I took the bait, I'm ashamed. I try to keep an open mind, so I read some of it.
The thrust of the article is that vaccines work well at preventing infection, but that if you do get infected, it is _possible_ that you might die at a higher rate. _Possible_ comes from a big spread in the standard deviation of P(death|infected & vaccinated). I didn't check his calculations, I took his numbers at face value and just looked at the charts.
The author uses big standard errors to make claims about what could or could not be true, because the standard errors straddle both favorable and unfavorable statistics for the vaccine.
However when the standard errors are too small for this dubious reasoning, they come out in favor of the vaccine.
The author appears to me to be using motivating reasoning to push an agenda.
Edit: sorry I am not completely right, because I wrote this comment before thoroughly reading the whole article. Sometimes his standard errors come out against the vaccine, and for all I know his math could be right. But I don't want to check the math because I have limited time and I don't care if my death rate goes up given conditions {x, y, z.}
I care if getting the vaccine makes me less likely to get sick and die, which it does, because if you look at all the tables and charts that's exactly what it shows.
It's even worse than that. Aside from deviations, this is a main point in the article:
> The death rate if infected was always going to be higher in the vaccinated groups if most of the vaccinated were those likely to die in the first place.
IOW: Those most likely to die were most likely to get vaccinated. Obviously true, but not interesting - a correlation we need to account for. When you do control for prior risk, the results are as expected, i.e., vaccines make you much safer.
The quoted tweet ("100x") was inaccurate and exaggerated. That's worth briefly correcting. But the lengthy article's points are only minor and pedantic, and do not change what we know, which is that our vaccines are very effective.
Yes. If I read the article correctly, the data in fact say you cannot determine whether the death rate in vaccinated people who become infected is different than in unvaccinated people.
However, 1) it isn't clear if Frieden is working from the same dataset as the author, and 2) a very slight re-wording of the initial statement would make it correct, i.e. "If you are _exposed_ to covid and you've been vaccinated you're about 100 times less likely to die."
There's several reason why even with highly effective vaccines, we might see worse infections when we do see them among the vaccinated. For example, if the vaccines work well, then there should be a bias where we see the most breakthrough infections among those for whom the vaccine doesn't do anything, i.e. people with compromised immune systems. So the people with normal immune systems will make up a much larger percentage of infections among the unvaccinated group, allowing them to drag down the average severity of infections - in the vaccinated group, they've been "removed" since most of them won't get infected, or have mild enough infections that they don't even notice, leaving the immunocompromised as a larger part of the infections that get counted in the statistics, and these infections are thus on average more severe.
> I care if getting the vaccine makes me less likely to get sick and die, which it does, because if you look at all the tables and charts that's exactly what it shows.
The author is actually in the range of an interesting point, even if the article is mostly foolish. If I had a hypothetical vaccine that was in fact composed of neurotoxin and killed the injectee instantly ... it would show 100% effectiveness in the cited tables, because the corpses would be unable to contract COVID. So you can't actually tell if the vaccine reduces your odds of getting sick and dying - we know the current crop are a bit rough and kill people in rare cases.
Covid deaths are so low in the 16-44 category that may be comparable to the vaccine. Although I think the hospitalisation numbers are still quite persuasive in favour of the vaccine.
There is a slippery joint conditional probability here. The article talks mostly of:
Pd1 = Prob(death GIVEN infected AND vaccinated)
vs
Pd2 = Prob(death GIVEN infected AND NOT vaccinated)
Which makes it easy to lose site of:
Pi1 = Prob(infected GIVEN vaccinated)
which is very very small compared to its opposite:
Pi2 = Prob(infected GIVEN NOT vaccinated)
Because Pi1 is so much smaller than Pi2 there are several consequences for Pd1 and Pd2:
- The small counts for "numerator" of Pd1 (zero in some cases) means large uncertainty in the ratio and that the centroid of the distribution is not particularly meaningful. Statistical fluctuations (just one more or one less death) will change Pd1 substantially.
- Statistically significant deviations between Pd1 and Pd2 do not point to a cause.
For example, it could be that those contributing to Pd1 all got a much higher viral load in order for the virus to break through the vaccine's defenses and high viral loads are known to correlate with death so once vax protection is defeated it is game over. The distribution of viral load exposure may even be the same in the Pd2 case, but for the unvaccinated a lesser load can be fatal. This explanation is consistent with P1d possibly being greater than P2d AND consistent with the given anecdotes of vaccinated tweeters saying covid is just the flu, bro.
The main take away is still: you do NOT want to get this shit and vaccine AND masking will help achieve that goal. And, unless you are a sociopath, you do NOT want to give this shit and vaccine AND masking will help achieve that goal.
The key insight of the article was that errors are reliably ignored or excused when they fit the desired narrative.
It's not hard to see how that would be effective anti-persuasion for persons inclined to have doubts about receiving an emergency authorized intervention. Rigorous honesty[1] would assuredly persuade at least some of those people and thus would raise the population vaccination rate, which I'm sure we all agree is a desirable outcome.
[1] For example, rather than chanting the mantra "safe and effective!" being honest about the tradeoffs and showing that proven risk management strategy indicates vaccination is the mathematically optimal choice.
> It's a given that you do not want to contract COVID-19. Vaccination and wearing a mask will help.
> You also do not want to give COVID-19 to others. Again, vaccination and wearing a mask will help.
How does that explain delta spread being biggest in highly vaccinated countries?
This is a great post and you can immediately see where they are going with this as soon as he quotes Dr. Tom Frieden.
It makes tons of sense that those who are vaccinated and get infected are more likely to die, simply because it would indicate a failure to develop antibodies and mount an effective immune response to the virus.
That being said I’m not sure such a rant is justified over a tweet like this. Dr. Frieden is more or less correct if you slightly re-word his sentence to: “Getting vaccinated decreases the chance of dying by approximately 100x”.
What I take home from this is that everyone is human, can make simple mistakes, and that we take things written on Twitter far too seriously. I’m sure that Dr. Frieden would not have made such a mistake if he wrote an article that was published somewhere reputable instead of a tweet that could have been sent while waiting for his coffee at Starbucks, or using the toilet.
There is a trade off in communicating science to the public in that using social media, TV, and newspapers reaches more people but almost always distorts the message. I’m not sure how exactly to balance this trade off, but I also think that some amount of responsibility lies with the reader to seek the truth instead of demanding that every sentence they read on the internet be as accurate as a peer reviewed journal.
The author makes a critical mistake when analyzing the BNT/Pfizer clinical trial. Over and over when discussing the results they mention calculating efficacy against "infections" using this data. But this specific clinical trial provided no data about infections. The metric measured was symptomatic COVID-19 disease.
This kind of mistake is understandable, but doesn't really inspire much confidence in a screed about the mistakes others are making in analyzing COVID-19 data.
> The real world data has shown that the death rate among the vaccinated, if infected with COVID, can be 3 to 5.7 times higher1 than the death rate of the unvaccinated.
Okay… but how much less likely are you to get “infected”?
How are you even defining “infected”, here? If virus gets in your body and then your immune system kills it before it does much damage, were you “infected”? If the answer is no, you've got pretty serious selection bias, because you're completely ignoring everyone who was completely protected by the vaccine in the “vaccinated” group, but paying attention to everyone who didn't really need a vaccine in the “unvaccinated” group: basically ignoring all the really-healthy, great-immune-systems, unlikely-to-die people in the “vaccinated” group so, proportionally, the sicker people take up more room. So you can't even say that the vaccine makes things worse for them! If it improves their chances by 100 times, as Dr. Tom said, but the selection bias is only paying attention to the 0.2% least immune people (those who got infected when exposed after vaccination), you'd expect to see a 5× higher death rate even though the actual death rate is 0.99× lower. (Made-up numbers.)
The author uses "evidence of infection" as the definition of infection, and labels all other definitions as misinformation, and writes at length about it as if it was not just a difference in definitions.
Obviously most of the people the author is criticizing are using "exposure that would cause detectable disease in non-immune individuals" as their definition, and I don't know if we have a better word for that.
> The author uses "evidence of infection" as the definition of infection,
So the author makes the implicit assumption that a successful vaccination doesn't prevent evidence of infection, and uses it to argue that vaccines are dangerous… then accuses everyone else of providing numbers without sufficient context?
Edit: no, actually.
> The reason, hidden in plain sight, is that a large number people who were never going to die, are no longer getting infected.
So the author does know… they were just getting spectacle in before explaining. And I do agree with the author's (eventual) point:
> Without careful control and understanding, one might erroneously conclude the Delta variant is is more lethal if you’ve been vaccinated, the vaccine is losing its efficacy, the vaccination is making people weaker, or some combination. While any of those are possible outcomes in this environment, by not being aware of the infection death rate issue from the start, because one is busy spreading misinformation about extra levels of protection that the data do not support, one misses how to properly control for these effects and analyze new data as it comes in.
But I really don't like the article. The author is the only one making those erroneous conclusions in the first place!
> 36 of 84611 in the unvaccinated versus 0 in 1066 in the vaccinated group. 36 in 84611 is roughly 1 in 2350, but we only had 1066 infected in the vaccinated group. There is not enough information to claim the death rate per infection is higher or lower, and that uncertainty is indicated in the graph above. That is worlds away from the relative immortality communicated by the efficacy number 100%.
Okay. But immediately after…
> In fact, if there had been 24 deaths in the vaccinated group the efficacy reported would have been 3%! Because it was looking at rates over time, 24 deaths would have been the death rate over time similar to 36 in the unvaccinated group. But clearly, among those infected, 36 in 84611 is a far lower death rate than 24 in 1066!
If? Now we’re onto hypotheticals. Here is an idea, you have so few because they were vaccinated!
Also the mental gymnastics of saying: “it’s better to not be vaccinated in case you get covid because you are less likely to die” is worthy of a mental gymnastics olympics gold medal.
I understand that the CDC guy may have worded things differently but directionally he is right.
But, again, one cannot just ignore the hypotheticals of Dr. Gator when it provides validation for your anti-vaxx stupid attitude.
In the “ What the numbers really showed” you use the lack of vaccinated sample size to justify something for the unvaccinated, but then immediately use that small number to your advantage to say “if 24 people had died then the vaccinated would have a higher death rate”. It’s disingenuous and doesn’t show anything.
Also if the vaccine prevents contraction then it should be counted as preventing death/hospitalization, but obviously those numbers aren’t obvious. Seems to me the vaccine IS preventing contraction so you would need to account for that wouldn’t you?
This article does actually make an interesting point, but its graphs are all bunk.
The vaccine works by priming the immune system to work quickly against a disease. It is quite likely that if someone develops symptoms despite their body being at peak alert for coronavirus then they are in trouble. The graphs bear that out, and do not alarm me. If we only select for cases where the person is physically susceptible enough for the virus to break through the vaccine protections to cause an infection it makes sense that the cases will be on average worse.
But there is an interesting point here that if the vaccine were killing people (eg, heart inflammation problems have been detected in some cases) then the "COVID-19-related-death" stats, etc, would be highly misleading. It would be more interesting to see all-cause mortality and the cited tables don't do that.
Readit News
The thrust of the article is that vaccines work well at preventing infection, but that if you do get infected, it is _possible_ that you might die at a higher rate. _Possible_ comes from a big spread in the standard deviation of P(death|infected & vaccinated). I didn't check his calculations, I took his numbers at face value and just looked at the charts.
The author uses big standard errors to make claims about what could or could not be true, because the standard errors straddle both favorable and unfavorable statistics for the vaccine.
However when the standard errors are too small for this dubious reasoning, they come out in favor of the vaccine.
The author appears to me to be using motivating reasoning to push an agenda.
Edit: sorry I am not completely right, because I wrote this comment before thoroughly reading the whole article. Sometimes his standard errors come out against the vaccine, and for all I know his math could be right. But I don't want to check the math because I have limited time and I don't care if my death rate goes up given conditions {x, y, z.}
I care if getting the vaccine makes me less likely to get sick and die, which it does, because if you look at all the tables and charts that's exactly what it shows.
> The death rate if infected was always going to be higher in the vaccinated groups if most of the vaccinated were those likely to die in the first place.
IOW: Those most likely to die were most likely to get vaccinated. Obviously true, but not interesting - a correlation we need to account for. When you do control for prior risk, the results are as expected, i.e., vaccines make you much safer.
The quoted tweet ("100x") was inaccurate and exaggerated. That's worth briefly correcting. But the lengthy article's points are only minor and pedantic, and do not change what we know, which is that our vaccines are very effective.
Dead Comment
However, 1) it isn't clear if Frieden is working from the same dataset as the author, and 2) a very slight re-wording of the initial statement would make it correct, i.e. "If you are _exposed_ to covid and you've been vaccinated you're about 100 times less likely to die."
The author is actually in the range of an interesting point, even if the article is mostly foolish. If I had a hypothetical vaccine that was in fact composed of neurotoxin and killed the injectee instantly ... it would show 100% effectiveness in the cited tables, because the corpses would be unable to contract COVID. So you can't actually tell if the vaccine reduces your odds of getting sick and dying - we know the current crop are a bit rough and kill people in rare cases.
Covid deaths are so low in the 16-44 category that may be comparable to the vaccine. Although I think the hospitalisation numbers are still quite persuasive in favour of the vaccine.
Dead Comment
Pd1 = Prob(death GIVEN infected AND vaccinated)
vs
Pd2 = Prob(death GIVEN infected AND NOT vaccinated)
Which makes it easy to lose site of:
Pi1 = Prob(infected GIVEN vaccinated)
which is very very small compared to its opposite:
Pi2 = Prob(infected GIVEN NOT vaccinated)
Because Pi1 is so much smaller than Pi2 there are several consequences for Pd1 and Pd2:
- The small counts for "numerator" of Pd1 (zero in some cases) means large uncertainty in the ratio and that the centroid of the distribution is not particularly meaningful. Statistical fluctuations (just one more or one less death) will change Pd1 substantially.
- Statistically significant deviations between Pd1 and Pd2 do not point to a cause.
For example, it could be that those contributing to Pd1 all got a much higher viral load in order for the virus to break through the vaccine's defenses and high viral loads are known to correlate with death so once vax protection is defeated it is game over. The distribution of viral load exposure may even be the same in the Pd2 case, but for the unvaccinated a lesser load can be fatal. This explanation is consistent with P1d possibly being greater than P2d AND consistent with the given anecdotes of vaccinated tweeters saying covid is just the flu, bro.
The main take away is still: you do NOT want to get this shit and vaccine AND masking will help achieve that goal. And, unless you are a sociopath, you do NOT want to give this shit and vaccine AND masking will help achieve that goal.
It's a given that you do not want to contract COVID-19. Vaccination and wearing a mask will help.
You also do not want to give COVID-19 to others. Again, vaccination and wearing a mask will help.
So this article is just pedantry over a statistical quirk.
It's not hard to see how that would be effective anti-persuasion for persons inclined to have doubts about receiving an emergency authorized intervention. Rigorous honesty[1] would assuredly persuade at least some of those people and thus would raise the population vaccination rate, which I'm sure we all agree is a desirable outcome.
[1] For example, rather than chanting the mantra "safe and effective!" being honest about the tradeoffs and showing that proven risk management strategy indicates vaccination is the mathematically optimal choice.
How does that explain delta spread being biggest in highly vaccinated countries?
It makes tons of sense that those who are vaccinated and get infected are more likely to die, simply because it would indicate a failure to develop antibodies and mount an effective immune response to the virus.
That being said I’m not sure such a rant is justified over a tweet like this. Dr. Frieden is more or less correct if you slightly re-word his sentence to: “Getting vaccinated decreases the chance of dying by approximately 100x”.
What I take home from this is that everyone is human, can make simple mistakes, and that we take things written on Twitter far too seriously. I’m sure that Dr. Frieden would not have made such a mistake if he wrote an article that was published somewhere reputable instead of a tweet that could have been sent while waiting for his coffee at Starbucks, or using the toilet.
There is a trade off in communicating science to the public in that using social media, TV, and newspapers reaches more people but almost always distorts the message. I’m not sure how exactly to balance this trade off, but I also think that some amount of responsibility lies with the reader to seek the truth instead of demanding that every sentence they read on the internet be as accurate as a peer reviewed journal.
This kind of mistake is understandable, but doesn't really inspire much confidence in a screed about the mistakes others are making in analyzing COVID-19 data.
Okay… but how much less likely are you to get “infected”?
How are you even defining “infected”, here? If virus gets in your body and then your immune system kills it before it does much damage, were you “infected”? If the answer is no, you've got pretty serious selection bias, because you're completely ignoring everyone who was completely protected by the vaccine in the “vaccinated” group, but paying attention to everyone who didn't really need a vaccine in the “unvaccinated” group: basically ignoring all the really-healthy, great-immune-systems, unlikely-to-die people in the “vaccinated” group so, proportionally, the sicker people take up more room. So you can't even say that the vaccine makes things worse for them! If it improves their chances by 100 times, as Dr. Tom said, but the selection bias is only paying attention to the 0.2% least immune people (those who got infected when exposed after vaccination), you'd expect to see a 5× higher death rate even though the actual death rate is 0.99× lower. (Made-up numbers.)
Obviously most of the people the author is criticizing are using "exposure that would cause detectable disease in non-immune individuals" as their definition, and I don't know if we have a better word for that.
So the author makes the implicit assumption that a successful vaccination doesn't prevent evidence of infection, and uses it to argue that vaccines are dangerous… then accuses everyone else of providing numbers without sufficient context?
Edit: no, actually.
> The reason, hidden in plain sight, is that a large number people who were never going to die, are no longer getting infected.
So the author does know… they were just getting spectacle in before explaining. And I do agree with the author's (eventual) point:
> Without careful control and understanding, one might erroneously conclude the Delta variant is is more lethal if you’ve been vaccinated, the vaccine is losing its efficacy, the vaccination is making people weaker, or some combination. While any of those are possible outcomes in this environment, by not being aware of the infection death rate issue from the start, because one is busy spreading misinformation about extra levels of protection that the data do not support, one misses how to properly control for these effects and analyze new data as it comes in.
But I really don't like the article. The author is the only one making those erroneous conclusions in the first place!
Okay. But immediately after…
> In fact, if there had been 24 deaths in the vaccinated group the efficacy reported would have been 3%! Because it was looking at rates over time, 24 deaths would have been the death rate over time similar to 36 in the unvaccinated group. But clearly, among those infected, 36 in 84611 is a far lower death rate than 24 in 1066!
If? Now we’re onto hypotheticals. Here is an idea, you have so few because they were vaccinated!
Also the mental gymnastics of saying: “it’s better to not be vaccinated in case you get covid because you are less likely to die” is worthy of a mental gymnastics olympics gold medal.
I understand that the CDC guy may have worded things differently but directionally he is right.
But, again, one cannot just ignore the hypotheticals of Dr. Gator when it provides validation for your anti-vaxx stupid attitude.
Also if the vaccine prevents contraction then it should be counted as preventing death/hospitalization, but obviously those numbers aren’t obvious. Seems to me the vaccine IS preventing contraction so you would need to account for that wouldn’t you?
The vaccine works by priming the immune system to work quickly against a disease. It is quite likely that if someone develops symptoms despite their body being at peak alert for coronavirus then they are in trouble. The graphs bear that out, and do not alarm me. If we only select for cases where the person is physically susceptible enough for the virus to break through the vaccine protections to cause an infection it makes sense that the cases will be on average worse.
But there is an interesting point here that if the vaccine were killing people (eg, heart inflammation problems have been detected in some cases) then the "COVID-19-related-death" stats, etc, would be highly misleading. It would be more interesting to see all-cause mortality and the cited tables don't do that.