System Optimization Via SMART Methodology with Applications to an Implementation Study of Improving Individual Interventions for Depression

A large class of implementation studies involves the selection of one or more efficacious treatment choices across time so that interventions delivered to individuals can match their individual-specific preferences, needs, and prior histories.  Ultimately, a comprehensive intervention regimen would need to provide good choices for an alternative intervention when the first-line intervention fails to produce a satisfactory response in an individual, and whether an intervention should be continued or stopped once a person achieves a satisfactory response. Such intervention regimens are more complex than those tested in standard randomized trials so a new genre of research designs is required. SMART trials are particularly suited for such dynamic treatment regimens. In this talk, I will introduce an enhanced version of the SMART design to optimize a treatment program of dynamic treatment regimens. The proposed method involves a novel application of adaptive randomization aimed to address three main concerns of an implementation study: it allows incorporation of historical data and/or prior knowledge, it includes randomization for learning purposes, and it aims to improve care via adaptation throughout the program. As an example we discuss a regimen for patients with depression following post-acute coronary syndrome. Using a simulation study, we illustrate that the inputs from historical data are important for the program performance measured by the expected outcomes of the enrollees, but also show that the adaptive randomization scheme is able to compensate for poorly specified historical inputs. The simulation results also confirm that the proposed design allows efficient learning of the treatments by alleviating the curse of dimensionality.  Finally, we describe how these enhanced designs address the dual goals of learning about effectiveness of the intervention regimens that are informed by responsiveness to first and improving the quality of care for patients enrolled into the program.