Comparing two stratification models with the same population

[Rachel Wiley]

Good day -

I'm not sure how to approach this problem. We have two risk stratification models, and we'd like to compare their performance using the same population. Our primary outcome is a dichotomous variable, and my mentor suggested comparing our low risk to low risk, med to med, high to high and then creating an ROC curve but I'm having trouble understanding/conceptualizing how to do this.

I guess my basic question would be,
How would I even create a variable with two different groups that overlap in the same population?

Thanks in advance