Estimating optimal treatment regimes from a classification perspective

Abstract

A treatment regime maps observed patient characteristics to a recommended treatment. Recent technological advances have increased the quality, accessibility, and volume of patient-level data; consequently, there is a growing need for powerful and flexible estimators of an optimal treatment regime that can be used with either observational or randomized clinical trial data. We propose a novel and general framework that transforms the problem of estimating an optimal treatment regime into a classification problem wherein the optimal classifier corresponds to the optimal treatment regime. We show that commonly employed parametric and semi-parametric regression estimators, as well as recently proposed robust estimators of an optimal treatment regime can be represented as special cases within our framework. Furthermore, our approach allows any classification procedure that can accommodate case weights to be used without modification to estimate an optimal treatment regime. This introduces a wealth of new and powerful learning algorithms for use in estimating treatment regimes. We illustrate our approach using data from a breast cancer clinical trial.

Dynamic Treatment Regimes

“…As highlighted by Qian and Murphy [56], the one stage decision making problem has a close connection with classification. Subsequently, methods based on classification [57, 58] have been proposed for estimating the decision rule. Other work in the one-stage decision setting include Cai et al [59] and Imai and Ratkovicz [60].…”

Estimation of optimal dynamic treatment regimes

“…We give a version of the Q-learning algorithm in the following section. Value maximization methods are based on forming an estimator of V(d) = E d Y and then directly maximizing this estimator over d in some class of DTRs, say D. Value maximization methods include outcome weighted learning [9, 14], augmented value maximization [10, 12, 13], and structural mean models [7]. We give a version of outcome weighted learning in the subsequent section.…”

“…A more direct approach is to postulate an estimator of V(d) = E d Y, say $\hat{V}false(\mathrm{normald}false)$, and then estimate the optimal DTR by searching for $\hat{d}={\mathrm{argmax}}_{\mathrm{normald}\in \mathrm{normalD}}\hat{V}false(\mathrm{normald}false)$, where D is a prespecified class of DTRs. Estimators of this form are called value maximization methods or policy search methods and have received a great deal of attention recently [7, 9, 10, 12–14]. A potential advantage of value maximization methods is that, because they need not rely on models for the Q-function, they may be more robust to model specification.…”

On estimation of the optimal treatment regime with the additive hazards model

“…Instead of directly maximizing the value function, Zhao et al (2012) proposed to estimate the optimal treatment regime by outcome weighted support vector machines in a weighted classification framework. Zhang et al (2012b) proposed a general classification framework for estimating the optimal treatment regime. These studies mainly focus on uncensored data.…”