Bayesovské modely pro vyhodnocení výsledků binárních diagnostických testů

The COVID-19 pandemic presented an unprecedented challenge to global healthcare systems. The implementation of rapid antigen tests (RATs) was crucial in limiting the spread of the virus; however, their true accuracy and reliability often did not match the claims made by manufacturers. This work focuses on the application of the Bayesian approach to the interpretation of antigen test results, aiming to explicitly propagate and reduce uncertainty in the diagnostic process using informative priors and a Bayesian hierarchical model that integrates the clinical and epidemiological context in the form of symptomatology and indication. The methodology includes simulation methods to generate posterior distributions of predictive values and the use of beta distributions to model sensitivity, specificity and prevalence. The hierarchical Bayesian model allows for the integration of uncertainty at multiple levels, providing a deeper insight into the probability of true infection. The results show that the use of informative priors and the hierarchical model significantly improves the accuracy of estimates and greatly reduces the uncertainty in diagnostic conclusions. The developed practical clinical decision support application enables healthcare professionals to interactively use this model, supports informative decisions and reduces the risk of erroneous diagnostic conclusions compared to the current state (uninformative prior). This work contributes to a better understanding of the importance of integrating clinical and epidemiological context into the diagnostic process and demonstrates how the Bayesian approach can improve the quality and efficiency of diagnostic methods. The proposed models have the potential to be applied to other diagnostic tools and clinical situations, contributing to better preparedness for future epidemics and increasing the informativeness of clinical decisions.