Regression analysis of incomplete medical cost data

Area of Science:

• Biostatistics
• Health Economics
• Survival Analysis

Background:

• Medical cost accumulation is a complex stochastic process, often observed incompletely due to limited time points and right censoring.
• Modeling these costs presents challenges due to the presence of death and censoring, complicating statistical inference.

Purpose of the Study:

• To develop regression models for analyzing medical cost accumulation influenced by time-dependent covariates.
• To address the complexities of death and censoring in cost accumulation data.
• To estimate the effects of covariates on both marginal and conditional means of cost accumulation.

Main Methods:

• Proposed regression models incorporating time-dependent covariates.
• Utilized generalized estimating equations (GEE) for longitudinal data.
• Applied inverse probability of censoring weighting (IPCW) technique for handling censoring.

Main Results:

• Developed consistent and asymptotically normal estimators for cost accumulation models.
• Variance estimators are simple and practical.
• Simulation studies confirmed the proposed inference procedures perform well.

Conclusions:

• The proposed statistical framework effectively models medical cost accumulation in the presence of death and censoring.
• Analysis of ovarian cancer data indicates disease aggressiveness impacts cost accumulation rates and lifetime costs.
• Less aggressive ovarian cancer patients incur lower initial costs but higher total lifetime costs due to increased longevity.