Kim Annink

83 Cerebral ultrasound scoring system for neonates with HIE Table 4: Performance of the model in cohort I and cohort II. Cut-off value* ≥3 ≥4 ≥5 ≥6 ≥7 Cohort I Sensitivity 93% (76–99) 79% (60–91) 45% (27–64) 28% (13–47) 17% (7–36) Specificity 86% (74–94) 88% (76–95) 92% (81–96) 98% (88–99) 100% (91–100) PPV 79% (62–91) 79% (60–91) 76% (50–92) 89% (59–99) 100% (46–100) NPV 96% (84–99) 88% (76–95) 75% (62–85) 71% (59–81) 68% (57–78) Cohort II Sensitivity 75% (47-92) 69% (41-88) 63% (36-84) 56% (31-79) 44% (21-69) Specificity 94% (68-100) 94% (68-100) 100% (76-100) 100% (76-100) 100% (76-100) PPV 92% (62-100) 92% (60-100) 100% (66-100) 100% (63-100) 100% (56-100) NPV 79% (54-93) 75% (51-90) 73% (50-88) 70% (47-86) 64% (43-81) * A cut-off value of ≥3 means that an ultrasound score of 3 or more is defined as abnormal. Validation: inter-observer variability Table 5 shows the agreement between the observers in cohort I. There is a moderate inter-observer agreement between all three observers. Table 5: Agreement between the observers in cohort I. Observer 1 vs. 2 Observer 1 vs. 3 Observer 2 vs. 3 Spearman’s rho 0.74 (p=0.001) 0.64 (p<0.001) 0.72 (p=0.001) Validation of the scoring system in cohort II The predictive values of the scoring system in cohort II, are shown in Table 5. The AUC was 0.89 (95%CI 0.77-1.00). To exclude the effect of the hospital on outcome, logistic regression was performed with the total CUS score, the hospital and their interaction term included in the analysis. The CUS score was significantly associated with adverse outcome (OR=2.5; 95%CI 1.8-3.4), the hospital and their interaction term were not. 4