CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Causal inference for semi-competing risks data Authors:  Daniel Nevo - Tel Aviv University (Israel) [presenting]
Malka Gorfine - Tel Aviv University (Israel)
Abstract: An emerging challenge for time-to-event data is studying semi-competing risks, where two event times are of interest: the non-terminal event (e.g. disease diagnosis) time, and a terminal event (e.g. death) time. The non-terminal event is observed only if it precedes the terminal event, which may occur before or after the non-terminal event, leading to the latter being unobserved or even undefined. Studying treatment or intervention effects on the event times is complicated because, for some units, the non-terminal event time may occur only under one treatment value but not the other. We will present and discuss new estimands, based on time-fixed stratification of the population. These estimands correspond to the scientific questions of interest, as we will exemplify using a real-data example of the effect of the APOE gene on Alzheimer's disease and death. We will then present novel assumptions utilizing the time-to-event nature of the data. The new assumptions enable partial identifiability of causal effects of interest, namely bounds. We will also present and discuss a sensitivity analysis approach based on semi-parametric frailty models. Finally, we will present non-parametric and semi-parametric estimators for the causal estimands.