Table 3.
Model performance parameters in the total cohort and after 5-fold cross-validation.
| Model discrimination | Model calibration | ||||
|---|---|---|---|---|---|
| Validation stage | C-statistica (95% CI) | Slopeb | Interceptb | Hosmer-Lemeshow test | |
| Apparent performancec (n = 350) | 0.820 (0.771–0.865) | 1.000 | 0.000 | P = 0.182 | |
| Internal validation | Optimism corrected performance (n = 350) | 0.792 (0.743–0.837)d | 0.968 | −0.022 | |
| Five-fold cross-validation | Fold 1 (n = 70) | 0.839 (0.737-0.941) | 0.950 | −0.227 | P = 0.682 |
| Fold 2 (n = 70) | 0.687 (0.543-0.831) | 0.698 | −0.066 | P = 0.090 | |
| Fold 3 (n = 70) | 0.678 (0.790-0.902) | 0.836 | 0.374 | P = 0.421 | |
| Fold 4 (n = 70) | 0.763 (0.647-0.879) | 0.936 | −0.026 | P = 0.310 | |
| Fold 5 (n = 70) | 0.801 (0.681-0.921) | 1.240 | −0.047 | P = 0.552 | |
a
C-statistic of 0.50 represents no discrimination and 1.00 represents perfect discrimination.
b
Intercept of 0 and slope of 1 represent perfect calibration.
c
Refers to performance estimated directly from bootstrap sample that was used to develop prediction model. Since these parameters are generated in 1,000 bootstrap samples, for each parameter the median is shown.
d
Average optimism = 0.028 determined by internal validation in bootstrap samples (1,000 samples with replacement).