Constructing Dynamic Treatment Regimes using STAR*D and CATIE
Both STAR*D and CATIE are large clinical trials in mental health in which patients are randomized and then re-randomized each time the patient shows insufficient response to treatment. One of the objectives of the trials is to formulate best sequences of treatment. In this talk we discuss how Q-learning can be used to construct more deeply tailored sequences of treatment, sometimes called adaptive treatment strategies or dynamic treatment regimes or policies. We discuss how one can provide measures of confidence as well.