Comment on "Dynamic treatment regimes: technical challenges and applications"

While optimal dynamic treatment regimes (DTRs) can be estimated without specification of a predictive model, a model-based approach, combined with dynamic programming and Monte Carlo integration, enables direct probabilistic comparisons between the outcomes under the optimal DTR and alternative (dynamic or static) treatment regimes. The Bayesian predictive approach also circumvents problems related to frequentist estimators under the nonregular estimation problem. However, the model-based approach is susceptible to misspecification, in particular of the "null-paradox" type, which is due to the model parameters not having a direct causal interpretation in the presence of latent individual-level characteristics. Because it is reasonable to insist on correct inferences under the null of no difference between the alternative treatment regimes, we discuss how to achieve this through a "null-robust" reparametrization of the problem in a longitudinal setting. Since we argue that causal inference can be entirely understood as posterior predictive inference in a hypothetical population without covariate imbalances, we also discuss how controlling for confounding through inverse probability of treatment weighting can be justified and incorporated in the Bayesian setting.

Precision medicine requires individualized treatment regime for subjects with different clinical characteristics. Machine learning methods have witnessed rapid progress in recent years, which can be employed to make individualized treatment regime in clinical practice. The idea of reinforcement learning method is to take action in response to the changing environment. In clinical medicine, this idea can be used to assign optimal regime to patients with distinct characteristics. In the field of statistics, reinforcement learning has been widely investigated, aiming to identify an optimal dynamic treatment regime (DTR). Q-learning is among the earliest methods to identify optimal DTR, which fits linear outcome models in a recursive manner. The advantage is its easy interpretation and can be performed in most statistical software. However, it suffers from the risk of misspecification of the linear model. More recently, some other methods not so heavily depend on model specification have been developed such as inverse probability weighted estimator and augmented inverse probability weighted estimator. This review introduces the basic ideas of these methods and shows how to perform the learning algorithm within R environment.

A dynamic treatment regime consists of a set of decision rules that dictate how to individualize treatment to patients based on available treatment and covariate history. A common method for estimating an optimal dynamic treatment regime from data is Q-learning which involves nonsmooth operations of the data. This nonsmoothness causes standard asymptotic approaches for inference like the bootstrap or Taylor series arguments to breakdown if applied without correction. Here, we consider the m-out-of-n bootstrap for constructing confidence intervals for the parameters indexing the optimal dynamic regime. We propose an adaptive choice of m and show that it produces asymptotically correct confidence sets under fixed alternatives. Furthermore, the proposed method has the advantage of being conceptually and computationally much simple than competing methods possessing this same theoretical property. We provide an extensive simulation study to compare the proposed method with currently available inference procedures. The results suggest that the proposed method delivers nominal coverage while being less conservative than alternatives. The proposed methods are implemented in the qLearn R-package and have been made available on the Comprehensive R-Archive Network (http://cran.r-project.org/). Analysis of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study is used as an illustrative example.

Ambiguous Dynamic Treatment Regimes: A Reinforcement Learning Approach

Recent developments in discovering dynamic treatment regimes (DTRs) have heightened the importance of deep reinforcement learning (DRL) which are used to recover the doctor’s treatment policies. However, existing DRL-based methods expose the following limitations: 1) supervised methods based on behavior cloning suffer from compounding errors; 2) the self-defined reward signals in reinforcement learning models are either too sparse or need clinical guidance; 3) only positive trajectories (e.g. survived patients) are considered in current imitation learning models, with negative trajectories (e.g. deceased patients) been largely ignored, which are examples of what not to do and could help the learned policy avoid repeating mistakes. To address these limitations, in this paper, we propose the adversarial cooperative imitation learning model, ACIL, to deduce the optimal dynamic treatment regimes that mimics the positive trajectories while differs from the negative trajectories. Specifically, two discriminators are used to help achieve this goal: an adversarial discriminator is designed to minimize the discrepancies between the trajectories generated from the policy and the positive trajectories, and a cooperative discriminator is used to distinguish the negative trajectories from the positive and generated trajectories. The reward signals from the discriminators are utilized to refine the policy for dynamic treatment regimes. Experiments on the publicly real-world medical data demonstrate that ACIL improves the likelihood of patient survival and provides better dynamic treatment regimes with the exploitation of information from both positive and negative trajectories.

Dynamic treatment regimes (DTRs) formalize medical decision-making as a sequence of rules for different stages, mapping patient-level information to recommended treatments. In practice, estimating an optimal DTR using observational data from electronic medical record (EMR) databases can be complicated by nonignorable missing covariates resulting from informative monitoring of patients. Since complete case analysis can provide consistent estimation of outcome model parameters under the assumption of outcome-independent missingness, Q-learning is a natural approach to accommodating nonignorable missing covariates. However, the backward induction algorithm used in Q-learning can introduce challenges, as nonignorable missing covariates at later stages can result in nonignorable missing pseudo-outcomes at earlier stages, leading to suboptimal DTRs, even if the longitudinal outcome variables are fully observed. To address this unique missing data problem in DTR settings, we propose 2 weighted Q-learning approaches where inverse probability weights for missingness of the pseudo-outcomes are obtained through estimating equations with valid nonresponse instrumental variables or sensitivity analysis. The asymptotic properties of the weighted Q-learning estimators are derived, and the finite-sample performance of the proposed methods is evaluated and compared with alternative methods through extensive simulation studies. Using EMR data from the Medical Information Mart for Intensive Care database, we apply the proposed methods to investigate the optimal fluid strategy for sepsis patients in intensive care units.

Tooth loss from periodontal disease is a major public health burden in the United States. Standard clinical practice is to recommend a dental visit every six months; however, this practice is not evidence-based, and poor dental outcomes and increasing dental insurance premiums indicate room for improvement. We consider a tailored approach that recommends recall time based on patient characteristics and medical history to minimize disease progression without increasing resource expenditures. We formalize this method as a dynamic treatment regime which comprises a sequence of decisions, one per stage of intervention, that follow a decision rule which maps current patient information to a recommendation for their next visit time. The dynamics of periodontal health, visit frequency, and patient compliance are complex, yet the estimated optimal regime must be interpretable to domain experts if it is to be integrated into clinical practice. We combine nonparametric Bayesian dynamics modeling with policy-search algorithms to estimate the optimal dynamic treatment regime within an interpretable class of regimes. Both simulation experiments and application to a rich database of electronic dental records from the HealthPartners HMO shows that our proposed method leads to better dental health without increasing the average recommended recall time relative to competing methods. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.

In recent sequential multiple assignment randomized trials, outcomes were assessed multiple times to evaluate longer-term impacts of the dynamic treatment regimes (DTRs). Q-learning requires a scalar response to identify the optimal DTR. Inverse probability weighting may be used to estimate the optimal outcome trajectory, but it is inefficient, susceptible to model mis-specification, and unable to characterize how treatment effects manifest over time. We propose modified Q-learning with generalized estimating equations to address these limitations and apply it to the M-bridge trial, which evaluates adaptive interventions to prevent problematic drinking among college freshmen. Simulation studies demonstrate our proposed method improves efficiency and robustness.

This Guide to Statistics and Methods provides an overview of using observational data and reinforcement learning strategies to develop dynamic treatment regimes for patients who are critically ill.

Sequential, multiple assignment, randomized trials (SMARTs) compare sequences of treatment decision rules called dynamic treatment regimes (DTRs). In particular, the Adaptive Treatment for Alcohol and Cocaine Dependence (ENGAGE) SMART aimed to determine the best DTRs for patients with a substance use disorder. While many authors have focused on a single pairwise comparison, addressing the main goal involves comparisons of >2 DTRs. For complex comparisons, there is a paucity of methods for binary outcomes. We fill this gap by extending the multiple comparisons with the best (MCB) methodology to the Bayesian binary outcome setting. The set of best is constructed based on simultaneous credible intervals. A substantial challenge for power analysis is the correlation between outcome estimators for distinct DTRs embedded in SMARTs due to overlapping subjects. We address this using Robins' G-computation formula to take a weighted average of parameter draws obtained via simulation from the parameter posteriors. We use non-informative priors and work with the exact distribution of parameters avoiding unnecessary normality assumptions and specification of the correlation matrix of DTR outcome summary statistics. We conduct simulation studies for both the construction of a set of optimal DTRs using the Bayesian MCB procedure and the sample size calculation for two common SMART designs. We illustrate our method on the ENGAGE SMART. The R package SMARTbayesR for power calculations is freely available on the Comprehensive R Archive Network (CRAN) repository. An RShiny app is available at https://wilart.shinyapps.io/shinysmartbayesr/.

The contributions of the review paper on the technical challenges and applications of dynamic treatment regimes are briefly discussed and potential extensions to continuous action spaces and high dimensional problems are indicated.

Appropriate analysis of observational data is our best chance to obtain answers to many questions that involve dynamic treatment regimes. This paper describes a simple method to compare dynamic treatment regimes by artificially censoring subjects and then using inverse probability weighting (IPW) to adjust for any selection bias introduced by the artificial censoring. The basic strategy can be summarized in four steps: 1) define two regimes of interest, 2) artificially censor individuals when they stop following one of the regimes of interest, 3) estimate inverse probability weights to adjust for the potential selection bias introduced by censoring in the previous step, 4) compare the survival of the uncensored individuals under each regime of interest by fitting an inverse probability weighted Cox proportional hazards model with the dichotomous regime indicator and the baseline confounders as covariates. In the absence of model misspecification, the method is valid provided data are available on all time-varying and baseline joint predictors of survival and regime discontinuation. We present an application of the method to compare the AIDS-free survival under two dynamic treatment regimes in a large prospective study of HIV-infected patients. The paper concludes by discussing the relative advantages and disadvantages of censoring/IPW versus g-estimation of nested structural models to compare dynamic regimes.

Inference for parameters associated with optimal dynamic treatment regimes is challenging as these estimators are nonregular when there are non-responders to treatments. In this discussion, we comment on three aspects of alleviating this nonregularity. We first discuss an alternative approach for smoothing the quality functions. We then discuss some further details on our existing work to identify non-responders through penalization. Third, we propose a clinically meaningful value assessment whose estimator does not suffer from nonregularity.

Purpose of ReviewChronic Kidney Disease-Mineral and Bone Disorder (CKD-MBD) is a complex condition impacting patients with kidney failure and characterized by inter-related features such as hyperparathyroidism, hyperphosphatemia, and hypocalcemia. Current treatments include active vitamin D sterols, calcimimetics, and phosphate binders alone and in combination. However, identifying optimal treatment is challenged by interdependency among CKD-MBD features, requiring new approaches to understand dynamic systems. In this review, we discuss challenges and opportunities for a more integrated view of CKD-MBD care.Recent FindingsFew clinical studies in CKD-MBD care have incorporated a dynamic understanding of the disorder and its treatment. Dynamic treatment regime methods are an evolving area of artificial intelligence (AI) that offer a promising approach for modeling and understanding CKD-MBD care. Efforts to date have included dynamic systems and quantitative systems pharmacology-based models to simulate the impact of alternative treatment regimes. Additional studies utilizing dynamic treatment regime approaches may help improve knowledge gaps in CKD-MBD care.SummaryAlthough preliminary research highlights the potential of dynamic treatment regime approaches in optimizing CKD-MBD management, further investigation and clinical validation are necessary to fully harness this approach for improving patient outcomes.

Healthcare applications pose significant challenges to existing reinforcement learning (RL) methods due to implementation risks, limited data availability, short treatment episodes, sparse rewards, partial observations, and heterogeneous treatment effects. Despite significant interest in using RL to generate dynamic treatment regimes for longitudinal patient care scenarios, no standardized benchmark has yet been developed. To fill this need we introduce Episodes of Care (EpiCare), a benchmark designed to mimic the challenges associated with applying RL to longitudinal healthcare settings. We leverage this benchmark to test five state-of-the-art offline RL models as well as five common off-policy evaluation (OPE) techniques. Our results suggest that while offline RL may be capable of improving upon existing standards of care given sufficient data, its applicability does not appear to extend to the moderate to low data regimes typical of current healthcare settings. Additionally, we demonstrate that several OPE techniques standard in the the medical RL literature fail to perform adequately on our benchmark. These results suggest that the performance of RL models in dynamic treatment regimes may be difficult to meaningfully evaluate using current OPE methods, indicating that RL for this application domain may still be in its early stages. We hope that these results along with the benchmark will facilitate better comparison of existing methods and inspire further research into techniques that increase the practical applicability of medical RL.