RobustInfer

The control of infectious animal diseases is a major challenge for sustainable animal husbandry, to limit public health risks (zoonoses), and to improve animal health and welfare. Better understanding and anticipation of the spread of diseases at territorial level strengthens decision-making capacities in the face of current and future threats.

As observed during recent health crises (ASF, covid-19), mechanistic models are powerful tools to describe spatio-temporal infection dynamics. They allow consideration of large scales, which are essential in a context of intensified and globalised animal trade. They help to support the decisions of health managers and public decision-makers, in particular by guiding surveillance efforts and the deployment of control interventions. However, if these models are not properly calibrated, the quality of predictions, and therefore their relevance, can be considerably reduced. Thus, they need to be fed with observational data, the collection of which becomes easier and more extensive. However, relevant procedures are missing to connect such models to data, especially for large and partially observed systems.

This project aims to better understand how to use available data for complex epidemiological systems on a large scale (region, territory) in order to parameterise the mechanistic models representing them. One of the objectives is to provide elements of understanding on how to build criteria (summary statistics) for summarising observational data, and to mobilise them in inference approaches adapted to the specificities of these systems (large scale, variability of dynamics, partial observation, etc.). This will make it possible to improve the integration of empirical information into mechanistic models and guarantee more reliable and realistic models, thus more useful for prioritising control strategies.

network_cropped_pdll.jpg