Defining, identifying, and estimating, causal effects of generalized time-varying treatment strategies on repeated, informatively measured outcomes truncated by death

Researchers often express interest in estimating causal effects of (possibly time-varying) treatment strategies on the mean of an outcome that is undefined after an individual dies.  For example, the Medications and Weight Gain in PCORnet (MedWeight) study was an observational study leveraging electronic health records (EHR) that aimed to estimate the effect of initiating (and subsequently adhering to) different medications for the same indication on future weight change at different times up to 24 months post treatment initiation in an adult, clinical population where some individuals die in every month. This renders the standard notion of an average total treatment effect undefined at any time.  Truncation by death is a ubiquitous problem, even in randomized trials, that cannot be solved with any particular choice of estimator (statistic).  Rather, truncation by death creates the more fundamental challenge of defining an alternative effect notion to the total effect that aligns with the true study motivation, which often involves some clinical decision. At the same time, an outcome like weight change is not only informatively measured, but also sparsely measured in an EHR at any specifically chosen follow-up time such that estimators which can time-smooth in available repeated outcomes beyond that chosen time are desirable for improved precision.  In this talk, we consider a studyable estimand for truncation by death settings that, under a testable isolation condition, is a causal effect which connects to a conditional version of the recently posed separable adherence effects.  We pose time-smoothed estimators of these effects for EHR data that can smooth over all available outcome measures for increased precision.