Table 3 Performance metrics for determining breast cancer recurrence using pre-specified variable algorithm and conditional tree algorithms

Sensitivity

83.3 (77.9–88.2)

83.2 (77.2–88.6)

81.1 (77.4–84.8)

78.0 (72.0–83.9)

73.7 (67.1–79.7)

68.5 (64.6–72.4)

Specificity

89.7 (85.0–93.6)

92.5 (88.1–95.8)

93.2 (90.6–95.4)

95.7 (92.5–98.4)

96.4 (93.4–98.8)

97.0 (95.4–98.5)

PPV^a

66.8 (57.9–76.7)

64.7 (54.0–76.8)

61.4 (53.7–70.2)

81.9 (72.0–92.4)

73.7 (63.5–91.1)

75.4 (66.3–85.7)

NPV^b

96.0 (94.2–96.9)

97.1 (96.1–98.0)

97.4 (96.8–97.9)

94.6 (93.2–96.0)

96.4 (94.7–96.7)

95.8 (95.3–96.3)

Correct classification

88.4 (84.5–91.9)

91.1 (87.6–94.2)

91.8 (89.4–93.8)

92.2 (89.5–94.6)

93.1 (90.3–95.6)

93.6 (92.1–94.9)

Scaled Brier

0.27 (0.03–0.49)

0.27 (−0.03–0.53)

0.21 (−0.01–0.40)

0.51 (34.2–66.4)

0.44 (0.20–0.64)

0.39 (0.24–0.51)