Table 3.

Prediction performance metrics for the total model and the models without a particular domain of exposures.

Model	Optimal threshold ROC curve					Sensitivity and specificity at a predefined point of 0.5
AUC (95% CI)	Threshold	Specificity	Sensitivity	Sensitivity + specificity	Accuracy	Specificity	Sensitivity
Total
0.707 (0.655–0.759)	0.789	0.725	0.593	1.318	0.704	0.777	0.787
Without demographic exposures
0.684 (0.630–0.739)	0.792	0.713	0.565	1.278	0.690	0.767	0.759
Without lifestyle exposures
0.695 (0.642–0.747)	0.875	0.494	0.815	1.309	0.545	0.774	0.806
Without environmental exposures
0.702 (0.650–0.754)	0.866	0.539	0.796	1.335	0.580	0.774	0.796
Without biological exposures
0.669 (0.611–0726)	0.811	0.645	0.611	1.256	0.640	0.730	0.685