Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Attributes and predictors of long Covid

nature.com

Post reply on HN
Print view View original

11–20 of 77 posts

Re: Attributes and predictors of long Covid

#11

Post by greenwich26 »

brb92 wrote: post #10

> For the reduced logistic regression model, the score was given by the following formula: S = 0.259503 × NumberSymptoms + 0.055457 × age − 0.633310 × sex − 3.20 (where sex is encoded as 1 − female/2 − male) Where ‘NumberSymptoms’ corresponds to the sum of different symptoms experienced over the first week among the list of 14 symptoms reported on daily logs. This score was then transformed to a probability using the…

Not an epidemiologist so don't weigh my opinion too highly, but as a data analyst, my first impression of this "model" is overfitted nonsense.

Re: Attributes and predictors of long Covid

#12

Post by vanderZwan »

greenwich26 wrote: post #9
vanderZwan wrote: post #7

Something I've noticed: when people in my surroundings bring up worries of long-term effects of the fast-tracked vaccines, they somehow tend to not consider the unknown long-term effects of covid. Even though the argument that we cannot know the long term effects yet should apply equally. When I bring this up they do seem receptive to this argument though, provided I'm not dismissive of their own skepticism. I'm not…

You can't neglect the prior. In most European countries, somewhere around 5% of the population might have been infected with covid. Up to around 10% in the US. Therefore, to have comparable posterior probabilities for long-term side effects, you would need the side effects argument not to merely "apply equally", but to favour the vaccine over the disease by a factor of about 10.

You're not wrong, but my priority is to ensure that my friends who rely on emotion-based arguments (fear of the unknown regarding the vaccines, dismissing the risks of covid and taking too few precautions as a result) to be safe and to not contribute to keeping the pandemic going. Going this deep into statistics would just lose them, because I'd come across as trying to intimidate them into taking my viewpoint on how to reason about these things.

edit: apparently none of you have ever been accused of being an ivory tower jerk in their lives, explicitly or implicitly, ever? We're not talking about a discussion between people trying to weigh uncertainties based on what they do and do not know and working it out on a paper napkin. We're talking about people being overwhelmed by the complexity of it all and as a result starting from what they're most scared of and reasoning backwards from there. Which is how most people out there function, like it or not.

Re: Attributes and predictors of long Covid

#13

Post by learnstats2 »

greenwich26 wrote: post #11
brb92 wrote: post #10

> For the reduced logistic regression model, the score was given by the following formula: S = 0.259503 × NumberSymptoms + 0.055457 × age − 0.633310 × sex − 3.20 (where sex is encoded as 1 − female/2 − male) Where ‘NumberSymptoms’ corresponds to the sum of different symptoms experienced over the first week among the list of 14 symptoms reported on daily logs. This score was then transformed to a probability using the…

Not an epidemiologist so don't weigh my opinion too highly, but as a data analyst, my first impression of this "model" is overfitted nonsense.

I checked the paper to try and defend it, but I was unable to find a table with significance testing or standard errors for this regression - it is just stated without justification.

This seems an unusual style of reporting results, to me. (Happy to be corrected)

Re: Attributes and predictors of long Covid

#14

Post by icegreentea2 »

brb92 wrote: post #10

> For the reduced logistic regression model, the score was given by the following formula: S = 0.259503 × NumberSymptoms + 0.055457 × age − 0.633310 × sex − 3.20 (where sex is encoded as 1 − female/2 − male) Where ‘NumberSymptoms’ corresponds to the sum of different symptoms experienced over the first week among the list of 14 symptoms reported on daily logs. This score was then transformed to a probability using the…

Gender differences wouldn't be unheard of. Also, in that model, being male reduces your predicted probably of long COVID.

This model does seem kinda wacky though. Just throwing in 1 symptom, age of 30, and female gives a probability of 13% which seems high. I'll have to read the paper more carefully to figure out what exactly that probability is supposed to mean.

Re: Attributes and predictors of long Covid

#16

Post by adolph »

vanderZwan wrote: post #12
greenwich26 wrote: post #9

Earlier quoted context omitted.

You can't neglect the prior. In most European countries, somewhere around 5% of the population might have been infected with covid. Up to around 10% in the US. Therefore, to have comparable posterior probabilities for long-term side effects, you would need the side effects argument not to merely "apply equally", but to favour the vaccine over the disease by a factor of about 10.

You're not wrong, but my priority is to ensure that my friends who rely on emotion-based arguments (fear of the unknown regarding the vaccines, dismissing the risks of covid and taking too few precautions as a result) to be safe and to not contribute to keeping the pandemic going. Going this deep into statistics would just lose them, because I'd come across as trying to intimidate them into taking my viewpoint on how…

>my friends who rely on emotion-based arguments

Isn’t every argument based in emotion? Do unpalatable facts have a strong track record of success in arguments?

Re: Attributes and predictors of long Covid

#17

Post by Retric »

adolph wrote: post #16
vanderZwan wrote: post #12

Earlier quoted context omitted.

You're not wrong, but my priority is to ensure that my friends who rely on emotion-based arguments (fear of the unknown regarding the vaccines, dismissing the risks of covid and taking too few precautions as a result) to be safe and to not contribute to keeping the pandemic going. Going this deep into statistics would just lose them, because I'd come across as trying to intimidate them into taking my viewpoint on how…

>my friends who rely on emotion-based arguments Isn’t every argument based in emotion? Do unpalatable facts have a strong track record of success in arguments?

No. A lot of modern physics for example still seems objectively wrong when hearing about it as a student. Physical experiments and math has converted a lot of skeptics into believing some very “strange” things about how our universe works.

Re: Attributes and predictors of long Covid

#18

Post by smhg »

wvh wrote: post #3
lofi_lory wrote: post #2

> Long COVID was characterized by symptoms of fatigue, headache, dyspnea and anosmia and was more likely with increasing age and body mass index and female sex. Experiencing more than five symptoms during the first week of illness was associated with long COVID. A bit disappointing information. So, first week severity indicates risk for long covid? Or is it multiorgan involvement independent from severity?

It's surprising how long it takes to draw solid conclusions, considering the amount of "live" cases to draw from.

Indeed, also taking into consideration that the pandemic is the most pressing issue in many countries. It can't be a budgetary issue. Could someone shed some light on why these questions are so hard?

Re: Attributes and predictors of long Covid

#19

Post by cortic »

vanderZwan wrote: post #7

Something I've noticed: when people in my surroundings bring up worries of long-term effects of the fast-tracked vaccines, they somehow tend to not consider the unknown long-term effects of covid. Even though the argument that we cannot know the long term effects yet should apply equally. When I bring this up they do seem receptive to this argument though, provided I'm not dismissive of their own skepticism. I'm not…

That the vaccine is better than the disease seems a ridiculously low bar to me. I'd be much more reassured if they listed the side effects with their likelihoods (1 in ??) subdivided this with age groups if appropriate, I've seen this in NHS prescriptions in the past and its always been very reassuring.

Re: Attributes and predictors of long Covid

#20

Post by IfOnlyYouKnew »

greenwich26 wrote: post #9
vanderZwan wrote: post #7

Something I've noticed: when people in my surroundings bring up worries of long-term effects of the fast-tracked vaccines, they somehow tend to not consider the unknown long-term effects of covid. Even though the argument that we cannot know the long term effects yet should apply equally. When I bring this up they do seem receptive to this argument though, provided I'm not dismissive of their own skepticism. I'm not…

You can't neglect the prior. In most European countries, somewhere around 5% of the population might have been infected with covid. Up to around 10% in the US. Therefore, to have comparable posterior probabilities for long-term side effects, you would need the side effects argument not to merely "apply equally", but to favour the vaccine over the disease by a factor of about 10.

Without vaccination, the chance of eventually catching the disease is just about 100%. Yes, any single person may try and hope to be protected by herd immunity. But there are serious doubts that can be achieved given some variants' higher virulence and expected levels of eventual vaccine uptake considering justifiable reasons.

Meanwhile, long-term harms from vaccination are entirely speculative. There isn't even a real theory as to any mechanism to cause any such effects: all components of the vaccines are well-known, except the COVID-specific parts which are a strict subset of the actual virus.

Post reply on HN