A major Stanford-led analysis has challenged conventional estimates of long COVID, finding that traditional study designs may vastly overestimate how many long-term health problems can truly be attributed to the virus.
According to The Stanford Daily's November 3 research round-up, the study analysed 14.4 billion American health insurance claims and examined two sources of bias in conventional studies: comparing positive COVID testers with people who were never tested, and grouping negative-testing subjects in hindsight instead of at the time of the test.
By enrolling subjects at their first polymerase chain reaction (PCR) test and then grouping them by positive or negative result, the researchers found 98 percent fewer COVID-attributable outcomes from four to 12 months after infection. Myopathy, hair loss, pulmonary embolism and respiratory failure stood as the only effects with substantially elevated risk — and by one year, nothing was clearly attributable to COVID, the study found.
The authors stress that the findings do not mean long COVID is 98 percent less common, or that certain symptoms are not genuine; only that it is unclear whether COVID causes them, the round-up reported. The study was published in the Proceedings of the National Academy of Sciences on October 27, co-senior-authored by computer science professor Jure Leskovec.
The result does not dismiss patients' experiences, but it does complicate the science: symptoms that follow an infection are not automatically caused by it, and careful comparison groups matter enormously when measuring long-term effects.
What happens next: researchers say future long COVID studies will need the same rigorous enrolment-at-test design if they want to separate genuine COVID effects from background illness in the general population.