Bayesian statistical meta-analysis of epidemiological data for QRA

We aim to pick up a maximum of information from data of epidemiological studies to introduce them in a quantitative risk assessment (QRA) model. Our objective is to estimate the probability of disease from epidemiological data sets. The proposed statistical meta-analysis describes all the probabilities of the available data from the parameters of a multinomial model based on an adapted partition of the whole population. Our target-parameter, the disease probability, is a function of the parameters of the multinomial distribution. To gather all the probabilities of interest in the same framework, we used a Bayesian approach assuming noninformative priors and priors based on expert opinions by means of a Dirichlet distribution on the multinomial parameters. This methodology was applied to acute gastroenteritis due to campylobacter. The proposed approach allows an estimation of the disease probability using all information around the disease reflecting uncertainty from all pieces of information