Markov Influence Diagrams.

Area of Science:

• Decision Analysis
• Health Economics
• Probabilistic Graphical Models

Background:

• Traditional decision trees and influence diagrams can be cumbersome for complex state-transition models.
• Modeling patient characteristics and history often requires complex state structures or discrete event simulation.
• Health economists prefer working within the state-transition model paradigm.

Purpose of the Study:

• Introduce Markov influence diagrams (MIDs) as an extension of influence diagrams for state-transition modeling.
• Demonstrate the utility of MIDs in building and evaluating cost-effectiveness models, particularly in medicine.
• Highlight the capabilities of MIDs in representing complex patient characteristics and history without increasing state complexity.

Main Methods:

• MIDs utilize causal graphs with multiple variables per cycle to model patient characteristics.
• The OpenMarkov software provides a graphical user interface for building and evaluating MIDs.
• MIDs facilitate cost-effectiveness analysis and various sensitivity analyses (deterministic and probabilistic).

Main Results:

• MIDs can represent patient history without requiring tunnel states, simplifying model structure.
• Complex models, previously unfeasible in spreadsheets or decision trees, can be effectively built using MIDs.
• Many problems solvable by discrete event simulation can be addressed within the MID framework.

Conclusions:

• Markov influence diagrams provide a powerful and flexible tool for state-transition modeling in health economics.
• The OpenMarkov tool democratizes the use of MIDs, enabling complex analyses without coding.
• MIDs offer a practical alternative to discrete event simulation, aligning with health economists' preferred modeling paradigm.