I haven’t looked into any papers in detail. I’ve read multiple EAs commenting on various papers and there seems to be ambiguity on how to interpret them. For instance, I remember someone (Katja Grace?) pointing out how the following points are statistically compatible:
Long Covid affects a significant portion of people
Among people who thought they had Covid because they recently started suffering from a mysterious chronic illness, seropositivity is lower than the baseline
That’s because there are indeed other causes of chronic illness that could be driving the belief “I probably had Covid.”
Another option is that Long Covid messes with seropositivity – I think I saw that hypothesized as well. It might be a somewhat conspiratorial hypothesis, but “Long Covid is psychosomatic” sounds just as conspiratorial. We know that post-viral syndromes exist and we know that SARS-1 led to extreme long-term issues (though this is highly confounded by disease severity).
I also don’t think it would help much if the mechanism was psychosomatic. That doesn’t change the problem that people cannot work anymore. (I guess you’re thinking that if you believe it isn’t a problem, you can make yourself immune? That may not work if you’re only 75% confident in the hypothesis. Illnesses can lower the clarity of one’s thinking and make you more anxious.)
In any case, I think there’s so much anecdotal data with Long Covid (even just in EA-related or personal/family circles) that anecdotes are useful here, so I’ve built up a picture that feels more robust than heavily contested studies based on self-reports. Of course, if all the studies said Long Covid wasn’t a thing, I’d believe them. It’s just that combined with priors I got from anecdotes, my interpretation of various studies is likely to be somewhere in the middle, rather than at the extremes.
That said, I also think we can probably rule out Long Covid rates (for >3 months out) >0.8%. In particular, I think we have some evidence from reports about athletes getting sick early on in the pandemic, and very few reports of “famous athlete can no longer play soccer because of Long Covid.” (It’s more common in women, though, and I mostly saw coverage of male athletes at the time.) (Edit: I googled soccer players with Long Covid and found quite a few results as well as statements saying very few professional players have even caught it, so I changed my estimate again from >0.5% to >0.8%.)
Edit to add: You estimated the expected health costs of Long Covid at about 1 week in your original comment. I think my view isn’t too far away from that (maybe I’d say it’s twice that). I was mostly reacting to “I doubt it’s even real” part of your comment.
Another approach: According to the Zoe Covid study, 280,000 people in the UK got Covid today. That’s 0.4% of the UK. So if that’s the baseline for one day, we get a 1% baseline for the 2.5 days of the conference.
I assume that EAs take more tests than the average person (esp. now that tests have scaled back) but that attending a conference is substantially more risky than the average behavior. I probably expect there to be more than 1% infections during EAG for that reason, but maybe not a lot more? With the 1% prior/anchoring, I’m thinking >5% is a bit too high, so I now put less than 25% probability mass on that. But my median is well above 1%.
I haven’t looked into any papers in detail. I’ve read multiple EAs commenting on various papers and there seems to be ambiguity on how to interpret them. For instance, I remember someone (Katja Grace?) pointing out how the following points are statistically compatible:
Long Covid affects a significant portion of people
Among people who thought they had Covid because they recently started suffering from a mysterious chronic illness, seropositivity is lower than the baseline
That’s because there are indeed other causes of chronic illness that could be driving the belief “I probably had Covid.”
Another option is that Long Covid messes with seropositivity – I think I saw that hypothesized as well. It might be a somewhat conspiratorial hypothesis, but “Long Covid is psychosomatic” sounds just as conspiratorial. We know that post-viral syndromes exist and we know that SARS-1 led to extreme long-term issues (though this is highly confounded by disease severity).
I also don’t think it would help much if the mechanism was psychosomatic. That doesn’t change the problem that people cannot work anymore. (I guess you’re thinking that if you believe it isn’t a problem, you can make yourself immune? That may not work if you’re only 75% confident in the hypothesis. Illnesses can lower the clarity of one’s thinking and make you more anxious.)
In any case, I think there’s so much anecdotal data with Long Covid (even just in EA-related or personal/family circles) that anecdotes are useful here, so I’ve built up a picture that feels more robust than heavily contested studies based on self-reports. Of course, if all the studies said Long Covid wasn’t a thing, I’d believe them. It’s just that combined with priors I got from anecdotes, my interpretation of various studies is likely to be somewhere in the middle, rather than at the extremes.
That said, I also think we can probably rule out Long Covid rates (for >3 months out) >0.8%. In particular, I think we have some evidence from reports about athletes getting sick early on in the pandemic, and very few reports of “famous athlete can no longer play soccer because of Long Covid.” (It’s more common in women, though, and I mostly saw coverage of male athletes at the time.) (Edit: I googled soccer players with Long Covid and found quite a few results as well as statements saying very few professional players have even caught it, so I changed my estimate again from >0.5% to >0.8%.)
Edit to add: You estimated the expected health costs of Long Covid at about 1 week in your original comment. I think my view isn’t too far away from that (maybe I’d say it’s twice that). I was mostly reacting to “I doubt it’s even real” part of your comment.
Another approach: According to the Zoe Covid study, 280,000 people in the UK got Covid today. That’s 0.4% of the UK. So if that’s the baseline for one day, we get a 1% baseline for the 2.5 days of the conference.
I assume that EAs take more tests than the average person (esp. now that tests have scaled back) but that attending a conference is substantially more risky than the average behavior. I probably expect there to be more than 1% infections during EAG for that reason, but maybe not a lot more? With the 1% prior/anchoring, I’m thinking >5% is a bit too high, so I now put less than 25% probability mass on that. But my median is well above 1%.