have been a discussion of the fact that so many people of color died early in the pandemic, that I suspect there are fewer people at high risk of death remaining in those communities.
Latinos, Blacks, and Native Hawaiians/Pacific Islanders represent over 32% of the US population, or about 106 million people. While the age distribution of PoC skews to younger ages, it is not so significant that all our older relatives died from Covid. This is the problem with percents vs numbers and using different groupings as the total, for example, covid illness, death, total US population, population of subgroups, etc. Yes, there is a concern about having lost so many elders and so much knowledge, but our cultural traditions are strong, especially in protecting elders.
Sorry, I did not mean to suggest that COVID significantly changed the age distribution. (That's a much longer standing issue.) A lot of my thinking revolved around unmeasured confounding--there's still a lot we don't know about what makes one person died from COVID and another survive. Age and comorbidities play a large role, but still, most 80 year olds survive COVID and some 40 year olds die, and we can't yet explain it. So in January 2020, there was a pool of people who were at very high risk of death (for biological reasons we can't yet entirely explain--and to be clear, biological reasons absolutely can be created by social determinants of health). Since then, that pool has been shrinking as it's members have succumbed. And they have succumbed at differential rates, so early on, the rate for people of color was much higher than for whites--so now, there are a lot of white people in that pool. Basically, I'm thinking about another Simpson's paradox, but with non-age related susceptibility to death.
Yes, great questions.
I'm sure David Leonhardt is a nice guy and means well, but he has been committing this type of innumeracy for two years. This latest piece isn't much different.
I don't want to judge him too harshly because what he is doing is nothing unique in journalism. Highly recommend John Allen Paulos' "A Mathematician Reads the Newspaper" for an accessible collection of essays on this phenomenon. (Most examples come from early 90's, but still interesting).
This particular line of thought he has been building on since March [1], though his recent piece on May 31 was probably more controversial.
[1] https://www.eugyppius.com/p/nyt-do-covid-precautions-work-yes?s=r
Katelyn - I write a Covid newsletter that focuses on my native state of Alabama. I've written over 450 letters, beginning in May 2020. In my state, the handful of Black Belt counties that have majority-black populations have generally fared better than white rural counties (located in northeast Alabama and the Wiregrass region of southeast Alabama) in terms of vaccination rates, infection rates and deaths. It makes me wonder if you might be making a similar mistake to the NYT. Adjusted for age is one thing, but what about if you adjust for geography. In the South, in particular, rural white citizens are overwhelmingly pro-Trump so they have lagged well behind in vaccination rates and other important metrics, such as mask-wearing and social distancing. Black citizens, on the other hand, have taken greater precautions, on the whole.
there are SO many confounders. this is why we build complex models (called multivariable regressions) to account for all of them to get closer to the "truth". I agree that geography is likely a confounder, same with religion, gender, marital status, zip code, and much more. Age, though, is a particularly strong confounder given it's the number one predictor of COVID19 death. like I said in the post, all of these questions are incredibly important. And they need to be answered but with the right models and peer reviewed process. Unfortunately that takes time and can't be answered in a hundred words and one univariate graph.
Multivariate analysis is the best we can do and I'm speaking as a person whose father was active in the MEPS group. But there is always an absolutely unavoidable reductionism in the analysis so all that can be hoped for is as tight an approximation as we have variables that don't descend to noise. Kudos to YLE for toiling to get us educated on this.
This discussion thread piques my interest because I am trying to build a better mental model of the statistical factors in general for problems like this. I have no formal stat training, but I definitely see the value in understanding better -- especially how phenomena such as ignoring confounding factors interplays with how misinformation originates and metastasizes.
On this particular topic, I'm wondering if you can clarify whether my current intuition is correct or not: While geographical-related variables do seem like very valid confounders, age-related variables appear to be much more dominant in the population overall?
Katelyn's specific statement in this post on the topic says "This is important because age is the strongest risk factor, by far, for dying of COVID." I suspect you've already established that position fairly firmly in previous posts, but I'm curious if you have any citations handy on the topic?
Age has always been the #1 driver of outcome, followed distantly by weight and the various comorbidities which go with obesity (hypertension, poor circulation, etc).
Consider there are roughly 50,000,000 people above 65 in the US, despite representing less than 17% of the total population that age group accounts for 75% of all covid deaths. If you extend that back to age 50+, you account for 94% of all Covid deaths.
By comparison under 30 deaths represents less than a single percent.
Here's a quick table showing covid deaths, total deaths, and %covid as cause of death per age group from CDC
https://imgur.com/a/lukbmEC
Source: https://data.cdc.gov/NCHS/Provisional-COVID-19-Deaths-by-Sex-and-Age/9bhg-hcku
I have a question. Why aren't Hispanics/Latinos in the second group of graphs? I didn't notice that at first, but that seems amazing, since we are the largest "minority" group - with 63% of Hispanics in the US having Mexican-ancestry. Did you find that odd? Thanks.
Thank you for clarifying this! I hope you and other scientists on this thread wrote in the NYT's comments section!