Dear List,
I have data relating to the transplantation of organs in human patients. There are two separate but correlated outcomes : organ failure and death. A multi-state model appears to be one way of modelling the data.
Everyone starts off in the same state, namely 'having been transplanted'.
STATE 1: Patients in the first group stay in this state until the end of the study, alive and with a functioning organ (transplanted, transplant succeeds, alive).
STATE 2: Patients in a second group lose their organ, but stay alive (transplanted, transplant fails, alive).
STATE 3: Patients in a third group die with a functioning organ (transplanted, transplant succeeds, dead).
STATE 4: Patients in a fourth group suffer organ failure and then die (transplanted, transplant fails, dead).
Clearly STATES 3 and 4 are absorbing (you cannot leave these states having entered them).
I wish to model the transitions between these states, using a selected group of predictors which have previously been found to be associated with patient and / or organ death. The outcome is time to the event (so survival analysis is the method).
I have been told that it may be possible to model these data using a multi-state model using the package MLWin, but am not certain how to enter the data.
Does anyone have any ideas ?