Using Parameter Constraints to Choose State Structures in Cost-Effectiveness Modelling.

Area of Science:

• Health economics
• Decision science
• Mathematical modeling

Background:

• The choice of state structure in cost-effectiveness multi-state models significantly impacts key outputs like treatment recommendations.
• Uncertainty exists in defining states, such as whether similar disease severities or clinical events should be separate or merged.
• Standard statistical methods for model comparison are invalidated by merging states, which aggregates data.

Purpose of the Study:

• To propose and illustrate a method for comparing state structures in cost-effectiveness models.
• To enable valid comparisons of different state structures without altering the underlying data.
• To facilitate informed decisions on treatment and research prioritization.

Main Methods:

• Re-expressing a model with merged states into a larger state space with constrained equal transition probabilities, costs, and utilities.
• Comparing maximised likelihoods or information criteria between constrained and unconstrained models to test state equality.
• Allowing for different structures for rates, costs, and utilities as needed.

Main Results:

• The proposed method yields identical cost-effectiveness estimates to models with merged states while preserving the original data.
• It allows for the testing of equality of costs and health consequences between states, justifying merging.
• The method enables the use of standard statistical techniques for model comparison.

Conclusions:

• State structures in cost-effectiveness models can be rigorously compared using standard methods by treating them as constrained versus unconstrained models.
• The approach was illustrated with applications in anti-depressant medication cost-effectiveness and coronary artery disease diagnostic test evaluation.