The difficulties of estimating long Covid

September 27, 2021


David Spiegelhalter and Anthony Masters write in The Guardian about the challenges of establishing a good figure for those affected by long Covid.

Many viruses can lead to prolonged effects after the infection has passed and Sars-CoV-2 is no exception: the NHS has compiled a long list of symptoms of long Covid. However, a recent Telegraph headline states: “More than half of Britons suffering from long Covid might not actually have it.” Is that what the data says?

The Office for National Statistics conducts regular surveys of private households, testing for Covid infections and asking about health. But there is no diagnostic test for long Covid, or even a consensus on a definition, so a range of exploratory statistical methods must be used.

The first approach uses a checklist of 12 classic symptoms, including fever, cough and loss of taste or smell, and found around one in 20 (5%) of participants reported any of these symptoms more than 12 weeks after testing positive. For similar “controls” who had tested negative, the estimate was around 3.5%; this does suggest most cases might have had those symptoms even without a Covid infection.

But there are problems with the Telegraph headline. First, this symptom list excludes “brain fog”, sleeplessness and other features linked to long Covid. Second, that proportion (5%) just refers to those who tick a box, rather than believing they have long Covid. The data is also just a snapshot. So ONS also tracks participants continuously, producing a much larger contrast: around 3.0% of those who have tested positive report one of the symptoms at consecutive visits over the whole 12 weeks, compared with just 0.5% in controls. Of course, this method fails to capture relapses.

A much larger estimate is found if people are directly asked if they have long Covid; after 12 weeks, the self-classified estimate was 11-12%, with around two-thirds of these saying their symptoms caused some limitation to their daily activities.

Other teams’ estimates have been both higher and lower and, unfortunately, there is no “correct” statistical approach. Many people are undoubtedly experiencing long-term consequences of Covid but, in the absence of an agreed definition, diagnosis and treatment are likely to become very challenging issues in the future.

David Spiegelhalter is chair of the Winton Centre for Risk and Evidence Communication at Cambridge. Anthony Masters is statistical ambassador for the Royal Statistical Society

Shopping Basket