Recurrent Event Survival Analysis


New Member
Hi there,

I'm working on a research on individuals who enter financial distress using the Cox proportional hazards model. In this case the individuals can get in financial distress multiple times within a time-window such that independency assumption is probably not satisfied. I stumbled upon some papers such as this ( which take into account recurrent events, however, in this case it is assumed that after an event took place, the clock starts ticking for the next event to happen.

I'll try to make it more clear using an example. Someone get's in financial distress at 1 May 2005 and goes out of financial distress on 23 May 2005, then he get's in financial distress at 7 July 2005 and goes out of financial distress on 08 August 2005. So in this case I want to ignore the period between 23 May and 7th of July.

Another example: the R frailtypack contains a dataset called "readmission" which can be used for recurrent event survival analysis. This dataset contains a column "id" stating the id of an individual, a column enum indicating whether it is the first, second, ... n-th event for a specific individual, t.start indicating the starting time of an interval, and a column t.stop indicating the end time of an interval. Now in this case the t.start of an episode equals zero or equals the ending time (t.stop) of the previous interval of an individual.
I would like to be able to work with t.start values which exceed the t.stop values of the previous interval. The problem is that I can not figure in what way that will effect the assumptions, outcomes of the model . An example of the dataset can be found through this link: on page 12.

The idea of this research is to get a prediction of time to recovery for an individual.

Is there anyone having some ideas how to cope with this?

Kind regards,

Last edited: