Recent advances in doubly-robust weighted ordinary least squares techniques for dynamic treatment regime estimation

TL;DR

This paper reviews the last decade of developments in doubly-robust weighted ordinary least squares (dWOLS) methods for estimating dynamic treatment regimes, highlighting extensions, applications, and practical challenges.

Contribution

It provides a comprehensive review of dWOLS advancements, including extensions to various treatment types and outcomes, and offers practical implementation guidance with R examples.

Findings

Extended dWOLS to binary, continuous, and multicategory treatments

Addressed challenges like model validation and variable selection

Provided R code for step-by-step implementation

Abstract

A dynamic treatment regime (DTR) is an approach to delivering precision medicine that uses patient characteristics to guide treatment decisions for optimal health outcomes. Numerous methods have been proposed for DTR estimation, including dynamic weighted ordinary least squares (dWOLS), a regression-based approach that affords double robustness to model misspecification within an easy to implement analytical framework. Initially, the dWOLS approach was developed under the assumptions of continuous outcomes and binary treatment decisions. Motivated by clinical research, subsequent theoretical advancements have extended the dWOLS framework to address binary, continuous and multicategory treatments across various outcome types, including binary, continuous, and survival-type. However, certain scenarios remain unexplored. This paper summarizes the last ten years of extension and application of the dWOLS method, providing a comprehensive and detailed review of the original dWOLS method and its extensions, as well as highlighting its diverse practical applications. We also explore studies that have addressed challenges associated with dWOLS implementation, such as model validation, variable selection, and handling measurement errors. Using simulated data, we present numerical illustrations along with step-by-step implementations in the \texttt{R} environment to facilitate a deeper understanding of dWOLS-based DTR estimation methodologies.