Ã山ǿ¼é

Title:ÌýNonadherence in Dynamic Treatment Regimes – Moving Beyond Intention-to-Treat Analyses.

´¡²ú²õ³Ù°ù²¹³¦³Ù:ÌýDylan Spicker is a postdoctoral fellow at Ã山ǿ¼é, supported by the CRM StatLab and CANSSI, under the supervision of Dr. Erica Moodie. Prior to their postdoc, Dylan completed their PhD and MMath in Statistics at the University of Waterloo. Dylan's research focuses on the development of statistical methods for causal inference, both using observational data in a longitudinal context and data which arise from respondent-driven sampling, with a particular interest in developing methods which are broadly accessible and applicable in a wide variety of settings. Website:

Dynamic treatment regimes (DTRs) are sequences of treatment rules that use patient-level information to inform treatment decisions. DTRs provide one framework for formalizing precision medicine in a longitudinal context. The estimation of optimal DTRs is an area of research which has received considerable attention, and which often assumes that patients perfectly adhere to their prescribed treatments. If patients may deviate from their assigned treatments, then applying standard DTR methods produces causal estimates for the impact of the treatment assignment, not the treatment itself. The causal impact of treatment is often of substantive interest. I will discuss the consequences of ignoring patient nonadherence in DTRs, paying attention to how this differs from nonadherence generally. I outline a framework which, through the modification of a commonly used DTR estimation technique, restores the desired causal interpretation. This allows estimation of optimal DTRs in the presence of nonadherent patients when treatment is a binary indicator.

Ìý

Via Zoom - Please visit our website at: Ìý

Ìý