Pseudo-value regression of clustered multistate current status data with informative cluster sizes

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Multistate current status data involves complex censoring and potential bias from informative cluster sizes.
• Existing methods may fail to account for the relationship between transition outcomes and cluster sizes, leading to inaccurate inferences.
• Periodontal disease studies often generate such complex clustered data.

Purpose of the Study:

• To develop a statistical approach for analyzing clustered multistate current status data with informative cluster sizes.
• To estimate covariate effects on state occupation probabilities in the presence of informative cluster sizes.
• To address potential biases in statistical inference caused by unadjusted informative cluster sizes.

Main Methods:

• Extension of the pseudo-value approach for clustered multistate current status data.
• Computation of marginal state occupation probabilities using nonparametric regression.
• Reweighting estimating equations with cluster size functions to adjust for informativeness.

Main Results:

• Simulation studies demonstrate the properties of the proposed pseudo-value regression under various informativeness scenarios.
• The method effectively adjusts for informative cluster sizes, reducing inferential bias.
• Validation through application to a periodontal disease dataset.

Conclusions:

• The proposed pseudo-value extension provides a robust method for analyzing clustered multistate current status data with informative cluster sizes.
• This approach enhances the accuracy of covariate effect estimation in complex health studies.
• The method is applicable to real-world clinical data, such as periodontal disease progression.