Table 3.
Model | Details | Estimated risk probabilities | |
---|---|---|---|
NE | SE | ||
Calibrated SELF | Mean observed | 0.2519 | 0.0801 |
Mean expected | 0.2555 | 0.0829 | |
Decile 1 | 0.0939 | 0.0198 | |
Decile 10 | 0.5532 | 0.2452 | |
Ratio decile10/decile1 | 5.89 | 12.39 | |
p of chi squared of GNDD test | 0.7296 | 0.5942 | |
Re-calibrated SCM | Mean expected | 0.1148 | 0.1908 |
Decile 1 | 0.0311 | 0.0652 | |
Decile 10 | 0.3107 | 0.4420 | |
Ratio decile10/decile1 | 10.0 | 6.78 | |
p of chi squared of GNDD test | <0.0001 | <0.0001 | |
Re-calibrated CM | Mean expected | 0.3516 | 0.0547 |
Decile 1 | 0.1161 | 0.0171 | |
Decile 10 | 0.3516 | 0.1419 | |
Ratio decile10/decile1 | 6.40 | 8.32 | |
p of chi squared of GNDD test | <0.0001 | <0.0001 | |
Re-calibrated C | Mean expected | 0.2624 | 0.0768 |
Decile 1 | 0.0802 | 0.0243 | |
Decile 10 | 0.6070 | 0.1957 | |
Ratio decile10/decile1 | 7.60 | 8.06 | |
p of chi squared of GNDD test | 0.2820 | 0.3776 | |
Re-calibrated WC | Mean expected | 0.2763 | 0.0938 |
Decile 1 | 0.0628 | 0.0150 | |
Decile 10 | 0.7129 | 0.3295 | |
Ratio decile10/decile1 | 11.34 | 21.94 | |
p of chi squared of GNDD test | <0.0001 | 0.0178 | |
Re-calibrated ALTS | Mean expected | 0.2141 | 0.0950 |
Decile 1 | 0.0765 | 0.0229 | |
Decile 10 | 0.4803 | 0.2774 | |
Ratio decile10/decile1 | 4.32 | 12.09 | |
p of chi squared of GNDD test | 0.0265 | 0.0072 |
NE = Northern Europe; SE = Southern Europe;
SELF = original model, respectively NE or SE; SCM = substituting S(0), coefficients and means of risk factors; CM = model substituting coefficients and means of risk factors; C = model substituting coefficients only; WC = model substituting coefficients with the widest coefficients that had the corresponding HR (95% CI): age 1.50 (1.28 1.77); BMI 1.21 (1.05 1.41); systolic blood pressure 1.55 (1.32 1.82); heart rate 1.05 (0.81 1.36); cholesterol 1.28 (1.18 1.39); ex-smokers 1.31 (0.75 2.26); smoker 2.18 (1.45 3.28); finally ALTS = substituting S(0) with altered S(0): for NE 0.9098 instead of 0.7698; for SE: 0.9254 instead of 0.9354.
GNDD = Greenwood-Nam-D’Agostino-Demler goodness of fit test for re-calibrated survival models: by this test significant difference (outlined here in bold) means that re-calibration is not good enough24.