iqLearn: Interactive Q-Learning in R.

Area of Science:

• * Biostatistics
• * Health Informatics
• * Clinical Trial Design

Background:

• * Chronic illness management requires adaptive, data-driven dynamic treatment regimes.
• * Traditional methods like Q-learning face challenges with complex data transformations.
• * Optimizing long-term patient outcomes necessitates robust treatment strategy estimation.

Purpose of the Study:

• * To introduce Interactive Q-learning (IQ-learning) as an alternative to Q-learning for dynamic treatment regimes.
• * To demonstrate the application of IQ-learning for estimating optimal treatment policies.
• * To provide an R package (iqLearn) for implementing both IQ-learning and Q-learning.

Main Methods:

• * IQ-learning models smooth, monotone data transformations, simplifying model fitting.
• * Q-learning, a common method, requires modeling nonsmooth, nonmonotone transformations.
• * The iqLearn R package facilitates the implementation of these algorithms.

Main Results:

• * IQ-learning provides a viable alternative to Q-learning for dynamic treatment regime estimation.
• * The iqLearn package enables practical application of these methods.
• * A simulated two-stage body mass index reduction trial illustrates the estimation process.

Conclusions:

• * IQ-learning offers a more accessible method for developing data-driven dynamic treatment strategies.
• * The iqLearn package supports researchers in applying advanced statistical methods for chronic disease management.
• * This approach can lead to more favorable average clinical outcomes for patients over time.