Extensions in Joint Modeling of Survival and Longitudinal Outcomes

abstract

In many medical studies both longitudinal and event history data are collected for each patient. A well-known and broadly studied example is found in AIDS research, where CD4 cell counts taken at different time points are related to the time-to-death. Often such outcomes are separately analyzed. However, in two particular settings a joint modeling approach is required. First, when focus is on the event times and we wish to study the association between the longitudinal responses and the risk for an event, where we need to take into account the fact that the longitudinal response is also an outcome generated by the same subject. Second, when focus is on the longitudinal outcome and events cause dropout. Under specific circumstances the dropout process needs to be accounted in the analysis in order to obtain valid inferences for the longitudinal outcome.