Abstract:
|
Semi-competing risks refers to the setting where interest lies in some non-terminal time-to-event outcome, the occurrence of which is subject to a terminal event (usually death). Key to semi-competing risks data is that they provide an opportunity to learn about whether and how the two events co-vary. Existing analysis approaches, however, fail to take advantage of this. We propose a novel framework for the analysis of semi-competing risks data that views the two outcomes through the lens of a longitudinal bivariate process on a partitioning of the time scale. At the core of the framework are three time interval-specific regression models, each specified in a manner analogous to a generalized linear model, with time-varying components represented via B-splines. Key to the framework is that it captures two distinct forms of dependence, “local” and “global” dependence, both of which have intuitive clinical interpretations. Estimation and inference is performed via penalized maximum likelihood, and can accommodate both right censoring and left truncation (as needed). The methods are motivated by and illustrated with data from the Adult Changes in Thought study.
|