Semiparametric Latent Class Analysis of Recurrent Event Data

Area of Science:

• Biostatistics
• Epidemiology
• Chronic Disease Research

Background:

• Recurrent events data are crucial for understanding chronic disease progression.
• Existing latent class methods for recurrent events data often rely on strong parametric assumptions and face algorithmic challenges.
• Population heterogeneity in disease trajectories requires advanced analytical techniques.

Purpose of the Study:

• To develop a flexible semiparametric latent class analysis for recurrent events data.
• To address limitations of existing methods by avoiding strong parametric assumptions.
• To provide a robust and efficient statistical framework for analyzing complex disease progression patterns.

Main Methods:

• Latent class analysis (LCA) applied to recurrent events data.
• Flexible semiparametric multiplicative modeling of event intensities.
• Novel adaptation of the conditional score technique for robust estimation.
• Utilizing multiplicative intensity modeling characteristics for stable computation.

Main Results:

• A robust and efficient estimation procedure for latent class analysis of recurrent events data.
• Demonstrated stable and efficient implementation using existing computational routines.
• Theoretical underpinnings and satisfactory finite sample performance confirmed through simulations.
• Successful application to a real-world Alzheimer's disease research dataset.

Conclusions:

• The proposed semiparametric latent class analysis offers a powerful and flexible tool for recurrent events data.
• The method effectively captures population heterogeneity in disease trajectories.
• This approach enhances the analysis of chronic disease progression and provides practical utility in research settings.