Danish Meat Trade College

The national control programme for Salmonella in Danish swine herds introduced in 1993 has led to a large decrease in pork-associated human cases of salmonellosis. The pork industry is increasingly focussed on the cost-effectiveness of surveillance while maintaining consumer confidence in the pork food supply. Using national control programme data from 2003 and 2004, we veloped a zero-inflated binomial model to predict which fanns were most at risk of Salmonella. We preferentially sampled these high-risk farms using two sampling schemes based on model predictions resulting from a farm's covariate pattern and its random effect.

This study reanalyses data using a new analytical approach, taking into account, that some risk-factors will increase the risk for a herd to be positive, whereas other risk factors tend to increase the risk of a high prevalence in positive herds. The zero-inflated model showed that use of pelleted feed increased the risk of being positive for the herd. In positive herds, use of pelleted feed, dry feed, increasing herd size and conventional (non-SPF) health status increased the prevalence.

The control programme for Salmonella infection in Danish swine has reduced the number of human cases attributable to pork consumption and the focus is now on cost-effectiveness. We applied time-series and longitudinal analyses to data collected between January 1995 and May 2005 to identify if there were predictable periods of risk that could inform sampling strategy; to investigate the potential for forecasting for early aberration detection; and to explore temporal redundancy within the sampling strategy. There was no evidence of seasonality hence no justification to change to targeted sampling at high-risk periods. The forecast of seropositivity made using an ARIMA (0, 1, 2) model had a root-mean-squared percentage error criterion of 8.4% indicating that accurate forecasts are possible. The lorelogram identified temporal redundancy at up to 10 weeks suggesting little value in sampling more frequently than this on the 'average' farm. These findings have practical applications for both farm-level sampling strategy and national-level aberration detection which potentially could result in a more cost-effective surveillance strategy.