Table 4.
Overview of models’ discrimination and overall performance in the validation.
| Method/model | AUC (95% CI) | AUCbias (95% CI) | Brierscaled | Slope (95% CI) | Intercept (95% CI) |
|---|---|---|---|---|---|
| 0 – no adjustments | 0.726 (0.719, 0.733) | – | 1.47% | 0.781 (0.752, 0.811) | 0.669 (0.539, 0.800) |
| 1 – calibration-in-the-large | 0.726 (0.719, 0.733) | – | 5.26% | 0.781 (0.752, 0.811) | −0.531 (−0.618, −0.444) |
| 2 – logistic calibration | 0.726 (0.719, 0.733) | – | 5.89% | 1.000 (0.962, 1.038) | 0.000 (−0.106, 0.106) |
| 3 – refitting | 0.738 (0.731, 0.745) | 0.737 (0.731, 0.744) | 6.53% | 1.000 (0.965, 1.035) | 0.000 (−0.098, 0.098) |
| 4 – refitting with different predictor assessment | 0.738 (0.731, 0.745) | 0.737 (0.731, 0.745) | 6.53% | 1.000 (0.965, 1.035) | 0.000 (−0.098, 0.098) |
| 5 – refitting with numerical predictors as continuous | 0.741 (0.734, 0.748) | 0.741 (0.734, 0.748) | 6.53% | 1.000 (0.966, 1.034) | 0.000 (−0.097, 0.097) |
| AUSDRISK | 0.723 (0.716, 0.730) | – | 4.42% | 0.956 (0.920, 0.991) | −0.514 (−0.600, −0.430) |
Abbreviations: AUC = area under the receiver-operator curve; AUCbias = bias-corrected AUC for refitted models; Brierscaled = scaled Brier score; CI = confidence interval.