Better methods for diagnostic test evaluations in infectious diseases

The thesis focuses on the application of epidemiological and biostatistical methods for better test accuracy assessment in infectious diseases, particularly in the pediatric population and in population health surveillance. We explore and propose a modelling approach that includes clinical and/or population health relevant constructs in the assessment of test accuracy and apply this to the respiratory pathogen SARS-CoV-2.
We examine the severity of the multi-inflammatory syndrome in children and the challenges in diagnosing past infections during the early pandemic. We provide seroprevalence estimates and risk factors for seropositivity from a prospective cohort study in Belgian schools. Additionally, we report on the development, initiation, and analysis of the data provided by the surveillance system of Flemish school COVID-19 cases.
Throughout the thesis, we use Bayesian latent class analysis to estimate the sensitivity and specificity of tests to diagnose SARS-CoV-2 past-infection rates. We then improve the more classically used latent class model for the assessment of the prevalence and the accuracy estimates, via the decomposition of the test accuracy question into its elements i) the tests under evaluation, ii) their measurands and iii) the target condition(s). We use directed acyclic graphs to visualize the model and Bayesian inference to obtain the estimates. We find more realistic estimates for the accuracy of the tests with regards to the measurands and the tests with regards to the different target conditions, clarifying their clinical performance to diagnose cases depending on the moment of testing. This corroborates the importance of explicitness regarding the target condition.