Study population mean as a reference for COX-regressions in SPSS


I'm a medical student doing my master thesis. In my analysis I intend to compare two different systems for subclassing stroke as per ethiology. I have 1300 patients subtyped according to both these systems. I also have information of longterm outcome for these patients in terms of recurrents stroke and mortality. I intend to compare the prognostic information for each subtype between the two subtyping systems. I do this through two COX-regressions; one having the subtypes from system 1 as a categorical variable with mortality as outcome and the second one with the subtypes of system 2 as a variable. To be able to compare the two systems hazard ratios I need a common reference. I would like this to be the overall mortality rate for patients that hasn't been subclassed. As both of the systems has been classifying the same patients this will be a common refernce.

How do I use the overall study population mean as a reference for COX-regressions in SPSS? Is this what deviation does?

Thank you in advance,


Less is more. Stay pure. Stay poor.
Not following the "types"?

You have predictors of stroke, stroke death, and other death (censoring)? You want to examine an ethnic based variable, yes. Perhaps write out the models using words:

risk of stroke = covariates + ethnic, now what is the second model?

You have data, so that is the hardest part!
Thanks for your reply Hlsmith!
I realize now that was very unclear..

*1300 patients(all of which have had a stroke).

*Variable 1(stroke subtype according to system 1): six groups, 1-6, categorical.

*Variable 2(subtype according to system 2): also six groups, categorical, these groups has the same names for the subtypes but the criterias differ slightly, meaning they include the patients a bit differently.

*Other covariables: age, level of symptoms at stroke.

*Status/outcome: Mortality in longterm follow up.

*Time: Time to mortality.

*COX-regression 1: multivariable analysis with mortality as event, age and level as symptoms as non categorical variables and most importantly "variable 1" as categorical. So far I've set it as indicator.

*COX-reg 2: like above but with variable 2 instead.

*I want to be able to compare the hazard ratios for each subtype between the two systems. To do so I need a common reference for the two of them. Ideally that would be the mortality rate for the study population as a whole, the risk of dying if you haven't been subclassed. There comes the problem, how do I get the overall group risk to be the reference for my subcategories?

Thanks again :)


Less is more. Stay pure. Stay poor.
That is much clearer. And you can't put both var1 and var2 in the model at the same time because they are too similar, correct?

A solution doesnt jump out at me but I will stew on it.