Table 3. Cross-validated model discrimination for different predictor sets and modelling techniques: Derivation cohort.

Predictor set	Model
	CPH		RF		GBC
	AUC	95% CI	AUC	95% CI	AUC	95% CI
QA (men only)	0.741	0.739, 0.743	0.754	0.752, 0.756	0.777	0.775, 0.779
QA (women only)	0.739	0.738, 0.740	0.755	0.754, 0.756	0.779	0.777, 0.781
QA	0.740	0.739, 0.741	0.752	0.751, 0.753	0.779	0.777, 0.781
QA+	0.751	0.750, 0.753	0.822	0.818, 0.826	0.834	0.833, 0.835
T	0.805	0.804, 0.806	0.825	0.824, 0.826	0.848	0.847, 0.849

For any given set of predictors, GBC outperforms the other 2 models. Similarly, for any given model, T predictors show the best predictive power.

AUC, area under the receiver operating characteristic curve; CPH, Cox proportional hazards; GBC, gradient boosting classifier; RF, random forest.