Quantitative Medicine Forum: Inference for Optimal Dynamic Treatment Regimes Using an Adaptive m-out-of-n Bootstrap Scheme

Speaker:
Dr. Bibhas Chakraborty
Assistant Professor of Biostatistics, Mailman School of Public Health, Columbia University

Host:
Dr. Benjamin Haaland
Assistant Professor, Centre for Quantitative Medicine, Office of Clinical Sciences, Duke-NUS Graduate Medical School

Date:
Thursday, June 28, 2012

Time:
12.30pm to 1.30pm 
(Lunch will be served at 12.00pm)

Venue:
Amphitheatre, Level 2
Duke-NUS Graduate Medical School
8 College Road, Singapore 169857
(opposite Singapore General Hospital, Block 6/7)

Contact Person:
Ms Megan Pooh, Centre for Quantitative Medicine, Office of Clinical Sciences, Duke-NUS
Tel: 6601 1719 or Email: megan.pooh@duke-nus. edu.sg

Synopsis:
A dynamic treatment regime (DTR) consists of a set of decision rules that dictate how to personalize treatment to an individual patient based on available treatment and covariate history. A common method for estimating an optimal DTR from patient data is Q-learning which involves non-smooth operations of the data. This non-smoothness causes standard asymptotic approaches for inference like the usual bootstrap or Taylor series arguments to break down if applied without correction. In this talk, we will discuss a novel adaptive m-out-of-n bootstrap scheme for constructing confidence intervals for the parameters indexing the optimal DTR. The proposed method produces asymptotically correct confidence intervals, and has the advantage of being conceptually simple. We will provide an extensive simulation study to compare the finite-sample performance of our proposed method with the currently existing inference procedures in this area. Analysis of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study will be used as an illustrative example of our methodology.

Biography:
Dr. Bibhas Chakraborty, PhD, is an Assistant Professor of Biostatistics in the Mailman School of Public Health, Columbia University. His primary research interest lies in statistical methods for constructing evidence-based, multistage, personalized interventions, often called dynamic treatment regimes. His other research interests include design of multi-component intervention trials and adaptive clinical trials, causal inference, and statistical machine learning and data mining. Prior to joining Columbia, he completed his Ph.D. in 2009 from the Department of Statistics, University of Michigan, under the supervision of Dr. Susan A. Murphy, one of the pioneers of the field of dynamic treatment regimes. In 2011, Dr. Chakraborty was awarded the Calderone Research Award for Junior Faculty by the Mailman School of Public Health, Columbia University.