Title: Estimation of optimal treatment regimes with censored time-to-event outcome: A classification perspective
Authors: Marie Davidian - North CArolina State University (United States) [presenting]
Abstract: Clinicians make a series of decisions at key points over the course of a patient's disease or disorder based on a synthesis of evolving information on the patient. A treatment regime is a sequence of decision rules mapping a patient's history to the set of feasible treatment options and thus formalizes this process. An optimal regime is one leading to the most beneficial outcome on average if used to make treatment decisions for the patient population. In many chronic disease contexts, the outcome is a possibly censored time to an event. We describe a method for estimation of an optimal regime with such outcomes, which is based on maximizing a locally efficient, doubly robust, augmented inverse probability weighted estimator for average outcome over a class of regimes. This optimization can be cast as a classification problem, which allows well-known methodology for classification to be exploited in a backward iterative algorithm. The performance of the method is demonstrated.