Table 3-.

Predictive performance of PREDICT29 in Brazil (A) and GC6 (B) Cohorts.

(A) Brazil
Model used	AUC	Sensitivity	Specificity
glmnet	0.929 (0.914, 0.944)	0.741 (0.703, 0.779)	0.867 (0.836, 0.897)
SVM	0.915 (0.900, 0.930)	0.756 (0.716, 0.796)	0.836 (0.807, 0.866)
ranger	0.867 (0.843, 0.891)	0.679 (0.641, 0.717)	0.830 (0.795, 0.866)
XGBoost	0.932 (0.919, 0.945)	0.794 (0.757, 0.830)	0.860 (0.828, 0.893)
AVERAGE	0.911 (0.894, 0.928)	0.742 (0.704, 0.780)	0.848 (0.816, 0.880)
(B) GC6
Model used	AUC	Sensitivity	Specificity
glmnet	0.664 (0.655, 0.674)	0.497 (0.459, 0.535)	0.787 (0.750, 0.824)
SVM	0.685 (0.676, 0.695)	0.535 (0.516, 0.554)	0.780 (0.766, 0.795)
ranger	0.683 (0.673, 0.693)	0.649 (0.616, 0.683)	0.651 (0.623, 0.678)
XGBoost	0.688 (0.677, 0.699)	0.549 (0.531, 0.567)	0.803 (0.787, 0.819)
AVERAGE	0.680 (0.670, 0.690)	0.558 .531, 0.585)	0.755 (0.732, 0.779)