A Bayesian Modeling Framework for Health Care Resource Use and Costs in Trial-Based Economic Evaluations

Area of Science:

• Health economics
• Biostatistics
• Clinical trial analysis

Background:

• Trial-based economic evaluations use individual-level data for effectiveness and costs.
• Health care resource use (HRU) data are essential for cost computation but often incomplete.
• Existing methods for missing HRU data rely on potentially unjustified assumptions, like imputing zero use.

Purpose of the Study:

• To present a general Bayesian framework for analyzing partially observed HRU data in economic evaluations.
• To accommodate data complexities like excess zeros, skewness, and missingness.
• To quantify the impact of missingness uncertainty on economic evaluation results.

Main Methods:

• Developed a flexible Bayesian framework for analyzing partially observed HRU data.
• Incorporated methods to handle data complexities including excess zeros, skewness, and missingness.
• Compared the proposed Bayesian approach with standard analyses using aggregated cost variables and ad hoc imputation.

Main Results:

• The Bayesian framework effectively handles partially observed HRU data and quantifies missingness uncertainty.
• Results differ significantly based on missingness assumptions and data aggregation levels.
• Analyses at the most disaggregated level, using all available trial data, are preferred over restrictive ad hoc imputation.

Conclusions:

• A comprehensive modeling approach is vital for handling partially observed HRU data in economic evaluations.
• Bayesian frameworks offer strategic advantages for modeling HRU data due to their flexibility and ability to incorporate uncertainty.
• Adopting disaggregated analyses within a Bayesian framework enhances the reliability and validity of economic evaluation findings.