Bayesian Cost-Effectiveness Analysis Using Individual-Level Data is Sensitive to the Choice of Uniform Priors on the...

Area of Science:

• Health Economics
• Biostatistics
• Decision Science

Background:

• Bayesian cost-effectiveness analysis (CEA) necessitates prior distributions for parameter estimation.
• Log-Normal distributions are often used for modeling costs in CEA.
• Wide Uniform priors on log-scale standard deviations of costs are common but their impact is unclear.

Purpose of the Study:

• To explore the impact of Uniform priors on cost standard deviations in log-normally distributed cost data within Bayesian CEA.
• To assess how prior choices affect CEA conclusions when costs are log-normally distributed.

Main Methods:

• Individual-level cost-utility data from a randomized controlled trial were analyzed.
• Costs and quality-adjusted life years (QALYs) were modeled using Log-Normal and Beta distributions, respectively.
• Uniform priors with varying upper bounds were applied to log-scale standard deviations, with comparisons to other distributional assumptions and a simulation study.

Main Results:

• The selection of Uniform priors on log-cost standard deviations can significantly alter cost estimates in Log-Normal models.
• These fluctuations may influence the determination of an intervention's cost-effectiveness.
• The presence of zero values in cost data appears to exacerbate these effects.

Conclusions:

• Bayesian CEA results can be sensitive to the upper bounds of Uniform priors for log-cost standard deviations in Log-Normal models.
• Caution is advised when employing Uniform distributions with large upper bounds, particularly with zero-valued cost data.