The elephant in the room: Causal inference in the face of competing events

Jessie K. Edwards & Jessica G. Young

These three papers review 1) traditional thinking around analysis of competing risks data, 2) use of a formal causal framework for defining, identifying and estimating counterfactual contrasts under different treatment levels of classical statistical estimands in the competing risks literature (two versions of risk and 3 versions of hazard), clarifying that contrasts in risks quantify either total or controlled direct effects when the treatment affects the competing event while contrasts in any hazard do not generally have a causal interpretation and 3) new definitions of causal effect when competing events exist (the separable effects) which allow decomposition of the total effect into direct and indirect effects, outside of and through the competing event, respectively, that overcome challenges in both controlled direct effects, as well as the survivor average causal effect which is popular in the causal inference literature, when the treatment may affect the competing event.

Leave a Reply