Differences in SARS-CoV-2 rapid diagnostic testing usage by COVID-19 vaccination status can bias vaccine effectiveness estimates against symptomatic SARS-CoV-2 infection from a test-negative design
Ryan E Wiegand, Gordana Derado, Allison Avrich Ciesla, Saba A Qasmieh, Jill M Ferdinands, Brendan Flannery, Ruth Link-GellesAbstract
Background
Use of rapid diagnostic tests (RDTs) may bias vaccine effectiveness (VE) estimates from test-negative design (TND) studies. We conducted individual-level simulations using national, US COVID-19 surveillance data to assess potential bias at varying likelihoods of visiting a healthcare facility and prevalence of RDT use.
Methods
Twenty-one scenarios that varied the probabilities of visiting a healthcare facility and RDT use by vaccination status were simulated. VE was calculated as 100 × (1 − odds ratio), comparing the odds of symptomatic SARS-CoV-2 infection between vaccinated and unvaccinated groups of people. VE estimates were compared to VE estimates from simulations without RDT testing to calculate the bias in percentage points (pp).
Results
VE was underestimated when either vaccinated people were more likely (probability of .5) than unvaccinated people (probability of .3) to visit a healthcare facility after a positive RDT [−5.8%age points (pp); 95% uncertainty interval (UI): −9.1, −2.4] or unvaccinated people were more likely than vaccinated people to visit a healthcare facility after a negative RDT (bias = −6.1 pp; UI: −9.9, −2.5). Conversely, VE was overestimated when either vaccinated people were more likely than unvaccinated people to visit a healthcare facility after a negative RDT (bias = 5.2; UI: 2.4, 7.7) or unvaccinated people were more likely than vaccinated people to visit a healthcare facility after a positive RDT (bias = 8.4 pp; UI: 5.6, 11.3).
Conclusion
VE estimates against symptomatic SARS-CoV-2 infection from a TND may be under- or overestimated, depending on differences in RDT usage between vaccination groups.