Doctoral thesis published by Mathilde Mercat

7 July 2021

Conception et évaluation d’une méthode d’estimation d’une probabilité d’infection d’un troupeau à partir de données hétérogènes: contribution au développement d’une surveillance épidémiologique basée sur la comparabilité des résultats

Mathilde Mercat
Alternative Title
Design and evaluation of a method for the estimation of a herd-level probability of infection from heterogeneous data : contribution to the development of output-based surveillance
On a territorial scale, programmes to contrai non-regulated infectious diseases of cattle have multiple benefits. They also create difficulties in exchanges between territories because the definitions of 'infection-free' status differ between programmes. Estimating a probability (of absence) of infection for each herd calculated independently of the surveillance modalities would make it possible to secure trade in animais between territories. This type of estimate could be used for output-based surveillance, a type of surveillance based on a result to be achieved and not on the means implemented. The objectives of this thesis work were to contribute to the development and evaluation of a method for estimating infection probabilities at the herd level, based on heterogeneous surveillance data. Using the example of bovine viral diarrhea virus infection, relevant and available information was identified and organised. The model developed is a hidden Markov model estimating a probability of infection at the herd level from repeated test results and risk factors for infection. lts performance was evaluated on simulated data representing a variety of infection dynamics and contrai programmes. The evaluation showed that the added value of the model was greater the lower the sensitivity of the diagnostic test. The added value of the risk factors was moderate in the range of situations evaluated. The use of this model requires further development for the classification of herds as free/infected based on predicted infection probabilities.