I think it strongly depend on what the model will be used to.
For rendering it on a still picture or an animation, if the end result is satisfying you can keep using crease if you want.
Now for exporting it into some engine (game or other) that use models in a real time environment, it will depend a lot if the format used in such engine support crease or edge marked as sharp , or not .
If the format of model used in such engine do not support crease or edges marked as sharp, you will need to add additional topology yourself. If the format use crease or edge marked as sharp, i think it’s then best to -not- add additional topology as you would then increase the face count a lot more than required.
A real time engines, like games, are all about face count optimisation (as it must be able to display environement+ model etc… in the same time, in the end adding all the faces of everything that can displayed in the field of view of the user can be high)