OBJECTIVES: Drug-eluting stents (DES) elute drugs that prevent coronary restenosis after a percutaneous revascularization. The cost of DES is quite high therefore imposing a relevant burden to hospital and NHS budgets. Patients selection and reimbursement policies need to be settled according to a large amount of clinical (number and anatomy of revascularized vessels, diabetes, unstable angina), economic (local cost of devices and procedures) and organizational variables (adoption rate of different types of DES and BMS, number of stent/procedure, shift from surgical to DES revascularization). The resulting decision models are usually blamed for high complexity but poor completeness and low transparency. METHODS: Bayesian networks (BN) are close acyclic graphs displaying value nodes, probability nodes and, possibly, decision nodes (influence diagrams). We explored the application of BN to policy selection for DES use, in both the NHS and the hospital perspective. Three commercial softwares were compared: Netica, TreeAgePro, HuginLite. Beta and gamma distributions were used to represent the probability of conditions/events and unit costs, respectively. Log-normal distributions were used to model relative risks. RESULTS: BN modeling of the DES problem was feasible with the 3 softwares and required 35-40 nodes. We tested 5 types of restriction policies to DES use, anchored to: 1) the presence of diabetes; 2) the number of vessels to be revascularized; 3) the overall risk of restenosis; 4) a cutoff adoption rate of DES; and 5) a cutoff budget. We could verify that flexibility, directness and transparency varied among the softwares. Also software modalities to manage uncertainty and to report the results had different appeal. CONCLUSIONS: BN allows rapid and transparent estimations of the local/specific economic impact of several different cost-containment policies, i.e. restriction for use of DES, along with the uncertainty surrounding the estimate.