Potential biases arising from epidemic dynamics in observational seroprotection studies medrxiv.org/content/10.110… #COVID19 w @rebeccajk13 @LeeKShaffer @yhgrad and Jamie Robins.
Seroprotection studies in #COVID19 willbe quite challenging.Because we are trying to study we got to the fact of one infection against future infection, Exposure and outcome are essentially the same event at different times. This creates many opportunities for confounding.
Often direction will be positive. People at higher infection risk in the past will continue to be at higher risk of re-infection. Our paper shows that can happen but also the opposite: confounding makes seropositiviy look protective when epidemic is controlled....
...so those enrolled earlier in their local epidemic are more likely seronrgative and exposed to higher hazard of infection.
Careful matching and stratification by risk can help but local network effects can still create bias. Most problematic if seropositives are unprotected or weakly protected. Less of a problem if strong immunity.
