Table 2

Discrimination and calibration statistics from recalibrated 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 (%)317.0 (9.7)317.0 (9.7)317.0 (9.7)208.0 (28.7)256.1 (35.4)259.7 (35.9)109.0 (4.3)60.9 (2.4)57.3 (2.3)
Brier score0.0520.0530.0530.1630.1660.1640.0200.0210.021
Spiegelhalter’s z-scorep = 0.2p = 0.5p = 0.5p<0.001p = 0.4p = 0.6p = 1.00p = 0.6p = 0.4
AUC0.9200.9210.9230.8300.8070.8110.8390.8260.833
Calibration-in-the-small0.00020.00010.00020.00850.00400.00430.0013<0.0001<0.0001
Calibration-in-the-large0000.0050<0.0001<0.00010.0004<0.0001<0.0001
Cox’s regression:
    Intercept α (SE)0000.28 (0.11)0.02 (0.10)0.01 (0.10)−0.93 (0.22)−0.24 (0.34)−0.08 (0.35)
    Slope β (SE)1110.72 (0.06)0.99 (0.08)1.03 (0.08)0.95 (0.08)0.94 (0.10)0.97 (0.11)
    Test: α = 0, β = 1p<0.001p = 0.9p = 0.9p<0.001p = 0.8p = 0.9
    Test: α = 0 | β = 1p<0.001p = 0.8p = 0.9p<0.001p = 0.7p = 0.9
    Test: β = 1 | αp<0.001p = 0.9p = 0.7p = 0.5p = 0.5p = 0.8
  • AUC, area under the curve; CRIB, Clinical Risk Index for Babies; CRIB II(−T), CRIB II without temperature at admission.