Descriptive spatial analysis of BSE in western FranceDavid Abriala, Didier Calavasb, Nathalie Lauvergneb, Eric Morignatb and Christian Ducrota
a Unité d'Epidémiologie Animale, INRA Theix, 63122 Saint-Genès-Champanelle, France
b Unité Epidémiologie, AFSSA Lyon, 31 avenue T. Garnier, 69364 Lyon Cedex 07, France
(Received 19 December 2002; accepted 15 April 2003)
The spatial heterogeneity of Bovine Spongiform Encephalopathy (BSE) was analysed on the 84 cases confirmed in western France (WF) between August and December 2000, when both the Mandatory Reporting System and an active surveillance on cattle at risk were running. Ninety-four percent of these cases were born between June 1993 and June 1996, and we analysed the location at birth. One disease mapping and two clustering methods (Scan of Kulldorff and the method of Besag and Newell) were used. In order to attenuate the contrasts artificially created by the standard disease mapping method (over-dispersion), we estimated the Standard Incidence Ratio (SIR) with a Bayesian method (Poisson-Gamma model) allowing a smoothing of the estimators. The geographical location of interest was the "canton", that divided the total area into 526 geographical units. The background population (2.6 million cattle) was obtained from the Agricultural Census 2000. We tested the hypothesis of a homogenous spatial distribution of the BSE risk where the expected number of BSE cases per unit area was obtained by applying the overall BSE rate in WF to each "canton", standardised on the type of breed, dairy versus beef suckler. The SIR ranged from 0.80 to 2.18 and the spatial distribution of BSE cases was significantly heterogeneous. Two spatial clusters were detected with the spatial scan statistics of Kulldorff and the method of Besag and Newell (18 to 20 observed BSE-cases per cluster with a radius of 45 km) centred on the "département" of Côtes-d'Armor and Mayenne. Another cluster was detected with the method of Besag and Newell (9 observed BSE-cases) in the "département" of Finistère. The results proved that the risk of BSE is linked to the geographical location in the area of the study.
Key words: disease mapping / Bayesian model / BSE / cattle
Correspondence and reprints: David Abrial firstname.lastname@example.org
© INRA, EDP Sciences 2003