If you want to see how an 'any symptom' present or not comparisons between positive cases vs non-positive cases massively underestimates long COVID prevalence- just look at the recent ONS study that some paediatricians are astonishingly hailing as 'reassuring'. Short thread. 
If you look at each symptom loss of smell/taste is *16%* in those who tested positive at 12 wks vs almost no one in the 'control' group which didn't have a positive test. So this would suggest that the prevalence of long COVID should be at least 16% of those infected, right?
After all if *one* symptom is present in 16% of those infected and almost none of those who didn't test positive, surely this is the *minimum* prevalence of the syndrome, right? When you consider more symptoms, this prevalence should increase not decrease.
So what happens when we do any symptom 'case-control' comparisons. The difference gets diluted - and now the difference is 8%! But how is that possible? Because we know that 16% of those infected had loss of smell/taste for 12 wks or more, when none of 'controls' did!
It's because inclusion of all these symptoms, and using an 'any symptom' approach actually substantially underestimates prevalence by diluting the signal. It's just basic math- you can read my thread about this here:
https://twitter.com/dgurdasani1/status/1487578265187405828?s=20&t=dPOrcSYtFA3uE-rNkzGrZw
I hope @ShamezLadhani and other commentators will take note.
Link to full analysis here:
https://twitter.com/dgurdasani1/status/1498393581366747143?s=20&t=AZCq9pIlKulJusbxpN5z_A
• • •
Missing some Tweet in this thread? You can try to
force a refresh
