| Abstract | Escherichia coli O157 emerged about two decades ago; Scotland has had among the highest incidences of human infection in the world. Two recent surveys (Chase-Topping et al 2007; Gunn et al 2007) concluded that E. coli O157 was present on c. 20% Scottish cattle farms, though with large variation in on-farm prevalence. Here, we considered dynamical processes for the spread of E. coli O157 infection between farms using stochastic models, and fitted the models by maximum likelihood to epidemiological survey data. Based on Akaike information Criterion (AIC), the most likely dynamical process is identified. The best model incorporates spread by the movement of cattle and by other routes including infection from an environmental reservoir. Adding decay of transmission with distance, seasonality, or the presence/absence of other livestock animals did not improve model fit. However, there was evidence of an effect of farm size. Further, a distribution of sets of parameter for the best model was generated by Markov Chain Monte Carlo methods. Using this distribution, the potential impacts of different control measures were simulated and sensitivity to changes in model parameters was explored. Reducing cattle movement between farms had only a small effect. Measures to reduce transmission rates (e.g. improved biosecurity) may be most effective if targeted at larger farms. |
|---|
