Table 2.

Summary of cross-validated prediction models on trained data (n = 3473)

Model	AUC	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Model 1
Logistic regression	0.73 [0.71–0.75]	64 [60–68]	70 [68–71]	32 [30–35]	89 [88–91]
Elastic net	0.74 [0.72–0.76]	68 [64–71]	69 [67–71]	33 [31–36]	90 [89–92]
Random forest	0.74 [0.72–0.76]	6 [4–8]	99 [99–99]	60 [47–71]	82 [81–83]
SVM	0.65 [0.62–0.68]	49 [45–53]	73 [71–74]	29 [27–32]	86 [85–88]
Model 2
Logistic regression	0.74 [0.72–0.76]	67 [63–70]	70 [68–72]	34 [31–36]	90 [89–91]
Elastic net	0.74 [0.72–0.76]	67 [63–70]	69 [67–71]	33 [30–36]	90 [89–91]
Random forest	0.74 [0.72–0.76]	47 [43–51]	84 [82–85]	40 [36–44]	88 [86–89]
SVM	0.73 [0.71–0.75]	72 [69–76]	63 [61–65]	31 [28–33]	91 [89–92]
Model 3
Logistic regression	0.71 [0.68–0.74]	64 [60–68]	68 [66–70]	31 [29–34]	89 [88–91]
Elastic net	0.71 [0.68–0.74]	64 [60–67]	67 [65–69]	31 [28–33]	89 [88–90]
Random forest	0.71 [0.68–0.74]	55 [51–59]	75 [73–77]	34 [31–37]	88 [87–89]
SVM	0.70 [0.67–0.73]	69 [65–73]	61 [59–63]	29 [27–31]	90 [88–91]

AUC = area under the ROC curve. SVM = support vector machine