Colistin resistance has been the subject of much attention since mcr genes encoding plasmid-mediated colistin resistance description in 2015. To date, surveillance data about resistance levels encountered in food-producing animals are scarce. In France, the Resapath dataset, consisting in a large collection of disk diffusion antibiogram results transmitted by a network of laboratories. It offers a unique opportunity to study the evolution of resistance towards colistin over the past 15 years in Escherichia coli isolated from diseased food-producing animals. This study used a Bayesian hierarchical Gaussian mixture model to estimate the resistant proportions from those data. This non-classical approach deals with the colistin-specific problem of overlapping distributions of diameters measured for susceptible and resistant isolates that makes the definition of epidemiological cut-off very hard. This model also considers the variability observed between the measurements performed by different laboratories. Proportion of resistant isolates has been calculated for several food-producing animals and most encountered diseases. From those estimations, a marked evolution of the proportions of resistant isolates is noticeable, for swine suffering from digestive disorders. In this group, an increase over the 2006-2011 period from 0.1% [ 0.0%, 1.2%] in 2006-28.6% [25.1%, 32.3%] in 2011 was followed by a decrease to reach 3.6% [2.3%;5.3%] in 2018. For isolates related to digestive disorders in calves, percentages increased and reached 7% in 2009 then decreased as for swine. In contrast, for poultry productions, estimated proportions and credibility intervals were constantly very close to zero.