TABLE 3.

Performance metrics of all five machine learning algorithm models

	Only PRS scores	Features I only	PRS + Features I	Minimal Features only	PRS + Minimal Features
AUROC	0.669 (0.634, 0.708)	0.750 (0.714, 0.781)	0.788 (0.758, 0.819)	0.751 (0.714, 0.784)	0.788 (0.757, 0.820)
Sensitivity	0.920 (0.877, 0.963)	0.807 (0.743, 0.870)	0.800 (0.736, 0.864)	0.807 (0.743, 0.870)	0.800 (0.736, 0.864)
Specificity	0.295 (0.289, 0.302)	0.552 (0.545, 0.560)	0.629 (0.622, 0.636)	0.563 (0.556, 0.570)	0.646 (0.639, 0.653)
DOR	4.819 (4.229, 5.410)	5.151 (4.744, 5.557)	6.783 (6.382, 7.184)	5.377 (4.971, 5.784)	7.299 (6.897, 7.700)
LR+	1.306 (1.244, 1.370)	1.802 (1.664, 1.953)	2.157 (1.986, 2.341)	1.846 (1.704, 2.000)	2.260 (2.081, 2.454)
LR‐	0.271 (0.157, 0.466)	0.350 (0.252, 0.485)	0.318 (0.231, 0.438)	0.343 (0.248, 0.476)	0.310 (0.225, 0.426)
FPR	0.705 (0.698, 0.711)	0.448 (0.440, 0.455)	0.371 (0.364, 0.378)	0.437 (0.430, 0.444)	0.354 (0.347, 0.361)

Abbreviations: AUROC, area under the receiver operating characteristic; DOR, diagnostic odds ratio; FPR, false positive rate; LR+, likelihood ratio positive; LR−, likelihood ratio negative; PRS, polygenic risk score.