Thank you for writing these Zvi. I read them, and I appreciate them.
You sort of “threw up your hands in confusion” at one point, and so I thought it might be worthwhile to respond with a note about how I’ve been purposefully trying to avoid having detailed models on the domain where the confusion might exist.
Specifically, I have long held off on worrying too much about the details of the medical biology of covid, out of humility or whatever.
Like: there are a zillion possible treatments, and different covid cases might evolve differently, and I’d tend to defer to an experienced clinician in their own clinic, if that clinic has good outcome numbers.
My model of local reactive skill and knowledge development predicts that two different clinics with two different genius clinicians in them could both learn to treat covid very effectively… and yet also in wildly different ways over the course of 6-18 months of a high volume of patients on which to “do their best, with variations, while observing carefully, and updating thoughtfully”. So: universality properties are helpful for noticing good and bad reasoning, and universality expectations are sparse here.
Not only do I expect “valid local practical empirical knowledge” to have heterogeneity, but also it just seems like covid has heterogeneous effects on humans.
(Also, I think maybe early studies shows that if you sample different compartments of a covid infected human, you find different genomes, so there might be some within-patient evolutionary effects? I haven’t heard much about this since like February of 2020 when only virology geeks and weirdo preppers were talking much about what I then tended to call “wuflu” in text messages, in private, to save on typing.)
Some people get anosmia, some don’t? Wat!?
It seems sorta weird to me how variable things are, and the thing I’d hope for is like: mechanistic variety to explain outcome variety.
It would be useful, perhaps, to do a GWAS on covid outcomes, and with the large N that has accumulated this might be possible?
So DESPITE all these predictable epistemic barriers to knowing about “covid x human interactions”...
...I’m coming around to thinking that it would be useful to have a new clean mechanistic model of “how a typical fatal covid case progresses, day by day, with milestones, and typical inter-milestone amounts of time”.
This seems to not yet be something that exists as part of common knowledge within the anglophone internet.
For example when I google [covid progression wikipedia] I have to click around to get to a not-very-good page about Symptoms of COVID-19 (which (right now) is more like a list of words than a mechanistic story about the handful of paths along which metaphorical “causal dominoes” can fall over in the way that dominoes could predictably fall over in general).
I’m not the first person to think “hey maybe detailed study of covid in the body of humans is useful to model” obviously.
Doctors have to think about this because this is their main job now. Hence… they’ve already DONE GWAS studies, and even started doing meta-analyses of covid GWAS studies.
All of this is to say that it might now, in 2022 (heading into 2023) be “too-wordcel-esque” at this point to fall back to mere “with/from” mortality distinctions?
My new theory is that these are mainly ‘with Covid’ deaths rather than ‘from Covid’ deaths and the definition somehow includes anyone with recent Covid, so people who have actually fully recovered, who then died of something else, are being counted as Covid deaths. I haven’t heard that expressed elsewhere, but if that’s not true, then I got nothing.
Like: at this point… I kind of expect to get covid several times over the rest of my life, and I think that I’ll probably have lowered life expectancy even if just from from the boring normal scar tissue that forms in quite a few human bodies from from such infections, sometimes with the loss of an IQ point or 5, maybe some pulmonary fibrosis towards the end?
(A separate question I have genuine empirical uncertainty about: maybe some people have versions of covid that linger in reservoirs, as with herpes, HIV, and hepatitis that does damage over time? I don’t know.)
If I get covid a couple times, and my body is damaged, and then my body gives up earlier than otherwise, and someone tries to placate me by saying “Jennifer, you’re not dying from covid, you’re just dying with or after covid” then I might uh.. spit at them? …and call them idiots? …and hope there’s an afterlife so I can “go all Karen” on The Demiurge, and ask why it is has this macabre fetish for imposing body horror on embodied souls.
As a practical upshot, I would love it if there was some body-system by body-system explanation of the progression of the virus.
What do the autopsies look like?
Does covid work its way down throats into lungs slowly and creepingly, or does viral shedding from the upper respiratory area fall to the bottom of the lungs, eventually “take”, and then climb back up and out, eating lung tissue as it goes? (Probably neither, but I don’t know. Does anyone know? If so can you please edit wikipedia with the knowledge?)
In a world with a competently benevolent health system, I think every death would come with a little Pearlian Causal Graph that includes some body systems, some medical choices, and some larger humanistic context.
Here are some possible things that one could read off of a (hypothetical (utopian?)) single person’s Official Causal Death Analysis in light of a larger graph theoretic causal reasoning framework (translated into english, with some trimming of the fuzzy leaf nodes):
(A) “ALICE was killed by lack of superfast medical care plus bleeding out, caused by lack of a real-time always-on heart monitor that can call EMH, plus a rare unskilled knife attack, whose wielding was caused by murderous intent, which was caused by discovery of marital infidelity”,
(B) “BOB was killed by congestive heart failure, caused by catching flu and having complications, during ‘normal’ flu season (caused by lack of communicable disease filtering at various borders), while deep into ‘early’ senescence, possibly due to a rare allele of CUX1 (...AKA a way to ‘die of old age’ while relatively young)”,
(C) and “CAROL was killed by traumatic brain damage, caused by a gun, caused by suicidal intent, maybe caused by bullying at a school, caused by teenagers in an authoritarian institution with low character, low morale, and inadequate leadership”,
(D) and so on… (they are kind of fun to write, if you subtract out the sadness intrinsic to the writing exercise, with each one sort of like the premise of a tiny little murder mystery story).
In a framework like this, in 2022, covid would show up. If covid doesn’t show up, you’re doing it wrong. Not as the first step, and not as the last step, but “in the story”.
Maybe, empirically, it would turn out that the data would show that people are mostly only causally “dying from covid” and never causally “dying with covid”… but I currently doubt that this is how the numbers would turn out, if such numbers were collected in a way that made it possible to tell these hypotheses apart, and then tabulated over large N, so that humans could simply know the answer.
I guess: it could be possible that someone on Earth already knows the answer to the practical empirical question here (even if they wouldn’t necessarily mention causal graphs)? But I haven’t found a good writeup yet that makes the knowledge easy to spread.
Thanks for this thoughtful analysis. Concrete example: there have been house fires in which loss of smell from Covid was a contributing factor.
Also people who are “hard of smelling” often develop cardiac issues because they tend to over-salt their food. My aunt has struggled with this for several decades now after a chemical accident, and recently had a heart attack.
Thank you for writing these Zvi. I read them, and I appreciate them.
You sort of “threw up your hands in confusion” at one point, and so I thought it might be worthwhile to respond with a note about how I’ve been purposefully trying to avoid having detailed models on the domain where the confusion might exist.
Specifically, I have long held off on worrying too much about the details of the medical biology of covid, out of humility or whatever.
Like: there are a zillion possible treatments, and different covid cases might evolve differently, and I’d tend to defer to an experienced clinician in their own clinic, if that clinic has good outcome numbers.
My model of local reactive skill and knowledge development predicts that two different clinics with two different genius clinicians in them could both learn to treat covid very effectively… and yet also in wildly different ways over the course of 6-18 months of a high volume of patients on which to “do their best, with variations, while observing carefully, and updating thoughtfully”. So: universality properties are helpful for noticing good and bad reasoning, and universality expectations are sparse here.
Not only do I expect “valid local practical empirical knowledge” to have heterogeneity, but also it just seems like covid has heterogeneous effects on humans.
(Also, I think maybe early studies shows that if you sample different compartments of a covid infected human, you find different genomes, so there might be some within-patient evolutionary effects? I haven’t heard much about this since like February of 2020 when only virology geeks and weirdo preppers were talking much about what I then tended to call “wuflu” in text messages, in private, to save on typing.)
Some people get anosmia, some don’t? Wat!?
It seems sorta weird to me how variable things are, and the thing I’d hope for is like: mechanistic variety to explain outcome variety.
It would be useful, perhaps, to do a GWAS on covid outcomes, and with the large N that has accumulated this might be possible?
So DESPITE all these predictable epistemic barriers to knowing about “covid x human interactions”...
...I’m coming around to thinking that it would be useful to have a new clean mechanistic model of “how a typical fatal covid case progresses, day by day, with milestones, and typical inter-milestone amounts of time”.
This seems to not yet be something that exists as part of common knowledge within the anglophone internet.
For example when I google [covid progression wikipedia] I have to click around to get to a not-very-good page about Symptoms of COVID-19 (which (right now) is more like a list of words than a mechanistic story about the handful of paths along which metaphorical “causal dominoes” can fall over in the way that dominoes could predictably fall over in general).
I’m not the first person to think “hey maybe detailed study of covid in the body of humans is useful to model” obviously.
Doctors have to think about this because this is their main job now. Hence… they’ve already DONE GWAS studies, and even started doing meta-analyses of covid GWAS studies.
All of this is to say that it might now, in 2022 (heading into 2023) be “too-wordcel-esque” at this point to fall back to mere “with/from” mortality distinctions?
Like: at this point… I kind of expect to get covid several times over the rest of my life, and I think that I’ll probably have lowered life expectancy even if just from from the boring normal scar tissue that forms in quite a few human bodies from from such infections, sometimes with the loss of an IQ point or 5, maybe some pulmonary fibrosis towards the end?
(A separate question I have genuine empirical uncertainty about: maybe some people have versions of covid that linger in reservoirs, as with herpes, HIV, and hepatitis that does damage over time? I don’t know.)
If I get covid a couple times, and my body is damaged, and then my body gives up earlier than otherwise, and someone tries to placate me by saying “Jennifer, you’re not dying from covid, you’re just dying with or after covid” then I might uh.. spit at them? …and call them idiots? …and hope there’s an afterlife so I can “go all Karen” on The Demiurge, and ask why it is has this macabre fetish for imposing body horror on embodied souls.
As a practical upshot, I would love it if there was some body-system by body-system explanation of the progression of the virus.
What do the autopsies look like?
Does covid work its way down throats into lungs slowly and creepingly, or does viral shedding from the upper respiratory area fall to the bottom of the lungs, eventually “take”, and then climb back up and out, eating lung tissue as it goes? (Probably neither, but I don’t know. Does anyone know? If so can you please edit wikipedia with the knowledge?)
In a world with a competently benevolent health system, I think every death would come with a little Pearlian Causal Graph that includes some body systems, some medical choices, and some larger humanistic context.
Here are some possible things that one could read off of a (hypothetical (utopian?)) single person’s Official Causal Death Analysis in light of a larger graph theoretic causal reasoning framework (translated into english, with some trimming of the fuzzy leaf nodes):
(A) “ALICE was killed by lack of superfast medical care plus bleeding out, caused by lack of a real-time always-on heart monitor that can call EMH, plus a rare unskilled knife attack, whose wielding was caused by murderous intent, which was caused by discovery of marital infidelity”,
(B) “BOB was killed by congestive heart failure, caused by catching flu and having complications, during ‘normal’ flu season (caused by lack of communicable disease filtering at various borders), while deep into ‘early’ senescence, possibly due to a rare allele of CUX1 (...AKA a way to ‘die of old age’ while relatively young)”,
(C) and “CAROL was killed by traumatic brain damage, caused by a gun, caused by suicidal intent, maybe caused by bullying at a school, caused by teenagers in an authoritarian institution with low character, low morale, and inadequate leadership”,
(D) and so on… (they are kind of fun to write, if you subtract out the sadness intrinsic to the writing exercise, with each one sort of like the premise of a tiny little murder mystery story).
In a framework like this, in 2022, covid would show up. If covid doesn’t show up, you’re doing it wrong. Not as the first step, and not as the last step, but “in the story”.
Maybe, empirically, it would turn out that the data would show that people are mostly only causally “dying from covid” and never causally “dying with covid”… but I currently doubt that this is how the numbers would turn out, if such numbers were collected in a way that made it possible to tell these hypotheses apart, and then tabulated over large N, so that humans could simply know the answer.
I guess: it could be possible that someone on Earth already knows the answer to the practical empirical question here (even if they wouldn’t necessarily mention causal graphs)? But I haven’t found a good writeup yet that makes the knowledge easy to spread.
Thanks for this thoughtful analysis. Concrete example: there have been house fires in which loss of smell from Covid was a contributing factor.
Also people who are “hard of smelling” often develop cardiac issues because they tend to over-salt their food. My aunt has struggled with this for several decades now after a chemical accident, and recently had a heart attack.
Fascinating. There’s a whole literature on sense/nutrition interactions.