|
Activity Number:
|
184
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 1, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #320710
|
|
|
Title:
|
Bayesian Additive Regression Trees (BART) and Precision Medicine
|
|
Author(s):
|
Brent Logan* and Rodney Sparapani and Robert McCulloch and Purushottam Laud
|
|
Companies:
|
Medical College of Wisconsin and Medical College of Wisconsin and The University of Chicago and Medical College of Wisconsin
|
|
Keywords:
|
Bayesian Additive Regression Trees ;
Precision Medicine ;
Individualized Treatment Regimes
|
|
Abstract:
|
Individualized treatment regimes (ITR) can improve health outcomes by recognizing that patients may respond differently to treatment and assigning therapy with the highest predicted outcome for each individual. Flexible and efficient prediction models are needed as a basis for such ITR's to handle potentially complex interactions between patient factors and treatment. Bayesian Additive Regression Trees (BART) perform well in fitting nonparametric regression functions to continuous and binary responses, even with moderately high dimensional covariate spaces. We investigate optimal treatment strategies which utilize individualized predictions of patient outcomes from BART models. Posterior distributions of patient outcomes under each treatment are used to assign the treatment that optimizes the posterior mean outcome. We also describe how to approximate the optimal treatment policy with a clinically interpretable ITR, and quantify the expected outcome of such a policy. The proposed method is compared in simulation studies with existing methods on ITR's such as parametric outcome models and policy search methods like outcome weighted learning.
|
Authors who are presenting talks have a * after their name.
