Structural and parameter uncertainty in Bayesian cost-effectiveness models.

Area of Science:

• Health Economics
• Biostatistics
• Decision Modeling

Background:

• Health economic decision models face uncertainty in parameters and structure.
• Bayesian frameworks can integrate prior evidence and data.
• Markov models are used for cost-effectiveness analysis, e.g., for implantable cardioverter defibrillators.

Purpose of the Study:

• To formally incorporate parameter uncertainty in cost-effectiveness estimates using Bayesian methods.
• To extend Bayesian methods to account for uncertainty in model structure selection.
• To demonstrate efficient computation using WinBUGS software.

Main Methods:

• Utilized Markov chain Monte Carlo (MCMC) posterior simulation for parameter uncertainty.
• Employed pseudo-marginal likelihood and deviance information criterion (DIC) for model structure uncertainty.
• Applied Bayesian model averaging to combine evidence from competing models.

Main Results:

• Parameter uncertainty was formally incorporated into cost and effectiveness estimates.
• Model structure uncertainty was addressed by weighting posterior distributions.
• Efficient calculation methods were demonstrated in WinBUGS.

Conclusions:

• Bayesian approaches effectively manage uncertainty in health economic models.
• The proposed methods enhance the robustness of cost-effectiveness analyses.
• WinBUGS provides an efficient platform for these advanced Bayesian analyses.