A Bayesian model averaging approach for cost-effectiveness analyses

Area of Science:

• Health Economics
• Biostatistics
• Clinical Trial Analysis

Background:

• Assessing technology cost-effectiveness relies on clinical trial data for costs and effects.
• Skewed and heavy-tailed cost distributions pose challenges for traditional probabilistic modeling.
• Accurate modeling of cost distribution tails is crucial for estimating population means.

Purpose of the Study:

• To develop a robust method for cost-effectiveness analysis (CEA) with complex cost data.
• To integrate model uncertainty into the analysis of cost data within CEA.
• To compare a novel Bayesian approach with existing semi-parametric methods.

Main Methods:

• Bayesian model averaging (BMA) was employed to handle uncertainty in cost data distributions.
• A set of plausible parametric models for costs was specified.
• Mean costs were estimated using a weighted average of posterior expectations, with weights from posterior model probabilities.

Main Results:

• The BMA approach effectively integrates uncertainty about cost data distributions.
• Results were compared against a semi-parametric method that makes fewer distributional assumptions.
• The study demonstrates a more realistic approach to modeling skewed cost data in CEA.

Conclusions:

• Bayesian model averaging offers a superior method for cost-effectiveness analysis when dealing with challenging cost data distributions.
• This approach enhances the reliability of cost-effectiveness assessments by accounting for model uncertainty.
• The findings support the use of BMA for more accurate health economic evaluations.