Estimating Optimal Treatment Regimes from a Classification Perspective

By Baqun Zhang, Anastasios A. Tsiatis, Marie Davidian, Eric B. Laber. 2012

279 citations recorded by the source corpus.

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. Copyright © 2012 John Wiley & Sons, Ltd.

Subfields: Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Learning When-to-Treat Policies (2019, background): “…led to a surge in interest in methods that leverage observational data to help inform how and to whom interventions should be applied (Manski 2004; Zhang et al. 2012; Zhao et al. 2012; Dudík et al. 2014; Swaminathan and Joachims 2015; Elmachtoub and Grigas 2017; Kallus and Zhou 2018; Kitagawa and…”
Efficient Policy Learning (2017, methodology): “, 2011; Swaminathan and Joachims, 2015), outcome-weighted learning (Chen et al., 2016; Qian and Murphy, 2011; Zhang et al., 2012; Zhao et al., 2012; Zhou et al., 2015), and empirical welfare maximization (Kitagawa and Tetenov, 2015).”
Quasi-oracle estimation of heterogeneous treatment effects (2017, background): “Another research direction that also use ideas from semiparametrics to estimate complex objects is centered on estimating optimal treatment allocation rules (Athey and Wager, 2017; Dud́ık et al., 2011; Luedtke and van der Laan, 2016; Zhang et al., 2012; Zhao et al., 2012).”
STATISTICAL INFERENCE FOR THE MEAN OUTCOME UNDER A POSSIBLY NON-UNIQUE OPTIMAL TREATMENT STRATEGY (2016, methodology): “Zhang et al. (2012a) make this assumption in the regularity conditions in their Web Appendix A when they consider an analogous empirical process term in deriving the standard error of an estimate of the optimal value in a restricted class.”
Faster Rates for Policy Learning (2017, background): “Exciting developments in policy learning over the last several years include outcome weighted learning [2, 3, 4], ensemble techniques [5], and empirical risk minimizers (ERMs) [1], to name a few.”
New Statistical Learning Methods for Estimating Optimal Dynamic Treatment Regimes (2015, methodology): “Since the proposed methods do not model the relationship between outcomes and DTRs, they may be more robust to model misspecification than statistical modeling alternatives such as Q-learning (Zhang et al. 2012a,b).”
Doubly Robust Learning for Estimating Individualized Treatment with Censored Data (2014, unknown): “Alternative choices have been proposed for continuous outcomes (Zhang et al., 2012b; Rubin & van der Lann, 2012), which can be generalized to the censoring data setup.”
Incorporating Patient Preferences into Estimation of Optimal Individualized Treatment Rules (2017, background): “The first three assumptions are standard (Robins et al., 2000; Zhang et al., 2012), whereas (C4) can be satisfied provided the conditions of Lemma (1) hold; nevertheless, this assumption can be weakened to A ⊥ H ∣∣X, W at the expense of postulating a model for the conditional mean of (H)Y and (H)Z g”
Precision Medicine. (2019, methodology): “To mitigate misspecification, one can use a flexible class of models to estimate Q(x, a), e.g., trees (Taylor et al 2015; Zhang et al 2012b), boosting (Kang et al 2014), generalized additive models (Moodie et al 2014), or non-linear basis expansions (Qian and Murphy 2011).”
Interactive Q-learning for Quantiles (2017, methodology): “Using the potential outcomes framework (Rubin 1974; Rosenbaum and Rubin 1983), Zhang et al. (2012c) developed methods for estimating quantiles of the potential outcomes from observational data.”
On Optimal Treatment Regimes Selection for Mean Survival Time (2014, unknown): “In addition, [8] proposed a general framework for estimating the optimal treatment regimes from a classification perspective.”
Offline Multi-Action Policy Learning: Generalization and Optimization (2018, background): “Note that since finding an optimal classification tree is generally intractable (Bertsimas and Dunn (2017)), and since classification can be reduced to a special case of policy learning5 (Zhang et al. (2012), Zhao et al. (2012)), policy learning with trees in general is intractable as well.”
Interpretable Dynamic Treatment Regimes (2016, methodology): “, 2013; Zhou and Kosorok, 2016); and tree-based methods (Zhang et al., 2012; Laber and Zhao, 2015; Zhang et al., 2015; Doove et al., 2015).”
Interactive Q-learning for Probabilities and Quantiles (2014, methodology): “Using the potential outcomes framework (Rubin, 1974; Rosenbaum and Rubin, 1983), Zhang et al. (2012c) develop methods for estimating quantiles of the potential outcomes from observational data.”
Q-Learning: Theory and Applications (2020, methodology): “In addition to problems of statistical inference, developments for finite Q-learning in statistics have included extensions to nonlinear model classes (Zhao et al. 2009, Laber et al. 2014a, Moodie et al. 2014, Zhang et al. 2015, Xu et al. methods (Zhang et al. 2012, Zhang et al. 2015).”
Q- and A-learning Methods for Estimating Optimal Dynamic Treatment Regimes (2012, background): “Related work includes Thall, Millikan, and Sung (2000), Thall, Sung, and Etsey (2002), Robins (2004), Moodie, Richardson, and Stephens (2007), Thall et al. (2007), Robins, Orellana, and Rotnitzky (2008), Almirall, Ten Have, and Murphy (2010) and Orellana, Rotnitzky, and Robins (2010).”