Table 2.  Areas under operating receiver characteristic curves and test characteristics (95% confidence intervals) for entire dataset for three risk assessment scales* Norton scale Braden scale Waterlow scale All patients (n=2190 patient weeks) Area under curve 0.56 (0.51 to 0.61) 0.55 (0.49 to 0.60) 0.61 (0.56 to 0.66) Sensitivity (%) 46.2 (37.7 to 54.7) 43.5 (35.0 to 52.0) 89.5 (83.8 to 95.1) Specificity (%) 60.4 (58.3 to 62.6) 67.8 (65.7 to 69.8) 22.4 (20.5 to 24.4) Positive predictive value (%) 7.1 (5.5 to 9.1) 8.1 (6.2 to 10.3) 6.7 (5.5 to 8.0) Negative predictive value (%) 94.5 (93.1 to 95.7) 94.9 (93.6 to 95.9) 97.2 (95.1 to 98.5) Subpopulation (n=1355 patient weeks) Area under curve 0.71 (0.65 to 0.77) 0.71 (0.64 to 0.78) 0.68 (0.61 to 0.74) Sensitivity (%) 78.7 (66.3 to 88.1) 72.9 (59.7 to 83.6) 95.9 (86.0 to 99.5) Specificity (%) 46.5 (43.7 to 49.3) 57.2 (55.4 to 60.0) 22.0 (19.5 to 24.5) Positive predictive value (%) 7.0 (5.2 to 9.2) 7.8 (5.7 to 10.4) 5.3 (3.9 to 7.0) Negative predictive value (%) 97.7 (96.1 to 98.8) 97.7 (96.3 to 98.7) 99.2 (97.0 to 99.9) * Cut-off points: at risk if <16 on the Norton scale, <18 on the Braden scale, or >9 on the Waterlow scale. Excluding patient weeks with preventive measures without pressure ulcers and surgical patients.