Table 2.

Predictor and trauma outcome variables

Model	AUC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Gap*	PPV (95% CI)	NPV (95% CI)	MCC (95% CI)
Child models
Goto LR20	0.78 (0.71 to 0.85)	0.54 (0.39 to 0.69)	0.91 (0.75 to 0.93)	0.55	0.01 (0.01 to 0.02)	0.990 (0.990 to 0.990)	–
Goto DNN20	0.85 (0.78 to 0.92)	0.78 (0.63 to 0.90)	0.77 (0.62 to 0.92)	0.45	0.01 (0.01 to 0.02)	0.990 (0.990 to 0.990)	–
Ours	0.86 (0.85 to 0.87)	0.78 (0.77 to 0.79)	0.78 (0.77 to 0.79)	0.44	0.09 (0.08 to 0.10)	0.992 (0.990 to 0.994)	0.626 (0.613 to 0.639)
Adult models
Raita LR21	0.74 (0.72 to 0.75)	0.50 (0.47 to 0.53)	0.86 (0.82 to 0.87)	0.64	0.07 (0.05 to 0.08)	0.988 (0.988 to 0.988)	–
Raita DNN21	0.86 (0.85 to 0.87)	0.80 (0.77 to 0.83)	0.76 (0.73 to 0.78)	0.44	0.06 (0.06 to 0.07)	0.995 (0.994 to 0.995)	–
Hong Triage DNN22	0.87 (0.87 to 0.88)	0.70	0.85	0.45	0.66	0.870	–
Ours	0.85 (0.85 to 0.85)	0.76 (0.76 to 0.76)	0.80 (0.80 to 0.80)	0.44	0.11 (0.11 to 0.11)	0.990 (0.989 to 0.991)	0.619 (0.614 to 0.624)
All ages models
Ours	0.85 (0.85 to 0.85)	0.74 (0.74 to 0.74)	0.81 (0.81 to 0.81)	0.45	0.12 (0.12 to 0.12)	0.989 (0.988 to 0.990)	0.602 (0.597 to 0.607)

*The gap between sensitivity and specificity. Calculated as follows: Gap=(1−Sensitivity)+(1−Specificity).

AUC, area under curve; DNN, Deep Neural Network; LR, logistic regression; MCC, Matthews Correlation Coefficient; NPV, negative predictive value; PPV, positive predictive value.