Student Seminar Series: Ashkan Ertefaie, Constructing Optimal Dynamic Treatment Regimes from Longitudinal Data in Observational Study

We develop a methodology for constructing optimal dynamic treatment regimes from longitudinal data collected in an observational study over a long period of time. The optimal regime (policy) is the one that maximizes the expected utility function. Existing methods rely on backward induction which, practically, limits their applications to cases with small number of decision points. We overcome this limitation and propose a method which can be used to determine the optimal regime in chronic diseases where patients are monitored and treated throughout their life. We parametrize the optimal regime and form an estimating equation based on Temporal Difference residuals to estimate the parameters in infinite horizon settings where there is no a priori fixed end of follow-up point. We derive large sample results necessary for conducting inference. We also simulate a cohort of patients with diabetes which mimics the third wave of the National Health and Nutrition Examination Survey and examine the performance of the proposed method in controlling the level of hemoglobin A1c.