Monte Carlo simulation of flock-level sensitivity of abattoir surveillance for ovine paratuberculosis

We used Monte Carlo simulation to estimate distributions for flock-level sensitivity of abattoir-based surveillance for ovine paratuberculosis as currently practised in New South Wales, Australia. Probability distributions were used as input variables for within-flock prevalence, years-infected and individual animal-level sensitivity and specificity of gross pathology as a screening test for the presence of paratuberculosis. Distributions used as inputs for the size of abattoir-slaughter groups were based on existing abattoir-surveillance data from NSW. Predicted flock-level sensitivity depended on within-flock prevalence and the number of animals examined and was sensitive to estimates of animal-level sensitivity and specificity. The median probability of detection of an infected flock based on the examination of one abattoir line was predicted not to exceed 0.95 unless the within-flock prevalence was '7%. If the within-flock prevalence was 2%, the probability distribution of flock-level sensitivity had a median of 0.73, with 80% of values lying between 0.55 and 0.84. Improvement in the flock-level sensitivity could be achieved by submitting more than three gross pathology-positive specimens per line, if available - but the degree of improvement depended on the number of sheep slaughtered (line size) and the within-flock prevalence. At 2% prevalence, a median flock-level sensitivity of 0.95 could be obtained in lines of >390 sheep if six gross pathology-positive specimens were submitted. We concluded that abattoir surveillance based on identification of gross pathology as a screening test is not a sensitive tool for detecting recently infected flocks or flocks which have a moderate or lower prevalence of infected animals. But - with relatively minor modifications of the protocol currently in use - it could become a key component of a surveillance programme which included additional testing strategies for small flocks.