Introduction to Decision-Theoretic Networks for Pharmacoeconomics

Cost-effectiveness evaluations or budget impact analysis represent a typical example of decision analysis problems, as they are performed for decision making regarding reimbursement of drugs or health care budget allocation under uncertainty. In contrast to traditional tools for decision making such as regression equations, decision trees and Markov models, Bayesian Networks and Influence Diagrams have not been frequently used in pharmacoeconomics. In this paper we aim to provide an introduction to these so called decision-theoretic networks. A formal introductory background is followed by an example of a cost-effectiveness evaluation of antifungal therapy. The example provides a comparison of the traditional decision tree approach and an equivalent Influence Diagram. Furthermore it will be shown how Influence Diagrams can be extended to incorporate parameter uncertainty and to perform probabilistic sensitivity analysis in a more efficient way. We believe that for their ability in representing probabilistic structure, facilitating communication between analysts and experts, and their computational facilities, decision-theoretic networks provide a useful expansion of the analytic tools in pharmacoeconomics.