[Competitive risks and multi-state models in epidemiology]

An introduction to multi-state models is presented. Survival models can be considered as simple multi-state models; models with competing risks allow to consider several mortality causes: they enable estimating cause-specific mortality rates; however, care must be exerted for the interpretation of survival functions in such models. The illness-death model is very useful for the study of chronic diseases. Transition intensities can be estimated using conventional Cox models when observations are done in continuous time. In that model too, one must avoid pitfalls when interpreting survival functions.