COST-EFFECTIVENESS ANALYSIS WITH MARKOV INFLUENCE DIAGRAMS

Purpose: To introduce Markov influence diagrams as a graphical tool for performing complex cost-effectiveness analysis

Method: Markov influence diagrams (MIDs) are a new type of probabilistic graphical model (PGM). The basic elements of a PGM are a probability distribution and a graph that represents the structural assumptions of the model, such as probabilistic conditional independencies. The nodes of the graph represent decisions, probabilistic events, and the decision maker’s values. The joint (global) probability distribution of a PGM is defined as the product of several small conditional probability distributions, each associated with a variable of the model. This modular representation makes it possible to build probabilistic models involving dozens of variables. MIDs, which extend influence diagrams in the same way as Markov decision trees extend decision trees, are suitable to represent state-transition models. Using a causal graph that may contain several variables per cycle, MIDs can model various features of the patient without multiplying the number of states; in particular, they can represent the patient history without using tunnel states. There are two main types of algorithms for evaluating state-transition models, including MIDs: cohort analysis and individual-level simulation. Many MIDs, including some of those that combine categorical and numeric variables, can be evaluated with cohort analysis, which is more efficient than individual-level simulation.

Result: OpenMarkov, an open-source tool, allows decision analysts to build MIDs and perform cost-effectiveness analysis—including several types of sensitivity analysis—with a graphical user interface, without writing any code. Using this tool we have built MIDs for several medical problems, which are publicly available on internet. It was much easier and faster than implementing the same models as spreadsheets, decision trees, or computer programs, especially because OpenMarkov’s graphical user interface saves a lot of time and drastically reduces the probability of making mistakes.

Conclusion: MIDs can be used to easily build and evaluate complex models whose implementation as spreadsheets or decision trees would be cumbersome or unfeasible in practice. Many problems that previously required discrete event simulation can be solved with MIDs, i.e., within the paradigm of state-transition models, in which many health economists feel more comfortable. However, there are other types of problems for which discrete event simulation or dynamic transmission models are more adequate.