An Investigation of the Dependability of Clinical Cutting Scores Using Generalizability Theory