In a post on long Covid back in July, I said that “estimates of the chance of reporting symptoms after 12 weeks range from less than 1% to almost 12%”. That 12% figure came from the ONS, who found that individuals who tested positive were 12 percentage points more likely than controls to report at least one symptom 12 weeks after infection.
In my post, I argued that 12% is probably an overestimate on the grounds that some people who tested positive might have been inclined to exaggerated their symptoms – to report things they normally wouldn’t have done (thanks to all the media attention on long Covid).
And I noted that a study published in Nature Medicine had observed a much smaller percentage of people still reporting symptoms 12 weeks after infection, namely 2.3%.
A new analysis by the ONS has obtained a figure almost identical to that observed in the Nature Medicine study, namely 2.5% (the difference between the blue and green lines in the chart below). This is clearly much lower than its previous estimate.
Interestingly, the reason for the discrepancy with the earlier figure isn’t the one I suggested (i.e., that some people who tested positive were inclined to exaggerate their symptoms). Rather, it’s a statistical issue.
In both their original and updated analyses, the ONS defined symptom discontinuation as two consecutive visits without reporting any symptoms. (Participants in the ONS’s survey were visited at regular intervals for the purpose of data collection.)
This means that someone would be classified as ‘having symptoms’ if they’d gone one, but not two, visits without reporting any symptoms. However, in their original analysis, participants were only followed for a median of 80 days (less than 11 weeks).
As a result, some participants who would have been classified as ‘not having symptoms’ if they’d been followed a little bit longer were still classified as ‘having symptoms’ at the end of their observation period. (In the jargon, their follow-up time was ‘right-censored’.) This is shown in the diagram below, taken from the ONS:
In the ONS’s updated analysis, which followed participants for a median of 204 days, individuals in the situation of Participant D above were correctly classified as ‘not having symptoms’ before the end of their observation period.
Using this revised method, the ONS found that less than 1% of children aged 2-11 continue to report symptoms 12 weeks after infection, with the figure rising to just 1.2% for those aged 12-16. Hence long Covid is particularly rare in children, further undermining the case for vaccinating that age-group.
While the ONS deserves credit for being completely transparent about the limitations of their original analysis, their updated analysis is still open to the criticism I mentioned above. This means that 2.5% should probably be considered an upper bound on the chances of getting long Covid, the true figure being somewhat lower.