R help - Jul 2007 - Statistics Question not R question: competing risks and non-informative censoring

All,

I am working with Emergency Department (ED) Length of Stay Data.  The ED
visit can end in one of a variety of ways (Admit, discharge, transfer,
etc...) Initially, I have modeled the time to event  by fitting a survival
model to  the time the outcome of interest and treat all other outcomes as
censoring. However I recently came across the cmprsk package in R which
seems to be developed specifically for competing risks survival models. I
have been able to gather that the key issue in deciding whether to model
each outcome separately or to use something like the cmprsk package is to
determine whether or not the censoring is "non-informative." I have
read a
up a little about informative vs. no-informative censoring, but remain
confused, is their a standard approach to determining whether or not
censoring is informative, can anyone suggest some good
(approachable/non-technical) references on the subject?

best,

Spencer

	[[alternative HTML version deleted]]