Bayesian zero-inflated predictive modelling of herd-level Salmonella prevalence for risk-based surveillance.

摘要: The national control programme for Salmonella in Danish swine herds introduced 1993 has led to a large decrease pork-associated human cases of salmonellosis. pork industry is increasingly focused on the cost-effectiveness surveillance while maintaining consumer confidence food supply. Using data from 2003 and 2004, we developed zero-inflated binomial model predict which farms were most at risk Salmonella. We preferentially sampled these high-risk using two sampling schemes based predictions resulting farm’s covariate pattern its random effect. Zero-inflated modelling allows assessment similarities differences between factors that affect herd infection status (introduction), those seroprevalence infected (persistence spread). Both (producing greater than 5000 pigs per annum), small less 2000 annum) significantly higher subsequent seroprevalence, when compared with medium sized annum). When being located elsewhere, south Jutland decreased infection, but increased pig an seropositive. suggested many where was not detected infected, low prevalence. cost sensitivity, results our under standard scheme, size, recently risk-based approach. Model-based sensitive show significant savings. Further refinements, methods evaluate their performance are important areas future work, should continue occur direct consultation authorities.