I’ve considered checking samples of public figures, but dismissed it because it’s really hard to get a good sense of who has Long Covid:
not everyone knows they have Long Covid
people don’t like to say they have it. I expect this to be especially the case for professional athletes whose career depends on it
I’m not sure how much top performance is affected in mild cases. I think in early stages it’s possible to push through a lot. The main symptom is fatigue post exertion. We would still expect to observe reduced performance though, but that’s harder to observe.
Retirement is a drastic decision and people would generally postpone that.
Due to these issues, I feel like disability data is a much more reliable sense check, and I think it fits the ONS UK numbers.
The article you link to isn’t good, because they probably had a lot of Long Covid cases in their control group. They used antibodies as sole diagnostic criterion of prior infection. But about 1 in 3 people do not create detectable amounts of antibodies (https://wwwnc.cdc.gov/eid/article/27/9/21-1042_article), antibodies fade over time, and there’s some rumors that Long Covid patients are now likely to not have antibodies but I haven’t checked that. The fact that there’s no easily accessible diagnostic tool makes it hard for all these prevalence studies.
On the study, even if the antibody test isn’t that accurate, one would still expect people with confirmed covid to have more long covid symptoms than people without confirmed covid. In fact, the study finds that belief in having had covid is a stronger predictor than confirmed covid, which suggests that the symptoms are caused by something else.
But people who have had COVID do have more Long Covid, if actually use an accurate measure (PCR testing). I report multiple studies in the post with control groups.
In the study, people with positive serology HAD more of 10 specific symptoms, even though serology is very inaccurate. Only when controlled for belief did that disappear. But belief in having had COVID has strong confounding effects:
if you have lasting effects, of course you’re more likely to identify a prior infection
if you had more clear acute symptoms, you’re more likely to have both belief you’ve had COVID, as well as that you’re more likely to develop Long Covid
they say that belief and serology were not correlated, but I’m confused by that. In the belief+ group, half had positive serology. In the belief- group, it’s like 2%?
If you control a weak predictor by a strong predictor correlated with the weak predictor, I’m not surprised that significant effects disappear.
The study also had data on PCR testing but didn’t use that in any way, which seems suspicious to me.
Also, in 2020 the base rate for other communicable diseases dropped a lot (flu dropped by factor 50x)
The problem with PCR test controls is that they would only catch an infection around the time you get infected, whereas antibody tests would catch infections further back in time.
I don’t see the evidence that belief in having had covid is a better predictor of having had covid than is a serology test.
On the economist article, the study didn’t find a significant drop, it found a reduction in minutes played of 2 minutes per game and a reduction in passes of 3 per 90 minutes 225 days post-covid. Although zero effect is outside of the confidence interval for the passes metric (but not minutes played) according to the study, the effect is so small, and the measure so noisy, that in my view it is almost certainly a statistical artefact.
Hi John,
I’ve considered checking samples of public figures, but dismissed it because it’s really hard to get a good sense of who has Long Covid:
not everyone knows they have Long Covid
people don’t like to say they have it. I expect this to be especially the case for professional athletes whose career depends on it
I’m not sure how much top performance is affected in mild cases. I think in early stages it’s possible to push through a lot. The main symptom is fatigue post exertion. We would still expect to observe reduced performance though, but that’s harder to observe.
Retirement is a drastic decision and people would generally postpone that.
Due to these issues, I feel like disability data is a much more reliable sense check, and I think it fits the ONS UK numbers.
The article you link to isn’t good, because they probably had a lot of Long Covid cases in their control group. They used antibodies as sole diagnostic criterion of prior infection. But about 1 in 3 people do not create detectable amounts of antibodies (https://wwwnc.cdc.gov/eid/article/27/9/21-1042_article), antibodies fade over time, and there’s some rumors that Long Covid patients are now likely to not have antibodies but I haven’t checked that. The fact that there’s no easily accessible diagnostic tool makes it hard for all these prevalence studies.
On the study, even if the antibody test isn’t that accurate, one would still expect people with confirmed covid to have more long covid symptoms than people without confirmed covid. In fact, the study finds that belief in having had covid is a stronger predictor than confirmed covid, which suggests that the symptoms are caused by something else.
But people who have had COVID do have more Long Covid, if actually use an accurate measure (PCR testing). I report multiple studies in the post with control groups.
In the study, people with positive serology HAD more of 10 specific symptoms, even though serology is very inaccurate. Only when controlled for belief did that disappear. But belief in having had COVID has strong confounding effects:
if you have lasting effects, of course you’re more likely to identify a prior infection
if you had more clear acute symptoms, you’re more likely to have both belief you’ve had COVID, as well as that you’re more likely to develop Long Covid
they say that belief and serology were not correlated, but I’m confused by that. In the belief+ group, half had positive serology. In the belief- group, it’s like 2%?
If you control a weak predictor by a strong predictor correlated with the weak predictor, I’m not surprised that significant effects disappear.
The study also had data on PCR testing but didn’t use that in any way, which seems suspicious to me.
Also, in 2020 the base rate for other communicable diseases dropped a lot (flu dropped by factor 50x)
The problem with PCR test controls is that they would only catch an infection around the time you get infected, whereas antibody tests would catch infections further back in time.
I don’t see the evidence that belief in having had covid is a better predictor of having had covid than is a serology test.
And here’s an Economist article analysing footballer performance after COVID infection: https://archive.ph/qGWKs
Average performance measures definitely dropped significantly long-term. But it doesn’t have data on all-out disability.
And this article lists a few names, but also mentions what you write: that surprisingly few athletes had Long Covid at the time of writing: https://www.washingtonpost.com/sports/2021/04/19/athletes-long-haul-covid-justin-foster/
On the economist article, the study didn’t find a significant drop, it found a reduction in minutes played of 2 minutes per game and a reduction in passes of 3 per 90 minutes 225 days post-covid. Although zero effect is outside of the confidence interval for the passes metric (but not minutes played) according to the study, the effect is so small, and the measure so noisy, that in my view it is almost certainly a statistical artefact.
Fair enough re: significance and effect size. I don’t think it’s an artefact though
Regarding public samples, I had been thinking of a political body like a parliament, but as this Senator with Long Covid says: many people are not public about their disability. https://twitter.com/wsbgnl/status/1505814009722798081?s=20&t=iLBZn1qk_BUJQcJx8kIQEg
(Not clear from the quote whether he refers to other senators, or colleagues in different positions)