Discrete-time competing-risks regression with or without penalization

Area of Science:

• Biostatistics
• Survival Analysis
• Computational Statistics

Background:

• Time-to-event data analysis often assumes continuous failure times.
• Discrete failure-time data, arising from inherent discreteness or measurement imprecision, presents analytical challenges.
• Existing methods may not adequately handle discrete-time data with competing risks.

Purpose of the Study:

• To introduce a novel estimation procedure for discrete-time survival analysis incorporating competing events.
• To provide a flexible framework that integrates with existing regularized regression and feature screening methods.
• To demonstrate the utility of the proposed method in a real-world clinical setting.

Main Methods:

• Development of a new estimation procedure for discrete-time survival data with competing risks.
• Integration with regularized regression and feature screening techniques.
• Validation through a comprehensive simulation study and application to intensive care unit (ICU) length of stay data.

Main Results:

• The proposed method effectively handles discrete-time survival data with competing events.
• The approach allows for straightforward application of advanced regression and feature selection techniques.
• Successful estimation of an ICU length of stay model with competing risks (discharge, transfer, death).

Conclusions:

• The new procedure offers a significant advantage for discrete-time survival analysis with competing risks.
• The method enhances the applicability of regularized regression and feature screening in this domain.
• The available Python package, PyDTS, facilitates the practical implementation of this advanced survival analysis technique.