01-10-2012 10:45 PM
I have 1000 patients divided into 3 groups depending upon the predictive mortality rate for patients. This predictive score was calculated based upon baseline risk profile of the patient.
3 groups are as follows:
Low risk group: those patients who have predictive mortality rate of 0 to 1.5%
Intermediate risk group: >1.5% - 3%
High risk group: >3%
So, I will have a mean±SD for the predictive mortality rate for each group.
In each of the 3 group: I have the "real" observed death rate as well.
I was wondering how best to compare the predictive score in each group to the observed death rate in each group? What kind of test should I do?
01-11-2012 09:41 AM
Consider the Hosmer-Lemeshow test. It is basically a one-way chi-square that uses the group sum of the predicted mortalities as the expected mortality to compare to the observed mortality.