So, deviation is a reasonably correct use of a statistical term. It refers to the difference between an observed value and some reference value - usually the reference value is the mean, but not always. All we have here is a reference value that's actually a range, but it isn't much of a stretch.
On the other hand, depending on the audience you may wish to avoid that term for this particular value. If it's being used by the public, the lack of statistical education would likely mean some would take it as being insulting, that being overweight or underweight is "deviant" in a moralistic sense.
A brief bit of digging suggests that any other term would be a much more novel coining, however. If the audience is statisticians or clinicians, you can definitely use deviation, but you should take clear to carefully define it before use (if this is for a document) or in unavoidable bits of user interface (if this is for software).