Development and validation of three machine-learning models for predicting multiple organ failure in moderately severe and severe acute pancreatitis

Table 5 Comparison of SVM, LRA, ANN and APACHE II score for predicting MOF

Variable	SVM (95% CI)	LRA (95% CI)	ANN (95% CI)	APACHE II score (95% CI)	P value
SEN	75.00%(63.16–84.13%)	79.17%(67.67–87.50%)	86.11%(75.48–92.78%)	80.56%(69.20–88.59%)	0.413
SPE	81.68%(75.30–86.75%)^{b, c}	77.49%(70.78–83.07%)	65.45%(58.19–72.07%)^a	65.45%(58.19–72.07%)^a	< 0.001
FPR	18.32%(12.84–23.81%)^{b, c}	22.51%(16.59–28.43%)	34.55%(27.79–41.31%)^a	34.55%(27.79–41.31%)^a	< 0.001
FNR	25.00%(15.00–35.00%)	20.83%(11.45–32.21%)	13.89%(5.90–21.88%)	19.44%(10.30–28.58%)	0.413
PPV	60.67%(49.72–70.69%)^c	57.00%(46.72–66.73%)	48.44%(39.58–57.39%)	46.77%(37.83–55.92%)	0.129
NPV	89.66%(83.91–93.58%)	90.80%(85.01–94.58%)	92.59%(86.45–96.16%)	89.93%(83.38–94.18%)	0.827
Accuracy	79.85%(74.10–83.80%)^{b, c}	77.95%(72.94–82.96%)	71.10%(65.62–76.58%)	69.58%(64.02–75.14%)^a	0.014
AUC	0.840 (0.783–0.896)	0.832 (0.773–0.890)	0.834 (0.777–0.890)	0.814 (0.759–0.869)	–

^aCompared with LRA, P < 0.05
^bCompared with ANN, P < 0.05
^cCompared with APACHE II score, P < 0.05
P value denoted the overall statistical result for the three models and APACHE II score
Entries in boldface showed significant difference