Table 2 Pre-PCI models

From: STEMI-OP in-hospital mortality prediction algorithms: Frailty-integrated machine learning in older patients undergoing primary PCI

Models

Selection methods

Accuracy (%)

Sensitivity (%)

Specificity (%)

G-mean (%)

F1 score (%)

AUC (95% CI) (%)

p value (DeLong’s test)

CatBoost

ElasticNet

90.55

59.74

95.69

75.33

90.21

92.16 (91.93–92.38)

Reference

Logistic regression

RFE

86.24

81.38

87.20

84.17

87.34

90.10 (89.83–90.37)

0.9615

GRACE 2.0

-

80.38

78.38

82.38

80.01

57.30

83.48 (71.02–92.36)

<0.0001

TIMI

-

76.21

74.07

78.36

76.15

49.94

81.76 (70.49–89.05)

<0.0001

CADILLAC

-

81.79

85.36

78.21

81.34

58.27

87.01 (79.49–91.86)

<0.0001

PAMI

-

70.91

64.04

77.77

70.02

44.91

75.12 (62.25–84.99)

<0.0001

NCDR CathPCI v4

-

73.35

78.89

67.81

71.68

43.14

77.64 (66.62–85.52)

<0.0001

ALPHA

-

78.04

71.68

84.40

77.54

56.33

78.38 (63.65–89.29)

<0.0001

APEX AMI

-

79.76

71.70

87.81

79.15

60.20

82.02 (68.88–91.34)

<0.0001

Zwolle

-

81.13

78.29

83.97

80.87

57.61

85.16 (77.67–90.51)

<0.0001

  1. Performance comparison of CatBoost model with ElasticNet feature selection, logistic regression model with RFE feature selection, and traditional models. AUC area under the curve, CI Confidence Interval.
Back to article page