Skip to main content

Table 6 Performance of the classification models: development and validation sets

From: Stratification of the severity of critically ill patients with classification trees

DEVELOPMENT (n = 1880)

Models

AUC (CI 95%)

HL-C

Brier

PPV (CI 95%)

PCC (CI 95%)

SMR (CI 95%)

APACHE II

0.81 (0.79 - 0.83)

68.2

0.17

0.72 (0.66 - 0.78)

0.75 (0.73 - 0.77)

1.30 (1.23 - 1.37)

SAPS II

0.82 (0.80 - 0.84)

77.2

0.16

0.74 (0.68 - 0.79)

0.74 (0.68 - 0.75)

1.31 (1.24 - 1.38)

MPM II 24

0.81 (0.79 - 0.83)

74.2

0.16

0.75 (0.70 - 0.80)

0.77 (0.75 - 0.79)

1.29 (1.22 - 1.36)

Logistic R

0.83 (0.81 - 0.85)

16.8

0.16

0.75 (0.70 - 0.80)

0.77 (0.76 - 0.79)

1.00 (0.92 - 1.10)

CART

0.78 (0.76 - 0.80)

------

0.17

0.67 (0.61 - 0.72)

0.75 (0.73 - 0.77)

1.00 (0.94 - 1.06)

CHAID

0.80 (0.78 - 0.82)

------

0.16

0.68 (0.63 - 0.73)

0.75 (0.73 - 0.77)

1.00 (0.93 - 1.08)

C4.5

0.80 (0.78 - 0.82)

------

0.16

0.69 (0.65 - 0.74)

0.78 (0.76 - 0.80)

1.00 (0.94 - 1.06)

VALIDATION (n = 804)

APACHE II

0.77 (0.74 - 0.81)

74.1

0.18

0.69 (0.60 - 0.70)

0.73 (0.70 - 0.76)

1.36 (1.26 - 1.47)

SAPS II

0.79 (0.76 - 0.83)

78.3

0.18

0.71 (0.63 - 0.78)

0.74 (0.71 - 0.77)

1.39 (1.28 - 1.49)

MPM II 24

0.79 (0.75 - 0.82)

66.9

0.18

0.71 (0.63 - 0.78)

0.74 (0.71 - 0.77)

1.36 (1.25 - 1.46)

Logistic R

0.81 (0.78 - 0.84)

41.5

0.17

0.73 (0.66 - 0.81)

0.75 (0.73 - 0.78)

1.22 (1.16 - 1.29)

CART

0.75 (0.71 - 0.81)

------

0.18

0.64 (0.57 - 0.72)

0.72 (0.69 - 0.75)

1.04 (0.95 - 1.31)

CHAID

0.76 (0.72 - 0.79)

------

0.18

0.64 (0.56 - 0.72)

0.72 (0.69 - 0.75)

1.06 (0.97 - 1.15)

C4.5

0.76 (0.73 - 0.80)

------

0.18

0.70 (0.63 - 0.76)

0.76 (0.73 - 0.79)

1.08 (0.98 - 1.16)

  1. AUC: Area under ROC curve; CI: Confidence interval; HL-C: Hosmer-Lemeshow test C (eight degrees of freedom); Brier: Brier score; PPV: Positive predictive value (cutoff 0.5); PCC: Percentage correctly classified (cutoff 0.5); SMR: Standardized mortality ratio. The severity scores (APACHE II, SAPS II and MPM II 24) were not developed in the development phase and recalibration was not performed.