Assessment of bovine tuberculosis surveillance effectiveness in French wildlife: An operational approach

Bovine tuberculosis (bTB) is a chronic, zoonotic, bacterial disease mostly caused by Mycobacterium bovis, which can affect both domestic and wild species. France was officially declared bTB-free in 2001 but faced since 2004 an increase of the prevalence in cattle. Since 2001, bTB has been detected in several wild species: red deer (Cervus elaphus), roe deer (Capreolus capreolus), wild boar (Sus scrofa) and badger (Meles meles). Infected wild species constitute a major threat, because they may contribute to the maintenance of the infection in cattle and prevent eradication. In 2011, a surveillance system, Sylvatub, was implemented nationwide, to monitor the epidemiological status of bTB in mainland France. Our objective in this study was to assess the effectiveness of one of Sylvatub's passive surveillance system components (SSCs), which is based on the visual inspection of hunted animals (wild boars, red and roe deer) throughout mainland France. The following effectiveness criteria were evaluated: individual-level and component group-level positive and negative predictive values, and individual-level and component group-level probabilities of type I error ("false-positive" error) and type II error ("false-negative" error). These criteria were estimated quantitatively, at the scale of the département (a French administrative area of similar size to a county), with a stochastic scenario tree model. Individual negative predictive values were high, whereas individual positive predictive values were poor, whatever the species considered, and the training of hunters did not improve these effectiveness criteria sufficiently. The individual-level probability of type I error was relatively low, but the individual-level probability of type II error was generally high and was therefore an issue. However, increasing the proportion of trained hunters decreased this probability effectively. At group level, the size of the population surveyed had a marked impact on the effectiveness criteria: both the component group-level negative predictive value and the component group-level probability of type II error decreased rapidly with increasing population size. The conclusions drawn from such rationale may directly benefit stakeholders and actors in the field in their day to day practice of the surveillance processes. Thus, the assessment method used in this article presents the advantage of being operational, as well as being applicable to any surveillance system.