1. Myself 62 vaccinated, well. First covid infection was mild mid 2023, couple of days. Subsequent fatigue lasted 7 months limiting usual activity such as walking locally and felt 10+ years older lying on the settee a lot. Would have been off work if wasn’t retired. All resolved gradually after 6 months.
2. A friend, 20s vaccinated. Post covid (2021) it took a year before he could run again due to fatigue.
3. Another friend in 60s had other physical post covid problems, but I don’t have permission to comment.
So there is severely debilitating LC and your mathematical stats review is great.
But there is also the less severely affected post covid cohort whose reduced capacity to function is significant to them. Many may not see a doctor (I had basic bloods checked eg for hypothyroidism, all ok) and hopefully most recover in time.
This should not be dismissed, the numbers must be high and the impact upon the economy and their own and other’s well being should be counted I think. I don’t know how that can be done and clearly those with severe illness need the scrutiny and treatment prioritised.
I guess at some point it will be possible and acceptable to count the milder as well as the severe and hopefully there will be less of it by then. Maybe rating scales such as with eg knee pain (does it hurt 1-5 and does it affect daily living activities 1-5 could be used) to assess.
Thanks, and this is all entirely fair (and just because the post-vaccination risk is lower on average, it doesn't mean it's non-zero, and I'm sorry that this happened to you).
My hope is that the ONS survey would at least capture some of the cases like the ones you mention, but I think ultimately all this data is going to be a bit imperfect.
Excellent analysis. My one doubt is that the ONS and Household Pulse Long COVID stats are likely overestimates. Both are, I understand, self-reported data without a control group. Some common symptoms of Long COVID have high baseline incidence and prevalence in the uninfected, particularly fatigue and breathlessness. Lack of control for the baseline would inevitably lead to an overestimate of the effect of Long COVID.
This study, in non-hospitalised and mostly unvaccinated patients pre-Omicron found about 5% incidence of Long COVID-like symptoms in the infected, vs about 4% in a matched cohort of the uninfected: https://www.nature.com/articles/s41591-022-01909-w
That study is subject to its own methodological limitations. In particular, there were likely undetected infections in the control group leading to an over-estimate of the baseline incidence of symptoms. However it does illustrate the need for a well-constructed control group to separate symptoms due to Long COVID from identical symptoms occurring for other reasons.
Thanks - I hadn't seen that, and it is indeed of interest! (I subscribed to Astral Codex Ten for a bit, but I was struggling to keep up with it all, so unsubscribed out of cowardice)
Not sure I agree with you that a numerate graduate would do the calculations you outline and conclude that because they don't 'add up' then they should through the paper away. I hope that they would do what you did and dig behind the numbers. Like you I also tried to find their (likely) sources and wanted to share some of my findings.
As you say the US household survey is almost certainly the source of the '1 in 20 Americans are currently suffering' bullet point. The data is for people self-reporting Long-Covid.
I think the source for the bullet point '10% of Omicron infections led to Long-Covid even with vaccination' comes from the following study which aims to develop a more rigorous way of defining Long-Covid. The data is based on a 'PASC score' derived from a detailed structured questionnaire taken by a sample of infected and non-infected people. The 10% figure is for those infected after Dec 1, 2021 who had Long-Covid 6 months after infection based on the 'PASC score'.
Finally, I struggled to find many papers that showed the long-term (2 year) levels of recovery from Covid. I missed the Lancet paper you found which seems the most likely candidate for the quote. The other papers I found, understandably, focused on those patients who suffered more acute Covid symptoms (e.g. hospitalised) and had a far higher percent still impacted after 2 years than 7%. This would support your contention to focus on the 'long haulers'. See the following paper for the rather worrying details:
Thanks, that makes sense about the NIH paper. I think the problem is that it still suffers from defining "COVID positive" to be symptoms or positive test, so it might still skew towards more severe infections. It feels a lot like the Household Pulse definition itself, which as you probably know reckons only 57% of Americans have had ever COVID (whereas I'd suspect the true figure is a lot closer to 95% say).
And that's interesting about the recovery paper. I'm pretty sure it's that one because I saw Eric Topol circulating it, and he's probably a super-spreader in terms of paper visibility. (And just to be clear, the 7% is the recovery rate - it's 93% not recovered (!) - so even if there's some reporting bias from e.g. people in the cohort getting better and not responding to questionnaires and the true figure is something like 70-80% not recovered then that's still spectacularly high)
Excellent work as ever Oliver. I love the way that you separate out that though it may not be as bad as it was once thought to be, for a proportion it is still truly awful, and they probably deserve more help and attention.
Regards Note 4, a quite minor thought - Are prior cases being tracked to ensure that those no longer reporting (who previously were) are better, rather than no longer able to report at all due to their condition worsening or having not survived? I hope it’s a very small fraction.
Thanks. And it's a good point - I don't know what they do in this case. My reading is that the ONS is a longitudinal survey, so tries to contact the same people every time (though of course some will drop out for one reason or another), whereas it reads to me like Household Pulse pick a different independent sample every time https://www.cdc.gov/nchs/covid19/pulse/long-covid.htm So it's possible you might get slightly different long-term behaviour from those, but I'm not sure if the effect would be big enough to skew the figures much.
I’d like to know more about the people who are suffering from PASC that wouldn’t necessarily be in the LC figures- for instance we know that cardiovascular diseases have increased since 2020- I know it’s hard to quantify but surely these are also important factors to consider when thinking about the overall risks?
I think it's a complicated thing to tease out (because things like longer hospital waiting lists presumably have an effect), but I think the bottom line is that now excess deaths have returned back to something like normal I find it hard to argue that there's anything *too* sinister going on.
There are always anecdotes …bear with me…
1. Myself 62 vaccinated, well. First covid infection was mild mid 2023, couple of days. Subsequent fatigue lasted 7 months limiting usual activity such as walking locally and felt 10+ years older lying on the settee a lot. Would have been off work if wasn’t retired. All resolved gradually after 6 months.
2. A friend, 20s vaccinated. Post covid (2021) it took a year before he could run again due to fatigue.
3. Another friend in 60s had other physical post covid problems, but I don’t have permission to comment.
So there is severely debilitating LC and your mathematical stats review is great.
But there is also the less severely affected post covid cohort whose reduced capacity to function is significant to them. Many may not see a doctor (I had basic bloods checked eg for hypothyroidism, all ok) and hopefully most recover in time.
This should not be dismissed, the numbers must be high and the impact upon the economy and their own and other’s well being should be counted I think. I don’t know how that can be done and clearly those with severe illness need the scrutiny and treatment prioritised.
I guess at some point it will be possible and acceptable to count the milder as well as the severe and hopefully there will be less of it by then. Maybe rating scales such as with eg knee pain (does it hurt 1-5 and does it affect daily living activities 1-5 could be used) to assess.
Not for you to sort out, but I wanted to say…
Thanks, and this is all entirely fair (and just because the post-vaccination risk is lower on average, it doesn't mean it's non-zero, and I'm sorry that this happened to you).
My hope is that the ONS survey would at least capture some of the cases like the ones you mention, but I think ultimately all this data is going to be a bit imperfect.
Excellent analysis. My one doubt is that the ONS and Household Pulse Long COVID stats are likely overestimates. Both are, I understand, self-reported data without a control group. Some common symptoms of Long COVID have high baseline incidence and prevalence in the uninfected, particularly fatigue and breathlessness. Lack of control for the baseline would inevitably lead to an overestimate of the effect of Long COVID.
This study, in non-hospitalised and mostly unvaccinated patients pre-Omicron found about 5% incidence of Long COVID-like symptoms in the infected, vs about 4% in a matched cohort of the uninfected: https://www.nature.com/articles/s41591-022-01909-w
That study is subject to its own methodological limitations. In particular, there were likely undetected infections in the control group leading to an over-estimate of the baseline incidence of symptoms. However it does illustrate the need for a well-constructed control group to separate symptoms due to Long COVID from identical symptoms occurring for other reasons.
The above argument is also made more eloquently and extensively in this article: https://ebm.bmj.com/content/early/2023/08/10/bmjebm-2023-112338
Thanks! You're right of course that controls are important, particularly now the prevalence is getting lower.
It's worth saying that ONS does have controls to some extent though, because it's longitudinal data (I think Household Pulse is just independent random samples?). It basically piggy backs off the ONS prevalence survey, where the same people were being tested every couple of weeks. They do make some sort of attempt to tie in "start of LC" with recent infection events - for example they excluded people who had those symptoms before the most recent infection https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/newonsetselfreportedlongcovidaftercoronaviruscovid19reinfectionintheuk/23february2023#measuring-the-data
Nothing to do with your post sorry (great as usual) just I thought you’d like this and can’t work out how to communicate properly on Substack! https://open.substack.com/pub/astralcodexten/p/some-unintuitive-properties-of-polygenic?r=7p1rx&utm_medium=ios&utm_campaign=post
Thanks - I hadn't seen that, and it is indeed of interest! (I subscribed to Astral Codex Ten for a bit, but I was struggling to keep up with it all, so unsubscribed out of cowardice)
Not sure I agree with you that a numerate graduate would do the calculations you outline and conclude that because they don't 'add up' then they should through the paper away. I hope that they would do what you did and dig behind the numbers. Like you I also tried to find their (likely) sources and wanted to share some of my findings.
As you say the US household survey is almost certainly the source of the '1 in 20 Americans are currently suffering' bullet point. The data is for people self-reporting Long-Covid.
I think the source for the bullet point '10% of Omicron infections led to Long-Covid even with vaccination' comes from the following study which aims to develop a more rigorous way of defining Long-Covid. The data is based on a 'PASC score' derived from a detailed structured questionnaire taken by a sample of infected and non-infected people. The 10% figure is for those infected after Dec 1, 2021 who had Long-Covid 6 months after infection based on the 'PASC score'.
The news release is here:
https://www.nih.gov/news-events/news-releases/large-study-provides-scientists-deeper-insight-into-long-covid-symptoms
.....and the published paper here:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10214179/?report=printable
Finally, I struggled to find many papers that showed the long-term (2 year) levels of recovery from Covid. I missed the Lancet paper you found which seems the most likely candidate for the quote. The other papers I found, understandably, focused on those patients who suffered more acute Covid symptoms (e.g. hospitalised) and had a far higher percent still impacted after 2 years than 7%. This would support your contention to focus on the 'long haulers'. See the following paper for the rather worrying details:
https://www.thelancet.com/journals/lanepe/article/PIIS2666-7762(23)00013-3/fulltext
Thanks, that makes sense about the NIH paper. I think the problem is that it still suffers from defining "COVID positive" to be symptoms or positive test, so it might still skew towards more severe infections. It feels a lot like the Household Pulse definition itself, which as you probably know reckons only 57% of Americans have had ever COVID (whereas I'd suspect the true figure is a lot closer to 95% say).
And that's interesting about the recovery paper. I'm pretty sure it's that one because I saw Eric Topol circulating it, and he's probably a super-spreader in terms of paper visibility. (And just to be clear, the 7% is the recovery rate - it's 93% not recovered (!) - so even if there's some reporting bias from e.g. people in the cohort getting better and not responding to questionnaires and the true figure is something like 70-80% not recovered then that's still spectacularly high)
Excellent work as ever Oliver. I love the way that you separate out that though it may not be as bad as it was once thought to be, for a proportion it is still truly awful, and they probably deserve more help and attention.
Regards Note 4, a quite minor thought - Are prior cases being tracked to ensure that those no longer reporting (who previously were) are better, rather than no longer able to report at all due to their condition worsening or having not survived? I hope it’s a very small fraction.
Thanks. And it's a good point - I don't know what they do in this case. My reading is that the ONS is a longitudinal survey, so tries to contact the same people every time (though of course some will drop out for one reason or another), whereas it reads to me like Household Pulse pick a different independent sample every time https://www.cdc.gov/nchs/covid19/pulse/long-covid.htm So it's possible you might get slightly different long-term behaviour from those, but I'm not sure if the effect would be big enough to skew the figures much.
I’d like to know more about the people who are suffering from PASC that wouldn’t necessarily be in the LC figures- for instance we know that cardiovascular diseases have increased since 2020- I know it’s hard to quantify but surely these are also important factors to consider when thinking about the overall risks?
I think it's a complicated thing to tease out (because things like longer hospital waiting lists presumably have an effect), but I think the bottom line is that now excess deaths have returned back to something like normal I find it hard to argue that there's anything *too* sinister going on.