Table 3 Proportion of patients that consumed an opioid stratified by model predictions across varying risk classification thresholds, tabulated with accompanying sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV)

From: Predicting opioid consumption after surgical discharge: a multinational derivation and validation study using a foundation model

Cut-off value

Low risk rate (%)

High risk rate (%)

Sensitivity (%)

Specificity (%)

PPV (%)

NPV (%)

10

4.8 (n = 20/421)

44.3 (n = 192/433)

90.6

62.5

44.3

95.2

20

7.1 (n = 37/521)

52.6 (n = 175/333)

82.5

75.4

52.6

92.9

30

10.3 (n = 62/604)

60.0 (n = 150/250)

70.8

84.4

60

89.7

40

13.5 (n = 90/665)

64.6 (n = 122/189)

57.5

89.6

64.6

86.5

50

16.9 (n = 122/721)

67.7 (n = 90/133)

42.5

93.3

67.7

83.1

60

20.5 (n = 161/784)

72.9 (n = 51/70)

24.1

97

72.9

79.5

70

23.3 (n = 193/828)

73.1 (n = 19/26)

9

98.9

73.1

76.7