Table 3.
Concordance index of models on training and validation data.
| Model C-index | Training | Optimism-adjusted |
Validation
(without refitting) |
Validation
(with refitting) |
| Clinical | 0.81 (0.66-0.94) |
0.76 (0.61-0.80) |
0.58 (0.37–0.76) |
0.76 (0.58–0.91) |
| Clinical + SD dose + 90th percentile density | 0.84 (0.71-0.95) |
0.77 (0.61-0.82) |
0.64 (0.37-0.80) |
0.81 (0.66-0.99) |
|
Clinical + SD dose + 90th percentile density
+ SD dose*90th percentile density |
0.88 (0.79-0.99) |
0.82 (0.69-0.93) |
0.56 (0.41-0.70) |
0.86 (0.71-1.00) |
Optimism-adjusted results represent the internally validated models on the training data. Validation without re-fitting represents a strict prediction model validation procedure, whereas with re-fitting is building a new model with all the identified variables in the training data for comparison.