Nonparametric competing risks analysis using Bayesian Additive Regression Trees. Stat Methods Med Res 2020 Jan;29(1):57-77
Date
01/08/2019Pubmed ID
30612519Pubmed Central ID
PMC6954340DOI
10.1177/0962280218822140Scopus ID
2-s2.0-85060347965 (requires institutional sign-in at Scopus site) 18 CitationsAbstract
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 and Gray models for the subdistribution hazard. In practice, regression relationships in competing risks data 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: 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.
Author List
Sparapani R, Logan BR, McCulloch RE, Laud PWAuthors
Purushottam W. Laud PhD Adjunct Professor in the Data Science Institute department at Medical College of WisconsinBrent R. Logan PhD Director, Professor in the Data Science Institute department at Medical College of Wisconsin
Rodney Sparapani PhD Associate Professor in the Data Science Institute department at Medical College of Wisconsin
MESH terms used to index this publication - Major topics in bold
Bayes TheoremBenchmarking
Computer Simulation
Graft vs Host Disease
Hematopoietic Stem Cell Transplantation
Humans
Incidence
Machine Learning
Regression Analysis
