Interpreting Covid-19 Tests and the Uncertainty by Bayesian Methodology

Tomáš Karel
Statistika, 105(2): 214–226
https://doi.org/10.54694/stat.2024.36

Abstract
The global impact of the Covid-19 pandemic has highlighted the urgent necessity for the development of rapid, effective diagnostic methods. Rapid antigen tests (RAT) have emerged as a key tool in this regard due to their speed and cost-efficiency. Nevertheless, the accurate interpretation of RAT results is challenging due to various factors, including the viral load, the quality of the sample, and the patient's status. This study demonstrates the advantages of Bayesian methods, which are capable of propagating posterior uncertainty in the form of the entire posterior distribution. It also highlights the benefits of using informative priors, which significantly reduce uncertainty in diagnostic parameters, lower false negative rates, and improve clinical decision-making. The results emphasize the need for precise interpretation of RAT results including uncertainty. Employing Bayesian simulations for posterior predictive values can reduce diagnostic errors and improve public health outcomes by upgrading the performance of RATs and explicitly propagating posterior uncertainty in clinical diagnosis, as described in this study.

Keywords
Bayesian statistics, posterior simulations, informative prior, Covid-19