🚀 go-pugleaf

I have a group of patients for whom I can calculate an individual predicted
probability of mortality using a multi-component index (Knaus' APACHE II)
I also know whether the patient actually survived the event.  What I
want to find out is how the actual survival of the group compared to
the prediction.

The actual study design is like this: there are two groups, one received
a specified treatment.  The prior probability of death can be calculated for
each patient (the method of calculating the prior probability is based on
multiple variables and was validated on other groups of patients in many
other situations around the world) and the outcome variable is DIED (Y/N).

Since the variable DIED takes two values, 1 if died 0 if not, I thought to
compare the actual to predicted by doing  a t-test on the means of DIED
and ProbDie. Is this a valid method? If so, how do I do it in SAS?

If not, what is a valid method?

I have about 50 patients in each group, and the death rate is about 20%.
These are acutely ill patients. So survival time is irrelevant.

Any help would be appreciated.

--
Aviel Roy-Shapira, M.D.         Ben-Gurion University Medical School
Dept. of Surgery A.             POB 151, Beer Sheva, Israel
avir@chen.bgu.ac.il             AVIR@BGUVM.BITNET