Joint latent class trees: A tree-based approach to modeling time-to-event and longitudinal data.

Area of Science:

• Biostatistics
• Statistical modeling
• Machine learning

Background:

• Traditional joint latent class models are often parametric, computationally intensive, and limited to time-invariant covariates.
• This restricts their ability to fully capture complex relationships in longitudinal and time-to-event data.

Purpose of the Study:

• To propose a novel semiparametric, tree-based joint latent class model (joint latent class tree).
• To overcome the limitations of existing parametric joint latent class models, particularly regarding computational cost and covariate flexibility.

Main Methods:

• Developed a tree-based approach for joint latent class modeling.
• Incorporated time-varying covariates in all modeling components, including survival risks and latent class memberships.
• Utilized simulated data and the PAQUID dataset for validation.

Main Results:

• The proposed joint latent class tree model demonstrated significant prognostic value by effectively utilizing time-varying covariates.
• Achieved superior prediction performance compared to the traditional joint latent class model.
• Showcased orders-of-magnitude speedup in computation time.

Conclusions:

• The joint latent class tree offers a computationally efficient and flexible alternative for joint analysis of longitudinal and time-to-event data.
• Its ability to incorporate time-varying covariates enhances predictive accuracy and prognostic capabilities.
• This method provides a valuable advancement for statistical modeling in various scientific domains.