A0366
Title: Bayesian inference of sampling weights in COVID-19 testing
Authors: Vasileios Giagos - University of Essex (United Kingdom) [presenting]
Abstract: Published COVID-19 testing results provide a daily source of information about the pandemic and, at the initial stages, testing prioritisation has been given to symptomatic patients. This prioritisation introduces an inherent selection preference towards symptomatic cases. We view COVID-19 daily testing results as a weighted sampling process from distinct subpopulations with the sampling weights being the parameters of interest. We incorporate the distribution of the weighted samples, the Wallenius distribution, in a Bayesian setting to estimate and compare their posterior distributions while we address identifiability challenges. In addition, we address computational challenges by proposing an efficient approximation based on the Linear Noise Approximation which demonstrates indistinguishable results under simulated experiments. Finally, we use the Diamond Princess COVID-19 outbreak as a case study to infer the testing priorities of the symptomatic/asymptomatic/healthy groups, and we showcase the flexibility of the Linear Noise Approximation by incorporating the testing mechanism in an epidemiological model to track the dynamics of the COVID-19 outbreak on board the Diamond princess.
