Ensemble Methods for Survival Data with Time-Varying Covariates

• Activity Number: 178 - Recent Development on the Analysis of Time-to-Event Data
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Lifetime Data Science Section
• Author(s): Weichi Yao* and Halina Frydman and Jeffrey S. Simonoff and Denis Larocque
• Companies: New York University and New York University and New York University and HEC Montreal
• Keywords: Continuous-time survival data; Ensemble methods; Time-varying covariates; Proportional hazards setting; Non-proportional hazards setting; Dynamic predictions

Abstract

We propose two new survival forests for left-truncated and right-censored data, which allow for time-varying covariates. They are generalizations of random survival forest and conditional inference forest - the traditional survival forests for right-censored data with time-invariant invariant covariates. We investigate the properties of these new forests, as well as that of the recently proposed transformation forest, and compare their performances with that of the Cox model via a comprehensive simulation study. In particular, we study the forests under the proportional hazards setting as well as the non-proportional hazards setting, where the forests based on log-rank splitting tend to perform worse than does the transformation forest. We provide guidelines for choosing among the considered forest methods. We also discuss the potential for these methods to provide dynamic updating of predictions as covariates change over time.