Event history graphs for censored survival data

Area of Science:

• Biostatistics
• Survival Analysis
• Data Visualization

Background:

• Traditional survival analysis often struggles to incorporate complex time-varying covariate data effectively.
• Visualizing the interplay between covariates and survival outcomes is crucial for accurate interpretation.

Purpose of the Study:

• To develop a compact graphical device that integrates survival data and time-varying covariate information.
• To enhance the visualization of Kaplan-Meier estimators with simultaneous covariate behavior display.

Main Methods:

• A novel graphical method is proposed, incorporating the Kaplan-Meier estimator for right-censored data.
• Time-dependent covariate levels undergo dimension reduction or smoothing to form continuous covariate functions.
• These functions are plotted as horizontal bars representing subject lifetimes, with censored data also incorporated.

Main Results:

• The union of horizontal bars visually reconstructs the Kaplan-Meier estimator of the survival function.
• The graphical device allows for a simultaneous display of covariate behavior and individual subject lifetimes.
• The method was implemented using a new S-plus function and validated through several applications.

Conclusions:

• The proposed graphical device offers a powerful and intuitive way to combine survival and time-varying covariate information.
• This method facilitates a deeper understanding of how covariates influence survival outcomes.
• The tool enhances the analysis of complex survival data in various scientific fields.