Grace Li

An Introduction to Joint Frailty Model for Recurrent Events and A Terminal Event

Recurrent event, which assumes that the same type of event can occur to a subject several times, arise frequently in statistical research. Cox proportional hazard models are not appropriate for studying recurrent events as they count only the first occurrence of an outcome in a subject and do not correctly reflect the natural history of the event. Moreover, researchers often encounter correlated data of different types from two survival processes, such as subjects’ repeated data together with time-to-event outcome. The observation of recurrent events could be terminated by survival outcome. The usual assumption of non-informative censoring (i.e. censoring time and survival time are independent conditional on the covariates) required by the Cox proportional hazard model can be violated. Considerable interest has been focused on joint models as they offer many advantages over separate analysis of two survival processes. In this talk, I will describe joint frailty models, explain their underlying assumptions, consider settings where these models are appropriate and discuss how to make inferences on these models. The method is illustrated by an analysis of data of recurrent events follicular lymphomas and a terminal event death. It provides new insights into studying the two survival processes over time as well as gives unbiased and efficient parameters.