Nonparametric competing risks analysis using Bayesian Additive Regression Trees

Rodney Sparapani, Brent R. Logan, Robert McCulloch, Purushottam W. Laud

Research output: Contribution to journalArticle

1 Citation (Scopus)

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 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.

Original languageEnglish (US)
JournalStatistical Methods in Medical Research
DOIs
StateAccepted/In press - Jan 1 2019

Fingerprint

Regression Tree
Competing Risks
Bayes Theorem
Risk Analysis
Regression
Proportional Hazards Models
Hazard
Cause-specific Hazard
Grey Model
Benchmarking
Cox Model
Model Misspecification
Transplantation
Stem Cells
Hazard Function
Hematopoietic Stem Cell Transplantation
Nonlinear Function
Parameter Space
Covariates
High-dimensional

Keywords

  • Cumulative incidence
  • graft-versus-host disease (GVHD)
  • hematopoietic stem cell transplant
  • machine learning
  • nonproportional
  • treatment heterogeneity
  • variable selection

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability
  • Health Information Management

Cite this

Nonparametric competing risks analysis using Bayesian Additive Regression Trees. / Sparapani, Rodney; Logan, Brent R.; McCulloch, Robert; Laud, Purushottam W.

In: Statistical Methods in Medical Research, 01.01.2019.

Research output: Contribution to journalArticle

@article{995e8b00532545c4b73bbe278ceb3b9b,
title = "Nonparametric competing risks analysis using Bayesian Additive Regression Trees",
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 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.",
keywords = "Cumulative incidence, graft-versus-host disease (GVHD), hematopoietic stem cell transplant, machine learning, nonproportional, treatment heterogeneity, variable selection",
author = "Rodney Sparapani and Logan, {Brent R.} and Robert McCulloch and Laud, {Purushottam W.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1177/0962280218822140",
language = "English (US)",
journal = "Statistical Methods in Medical Research",
issn = "0962-2802",
publisher = "SAGE Publications Ltd",

}

TY - JOUR

T1 - Nonparametric competing risks analysis using Bayesian Additive Regression Trees

AU - Sparapani, Rodney

AU - Logan, Brent R.

AU - McCulloch, Robert

AU - Laud, Purushottam W.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - 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.

AB - 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.

KW - Cumulative incidence

KW - graft-versus-host disease (GVHD)

KW - hematopoietic stem cell transplant

KW - machine learning

KW - nonproportional

KW - treatment heterogeneity

KW - variable selection

UR - http://www.scopus.com/inward/record.url?scp=85060347965&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85060347965&partnerID=8YFLogxK

U2 - 10.1177/0962280218822140

DO - 10.1177/0962280218822140

M3 - Article

JO - Statistical Methods in Medical Research

JF - Statistical Methods in Medical Research

SN - 0962-2802

ER -