Scenario tree modelling to inform surveillance design for maintaining freedom from Coxiella burnetii infection in Australian commercial dairy goat herds.

We used scenario tree methods to determine how different disease detection methods might be used to provide quantitative evidence that Australian dairy goat herds are free of coxiellosis. The aim of our proposed C. burnetii surveillance programme is to find evidence of the absence of antigen as well as evidence of an absence of an immune response to C. burnetii infection in individual dairy goat herds. We defined a C. burnetii infected dairy goat herd as a herd in which at least one doe was showing evidence of either active infection or past C. burnetii exposure using four candidate surveillance system components (SSCs): (1) testing of individual doe whole blood using the C. burnetii com1 PCR; (2) testing of individual doe whole blood using the C. burnetii ELISA; (3) testing bulk tank milk (BTM) using the com1 PCR and the C. burnetii ELISA; and (4) investigations of abortions and stillborn kids submitted to a diagnostic laboratory for testing. Of eight candidate surveillance strategies (combinations of the SSCs listed above) individual doe ELISAs every six months combined with monthly BTM PCR and ELISA testing returned the highest surveillance system sensitivity of 0.963 (95% probability interval [PI] 0.911-0.982) for the lowest cost, at AUD 28.94 (95% PI 28.38-30.59) over a 12-month period, for every one percent increase in surveillance system sensitivity. Assuming a probability of disease freedom of 0.10 at the start of the surveillance program and a probability of C. burnetii introduction per month of 0.01 we estimate that 95% confidence that C. burnetii was absent from a herd could be achieved after a single round of individual doe ELISAs followed by period of 6 consecutive monthly BTM PCR and ELISA tests. The results of this study show that selection of the most efficient combination of surveillance system components requires a good understanding of initial herd C. burnetii status and the probability of introduction of infection and how this may change over time. Scenario tree analyses results have provided insight into the key determinants of C. burnetii detection ability.