Multilevel joint analysis of binary outcomes or Binary GEE?

Question

I have repeated measures of systolic blood pressure for approximately 200 different kidney donors, taken at random times from nurses in the withdrawal from treatment to death period (which generally doesn't last longer than 4 hours). This has led to highly unbalanced and unequally spaced data. There is a minimum of 2, maximum of 50 and mean of 13 repeated measures.

I want to know whether different characteristics of these blood pressure trajectories are predictive of whether the recipients are likely to incur delayed graft function (a binary response variable that I have, indicative of successful transplant outcome). I want to adopt a formal modelling approach, but am not sure which would be more appropriate.

An extra complication to the correlation structure is the recipient centre variable that I have, which contains 22 levels.

I would be very grateful if someone could give me some hints to the following:

1. Is it possible to use a Multilevel Joint model for longitudinal and binary outcomes for this sort of data, as described in http://d-scholarship.pitt.edu/16754/1/SeoyeonHong_11262012.pdf ?
2. If so, is there any package for implementing this sort of model in R?
3. Is there any other type of model that one could recommend to be more appropriate for this analysis (such as Binary GEE)? (As I have not found much literature behind the Joint model for longitudinal and binary data).

Answer 1

A joint model which i built using a mixed model and a logisitc regression is a possible way to do it, there is no pacakge available to apply it but it is easy to program it your self using R or winbugs .