Best practices for handling values?

When working with value data, that is, information such as greyscale maps, is it more efficient to use a math node over a color node? For example, when inverting values frequently you’ll see people placing an “invert” node. In this case would it be more efficient to use a subtraction node instead?