3. Predictive Validity
The overall scores of modified GUSS and VFSS illustrated high negative correlation (rs=-0.931, P<0.01), and the overall severity rating of results of modified GUSS and VFSS achieved excellent agreement (κ=0.848, P<0.01, P0=0.900). The modified GUSS rater found 16 patients (40.0%) with severe dysphagia, but the VFSS rater found 19 (47.5%) patients with severe dysphagia. Predictive values of severe dysphagia by Modified GUSS: sensitivity 84.2%, specificity 100.0%, positive predictive value 100.0% and negative predictive value 87.5% when compared with VFSS (κ=0.848, P<0.01, P0=0.925). The modified GUSS rater classified 18 (45.0%) patients as having severe and moderate dysphagia, whereas the VFSS rater classified 20 (50.0%) patients. Predictive values of severe and moderate dysphagia by Modified GUSS: sensitivity 90.0%, specificity 100.0%, positive predictive value 100.0% and negative predictive value 90.9% when compared with VFSS (κ=0.900, P<0.01, P0=0.950). Both raters confirmed 7 (17.5%) patients having no dysphagia in the sample. Predictive values of dysphagia by Modified GUSS: sensitivity、specificity、positive predictive value and negative predictive value were all 100.0% when compared with VFSS (κ=1, P<0.01, P0=1.00) (Table. 3 and Table. 4). The results of modified GUSS and the days from starting oral intake to removing gastric tube were demonstrated to have substantially negative correlation (rs= -0.664, P<0.01).