Activity Number:
|
98
- New Developments in Bayesian Additive Regression Trees
|
Type:
|
Invited
|
Date/Time:
|
Monday, July 30, 2018 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Bayesian Statistical Science
|
Abstract #326949
|
Presentation
|
Title:
|
Individualized Treatment for Time-To-Event Outcomes with BART
|
Author(s):
|
Brent Logan*
|
Companies:
|
Medical College of Wisconsin
|
Keywords:
|
Bayesian Additive Regression Trees;
Bayesian nonparametric;
ensemble model;
ITR;
Personalized Medicine;
Precision Medicine
|
Abstract:
|
Individualized treatment rules (ITR) can improve health outcomes by recognizing that patients may respond differently to treatment and assigning therapy with the most desirable predicted outcome for each individual. Flexible and efficient prediction models are desired as a basis for such ITRs to handle complex interactions between patient factors and treatment. Bayesian Additive Regression Trees (BART) perform well in fitting nonparametric regression functions even with many covariates, and are computationally efficient for practical use. We propose an ITR strategy for time-to-event outcomes which utilizes individualized predictions of patient survival functions from nonparametric BART survival models. Posterior distributions of patient survival functions under each treatment are used to assign the treatment that maximizes an expected posterior utility. This strategy provides posterior uncertainty quantification of patient specific treatment decisions as well as the population wide value of an ITR and the Optimal ITR. We illustrate this method in a study of patients undergoing hematopoietic cell transplantation.
|
Authors who are presenting talks have a * after their name.