Table 3 Outcomes of the internal validation test after the training with two data sets and of the external validation test using the excluded data set (Phase 1).

	Training and internal validation test							External validation test					P-value of difference between two validation (95% CI)
	Data set	AUC	Highest accuracy	Highest Youden index	Sen	Spe	Optimal cut-off value	Data set	AUC	Youden index	Sen	Spe	P-value of difference between two validation (95% CI)
(A)	A + B	0.966 (0.955, 0.975)	0.952	0.818	0.913	0.905	0.02	C	0.811 (0.784, 0.835)	0.421	0.793	0.628	< 0.001* (0.128, 0.183)
(B)	B + C	0.971 (0.959, 0.980)	0.943	0.848	0.943	0.905	0.06	A	0.895 (0.874, 0.913)	0.431	0.947	0.485	< 0.001* (0.059, 0.102)
(C)	C + A	0.946 (0.926, 0.961)	0.927	0.764	0.913	0.851	0.01	B	0.844 (0.828, 0.858)	0.493	0.651	0.842	< 0.001* (0.080, 0.125)

(A) External validation with set C after training and internal validation with sets A + B, (B) External validation with set A after training and internal validation with sets B + C, (C) External validation with set B after training and internal validation with sets C + A, (D) Internal validation after training with sets A + B + C. Positive, with intussusception; negative, without intussusception. AUC, area under the receiver operating characteristic curve (ROC). Accuracy, the fraction of the correct predictions over the total number of predictions. The Youden index, sensitivity + specificity – 1—that is, the vertical distance between the 45° line and the point on the ROC curve. In the external validation tests, we selected the optimal cut-off value based on the highest Youden index value in the internal validation tests. CI, confidence interval. Sen, sensitivity. Spe, specificity.
*P-values < 0.05 indicate a statistically significant difference.

Quick links

Search