ROC-Guided Survival Trees and Ensembles

TL;DR

This paper introduces a new framework for survival trees and ensembles that handle time-dependent covariates, guided by ROC-based criteria, and demonstrates improved prediction accuracy through extensive simulations and an AIDS study.

Contribution

The paper presents a novel ROC-guided approach for constructing survival trees and ensembles that incorporate time-dependent covariates and decision-theoretic criteria.

Findings

Enhanced prediction accuracy demonstrated in simulations

Effective handling of time-dependent covariates in survival analysis

Ensemble method reduces instability compared to single trees

Abstract

Tree-based methods are popular nonparametric tools in studying time-to-event outcomes. In this article, we introduce a novel framework for survival trees and ensembles, where the trees partition the dynamic survivor population and can handle time-dependent covariates. Using the idea of randomized tests, we develop generalized time-dependent Receiver Operating Characteristic (ROC) curves for evaluating the performance of survival trees. The tree-building algorithm is guided by decision-theoretic criteria based on ROC, targeting specifically for prediction accuracy. To address the instability issue of a single tree, we propose a novel ensemble procedure based on averaging martingale estimating equations, which is different from existing methods that average the predicted survival or cumulative hazard functions from individual trees. Extensive simulation studies are conducted to examine the performance of the proposed methods. We apply the methods to a study on AIDS for illustration.