I have data for patients that were subjected to either one treatment or multiple treatments at various points in time and I need to analyse their survival times after treatments. This of course means that some patients appears only once in the dataset (those had been subjected to only one treatment) and some several times (those had been subjected to multiple treatments). So it seems I have clustered data.


So the question is: what is the best model to use in such a setting. I briefly looked at frailty models (see: https://www.jstatsoft.org/article/view/v047i04) and it seems that this may be a way to go. However, I have some doubts as for how to deal with the fact that of course if a patient was subjected, lets say, to two different treatments, then the survival time for the one that was conducted first will be longer, no matter what. So I wonder what is the best way to address this problem?


1 Answer 1


You may find useful a relatively new type of model, which is a combination of mixed and relative risk models: "Joint Models for Longitudinal and Survival Data".

Here are some links: link1, link2, link3, link4, link5, link6.

  • $\begingroup$ But does it means also that you share my doubts concerning frailty models? That is, that they may be not really appropriate for this kind of problem? $\endgroup$
    – sztal
    Jul 12, 2016 at 10:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.