Frailty models for multivariate time-to-event data
Steffen Unkel, University Medical Center Göttingen
Andreas Wienke, Martin-Luther-University Halle-Wittenberg
Day: Sunday, September 6
Length of tutorial: half day
Technical level: medium
In this tutorial we consider the case of multivariate time-to-event data such as (i) event times of related individuals (e.g. family members, patients in study centers), (ii) times to occurrence of different non-lethal diseases within the same individual (e.g. infections), and (iii) successive observations of recurrent events (e.g. hospitalisations), where the event of interest can occur more than once in an individual. In the example (i), each family or study center can be seen to form a cluster and the times to the event of interest observed from the individuals are observations within the cluster. In examples (ii) and (iii), each individual can be treated as a cluster and the times to the events observed from the individual are observations in the cluster. In each of the situations (i)-(iii), it is quite probable that there is some association within clusters of survival times in the sample as well as some heterogeneity among clusters.
Frailty models are random effects models for time-to-event data and provide a conceptually simple and appealing way of generating association between event times and of representing heterogeneities resulting from factors which may be diffcult or impossible to measure.
This tutorial will provide an introduction into the broad spectrum of frailty models for multivariate time-to-event data as well as into the mathematical foundations underpinning such models. The focus of the course is on the application of tools available in the software R to implement the methods using modern examples from the life sciences.
Institute of Biometry and Clinical Epidemiology
Charité - Universitätsmedizin Berlin & Berlin Institute of Health