A two-part model for censored medical cost data

Area of Science:

• Biostatistics
• Health Economics
• Medical Data Analysis

Background:

• Medical cost data often exhibit excess zeros and high skewness, complicating analysis.
• Right-censoring due to incomplete follow-up can be informative, posing challenges for cost data analysis.

Purpose of the Study:

• To develop a robust statistical model for analyzing right-censored medical cost data with informative censoring.
• To apply inverse probability weighting to a two-part model for accurate cost estimation.
• To introduce efficiency augmentation and model-checking techniques for improved analysis.

Main Methods:

• Utilized a two-part model combined with inverse probability weighting for right-censored medical costs.
• Introduced functionals based on intermediate cost history with efficiency augmentation.
• Developed a model-checking technique using cumulative residuals.
• Conducted simulation studies to assess finite sample performance.

Main Results:

• The proposed inverse probability weighting method effectively handles informative censoring in two-part models.
• Simulation studies demonstrated the good finite sample performance of the developed statistical techniques.
• The method was successfully illustrated using cardiovascular disease-related Medicare costs.

Conclusions:

• The proposed inverse probability weighting approach provides a reliable method for analyzing complex medical cost data.
• The techniques enhance the analysis of skewed, censored health economic data, particularly for chronic diseases.
• This methodology offers practical tools for researchers and policymakers dealing with health cost data.