I don't think SAS uses the term "JointPoint regression." This thread gives more common names, including "segmented models":
Solved: Joinpoint regression model - SAS Support Communities
Maybe I don't understand your question, but it seems like if you are looking for a segmented regression model, the impact is in the X variables, not the Y variables. I don't think it would matter if the Y variable is continuous, binary, or multinomial. Perhaps you could explain your data and what you are looking for?