Interaction effects

I have a variable that has 4 levels tied to the type of disability. I have a theory that the type of disability influences the impact of services on income gains for customers. I am not sure of the best way to specify an interaction effect for this. Say spending is A, the original 4 level categorical variable is d. One way is just to specify Y =B0 +Bspending + Bd +bsd (the interaction term).

Another way is to specify three dummies for the four levels and then specify interaction effects with the dummies and spending. There would be three interaction terms. I assume not specifying dummies is the correct approach (that is using the original categorical variable not dummies) although I was warned once up a time by a stats professor that specifying variables that are categorical with more than one level could generate nonsensical results.