In semicompeting risks problems, non-terminal time-to-event outcomes such as time to hospital readmission are subject to truncation by death. Such settings are often evaluated with parameters from illness-death models, but evaluating causal treatment effects with such models is problematic due to the evolution of incompatible risk sets over time. As an alternative, the survivor average causal effect (SACE) is a principal stratum causal effect of a treatment on the non-terminal event among units that would survive regardless of the assigned treatment. Traditional SACE formulations specify a single time point past which an individual is deemed to have always survived.
We propose a new causal estimand, the time-varying SACE (TV-SACE), for non-terminal events in the semicompeting risks setting. We adopt a Bayesian estimation procedure that is anchored to parameterization of illness-death models for both treatment arms but maintains causal interpretability. We outline several frailty specifications and highlight their connection to assumptions from the principal stratification literature. The method is demonstrated with data on hospital readmission for pancreatic cancer patients.