Interpreting and optimising bulk milk testing for BVD and IBR considering numerical test outcomes

Many diagnostic assays provide a numerical result that is then turned into a positive/negative outcome by applying a cut-off, with a loss of potentially valuable information. This research project proposes to develop and apply Bayesian latent class models (BLCMs) to support interpretation of serological diagnostic tests using additional information derived from the numerical outcome. The project builds on the results of a previous CoVetLab project that successfully proved the feasibility of inter-laboratory diagnostic test evaluation for the serological detection of infectious bovine rhinotracheitis (IBR) and bovine viral diarrhoea (BVD). We will now focus on bulk-milk testing, which will greatly benefit from the use of numerical results and for which diagnostic test evaluation of numerical test results has not yet been done. The ultimate outcome of the project is to determine the probability of a herd being truly seropositive for BVD/IBR given a certain bulk milk test result. The key objectives are: (1) Develop a BLCM method to analyse numerical test outcomes from multiple diagnostic tests. (2) Compare the test performance of serological tests currently applied for routine bulk milk diagnostics of BVD and IBR in the partner institutes. (3) Develop tools to optimise test performance for country-specific requirements. (4) Develop a dashboard for test result interpretation that provides the probability of infection given the numerical test outcome. As well as knowledge sharing and harmonization between institutes, the collaboration within the CoVetLab framework is a requirement for accessing the sample material used for this inter-laboratory diagnostic test evaluation.