Table 1

Discrimination and calibration statistics from published predicted probability algorithms

All22–27 weeks28–32 weeks
CRIBCRIB IICRIB II(−T)CRIBCRIB IICRIB II(−T)CRIBCRIB IICRIB II(−T)
Observed deaths, n (%)317 (9.7)259 (35.8)58 (2.3)
Predicted deaths, n (%)324.4 (9.9)255.2 (7.8)216.6* (6.6)205.8 (28.4)206.0 (28.4)175.6* (24.2)118.6 (4.7)49.2 (1.9)41.1* (1.6)
Brier score0.0520.0550.0570.1630.1740.1830.0200.0210.021
Spiegelhalter’s z-scorep = 0.8p = 0.003p<0.001p<0.001p = 0.004p<0.001p = 1.00p = 0.09p = 0.004
AUC0.9200.9210.9230.8300.8070.8110.8390.8260.833
Calibration-in-the-small0.00020.00180.00440.00840.01140.02280.00140.00010.0002
Calibration-in-the-large<0.00010.00040.00090.00540.00540.01330.0006<0.0001<0.0001
Cox’s regression:
    Intercept α (SE)0.08 (0.11)0.50 (0.11)0.91 (0.13)0.34 (0.12)0.52 (0.12)0.94 (0.15)−0.86 (0.23)0.24 (0.39)0.80 (0.44)
    Slope β (SE)1.08 (0.05)1.08 (0.05)1.16 (0.05)0.78 (0.06)1.08 (0.09)1.20 (0.10)1.02 (0.09)1.02 (0.11)1.13 (0.12)
    Test: α = 0, β = 1p = 0.2p<0.001p<0.001p<0.001p<0.001p<0.001p<0.001p = 0.4p = 0.02
    Test: α = 0 | β = 1p = 0.6p<0.001p<0.001p<0.001p<0.001p<0.001p<0.001p = 0.2p = 0.01
    Test: β = 1 | αp = 0.1p = 0.08p = 0.002p = 0.0007p = 0.4p = 0.04p = 0.8p = 0.9p = 0.3
  • *Calculated using published predictive algorithm for CRIB II.

  • AUC, area under the curve; CRIB, Clinical Risk Index for Babies; CRIB II(−T), CRIB II without temperature at admission.