Semiparametric analysis of clustered interval-censored survival data using soft Bayesian additive regression trees...

Area of Science:

• Biostatistics
• Survival Analysis
• Machine Learning

Background:

• Traditional survival models struggle with complex clustered and interval-censored data.
• Accurate survival prediction is crucial for understanding disease progression and treatment efficacy.

Purpose of the Study:

• To develop a flexible and robust semiparametric model for clustered, interval-censored survival data.
• To improve survival prediction accuracy by incorporating complex covariate effects and data dependencies.

Main Methods:

• Proposed a novel semiparametric hazards regression model using soft Bayesian additive regression trees (SBART).
• Modeled the hazard function as a product of a parametric baseline and a nonparametric SBART component.
• Implemented the methodology using a data augmentation scheme compatible with Bayesian backfitting algorithms.

Main Results:

• The SBART-based model demonstrated excellent predictive accuracy for clustered, interval-censored survival data.
• The method effectively handles unknown covariate effects, clustering, and interval censoring.
• Simulations and a prostate cancer surgery dataset analysis validated the model's practical implementation and advantages.

Conclusions:

• The proposed robust semiparametric model offers a flexible framework for complex survival data.
• This approach is applicable to various censoring types (left, right, interval) and high-dimensional data.
• The method enhances survival prediction in studies with inherent data clustering and complex associations.