Using Markov Decision Processes to define an adaptive strategy to control the spread of an animal disease

A strategy to control an animal disease within an area is often based on one or more actions systematically implemented. In this paper, we illustrate how to use a Markov Decision Process (MDP) to compute an adaptive strategy depending on the pathogen spread within a group of farmers with only one decision-maker for the group. The objective at the group level is to decrease the cost of the disease and its control. Status for each farm is assumed to be exactly known each year by the decision-maker. Possible actions each year are Doing nothing or Vaccinating. The computed MDP policy results in a non-systematic vaccination. Although the objective is only based on the total costs, the computed MDP policy reduces the prevalence, that is the amount of infected herds, compared to a systematically Doing nothing strategy.

► We compute an optimal and adaptive strategy in terms of disease and control costs.
► The strategy is a threshold between systematically Vaccinating or Doing nothing.
► Our strategy reduces the prevalence with similar costs to the Do Nothing strategy.