Nonparametric competing risks analysis using Bayesian Additive Regression Trees (BART)

TL;DR

This paper introduces a flexible Bayesian Additive Regression Trees (BART) approach for analyzing competing risks in time-to-event data, addressing complex relationships and model misspecification issues.

Contribution

It presents a novel BART-based method for competing risks analysis, improving prediction accuracy over traditional models and random survival forests.

Findings

BART outperforms standard regression techniques in simulation studies.

The method effectively captures nonlinearities and interactions.

Application to stem cell transplantation data demonstrates practical utility.

Abstract

Many time-to-event studies are complicated by the presence of competing risks. Such data are often analyzed using Cox models for the cause specific hazard function or Fine-Gray models for the subdistribution hazard. In practice regression relationships in competing risks data with either strategy are often complex and may include nonlinear functions of covariates, interactions, high-dimensional parameter spaces and nonproportional cause specific or subdistribution hazards. Model misspecification can lead to poor predictive performance. To address these issues, we propose a novel approach to flexible prediction modeling of competing risks data using Bayesian Additive Regression Trees (BART). We study the simulation performance in two-sample scenarios as well as a complex regression setting, and benchmark its performance against standard regression techniques as well as random survival forests. We illustrate the use of the proposed method on a recently published study of patients undergoing hematopoietic stem cell transplantation.