They say they haven’t accounted for sampling bias, though, which makes me doubt the methodology overall, as sampling bias could be huge over 90 day timespans.
Yes, the article doesn’t describe the exact methodology, but they could be well deriving the percentages from people who choose to self-report how they’re doing after 30 and 90 days. These would be far more likely to be people who still feel unwell.
As a separate point, and I’m skirting around using the word “hypochondria” here, asking people is they still feel unwell or have symptoms a month or three after first contracting covid is going to get some fairly subjective answers. All in all I don’t think this particular study tells us much about the likelihood of covid causing permanent damage.
What about as an upper bound? I’m having a harder time generating confounders that make this an underestimate.
Yes, it seems more reasonable to treat it as evidence of upper bound. Still weak evidence IMO, due to the self-reporting of perceived symptoms.