A stochastic predictive model for the natural spread of bluetongue

Abstract In recent years the vector-borne diseases (VBD) are (re)-emerging and spreading across the world having a profound impact on human and veterinary health, ecology, socio-economics and disease management. Arguably the best-documented example of veterinary importance is the recent twofold invasion of bluetongue (BT) in Europe. Much attention has been devoted to derive presence-absence habitat distribution models and to model transmission through direct contact. Limited research has focused on the dynamic modelling of wind mediated BT spread. This paper shows the results of a stochastic predictive model used to assess the spread of bluetongue by vectors considering both wind-independent and wind-mediated movement of the vectors. The model was parameterised using epidemiological knowledge from the BTV8 epidemic in 2006/2007 and the BTV1 epidemic in 2008 in South-France. The model correctly reflects the total surface of the infected zone (overall accuracy = 0.77; sensitivity = 0.94; specificity = 0.65) whilst slightly overestimating spatial case density. The model was used operationally in spring 2009 to predict further spread of BTV1. This allowed veterinary officers in Belgium to decide whether there was a risk of introduction of BTV1 from France into Belgium and thus, whether there was a need for vaccination. Given the far distance from the predicted infected zone to the Belgian border, it was decided not to vaccinate against BTV1 in 2009 in Belgium.