The impact of inaccurate assumptions about antibody test accuracy on the parametrisation and results of infectious disease models of epidemics

The parametrisation of infectious disease models is often done based on epidemiological studies that use diagnostic and serology tests to establish disease prevalence or seroprevalence in the population being modelled. During outbreaks of an emerging infectious disease, tests are often used, both for disease control and epidemiological studies, before studies evaluating their accuracy in the population have concluded, with assumptions made about accuracy parameters like sensitivity and specificity. In this simulation study, we simulated such an outbreak, based on the case study of COVID-19, and found that inaccurate parametrisation of infectious disease models due to assumptions about antibody test accuracy in a seroprevalence study can cause modelling results that inform public health decisions to be inaccurate; for example, in our simulation setup, assuming that antibody test specificity was 0.99 instead of 0.90 when it was in fact 0.90 led to an average relative difference of 0.78 in model-projected peak hospitalisations, even when test sensitivity and all other parameters were accurately characterised. We therefore suggest that methods to speed up test evaluation studies are vitally important in the public health response to an emerging outbreak.