Table 2.

Experiment 1: prognostic accuracy (Harrell’s C) for the original model (M1, diagnostic spectra) developed through Clinical-learning (a priori clinical knowledge) vs machine learning (LASSO and RIDGE). The EPV is >20 (55.6)

Method	Derivation Data Set (N = 33,820)			Validation Data Set (N = 54,716)			Optimism
	Harrell’s C	SE	95% C.I.	Harrell’s C	SE	95% C.I.
Unregularized	0.800	0.008	0.784–0.816	0.791	0.008	0.775–0.807	0.009
Lasso	0.798	0.008	0.782–0.814	0.789	0.008	0.773–0.805	0.009
Ridge	0.810	0.008	0.794–0.826	0.788	0.008	0.772–0.804	0.022