HC Dynamic Treatment Regimes

Clinical management of a chronic illness requires the continued adjustment of treatment in response to the evolving health status of the patient. Dynamic treatment regimes operationalize this process as a sequence of decision rules. Each decision rule corresponds to a time point at which current treatment may or must be adjusted; each rule takes as input an up-to-date patient history and give as a output a recommended next treatment. Recently, there has been vigorous research on the estimation of and inference for optimal dynamic treatment regimes form data collected in an observational study or sequentially randomized clinical trials. Nevertheless, there are a large number of important open problems in this area. In broad strokes, we need methodology for:
* Estimation of and inference for an optimal dynamic treatment regime where there are an infinite number of decisions (i.e., the infinite horizon problem).
* Robust estimation.
* Adjusting to patient individual preference with respect to competing outcomes, e.g., side-effects vs. efficacy.
* Sample size calculations.
* Accommodating missing data.
* Making optimal decisions in real-time using data from mobile devices (mHealth).
* Exploratory data analysis.
* And more!
The goal of this working group is to bring together ideas from statistics and operations research to develop new tools for data-driven medical decision making.