Competing Models for Competing Risks

• Activity Number: 622
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): John D. Kalbfleisch and Yining Ye and Douglas E. Schaubel*+
• Companies: University of Michigan and Amgen and University of Michigan
• Keywords: competing risks ; hazard modeling ; cumulative incidence

Abstract

Models and methods for competing risks are important in those situations in which each individual under study may experience one of two or more different types (or causes) of failure; for each individual we observe the time to failure or censoring, and in the case of failure, the type of failure. Various approaches to analyzing data on competing risks have been taken in the literature, including direct modeling of cause specific hazard functions, modeling latent failure times, and direct modeling of cumulative incidence functions. The methods are reviewed with respect to: i) estimating or evaluating the relationships between measured covariates and the occurrence of failures of specific types; and ii) assessing the interrelationships that exist between failure types under a specific set of study conditions. Main criteria considered include simplicity of inference, transparency of the models and methods, and plausibility of the underlying assumptions. It is argued that the formulation of the problem in terms of the basic cause specific hazard functions is most easily interpreted and provides the needed flexibility. Some related comments are made on semi competing risks.