Table 3. Typical Statistical Measures and Methods for Evaluating the Performance of Survival Prediction Models.
| Aspect of Performance | Statistical Measure or Method | Nature of Model Output | Meaning |
|---|---|---|---|
| Discrimination | C index | Time-independent | Agreement between the predicted vs. the observed order of the events |
| Time-dependent ROC AUC | Time-dependent | Probability that the model predicts a randomly chosen patient who encountered the event before the specific time point as having higher risk than a randomly chosen patient who did not encounter the event by then | |
| Calibration | Calibration plot (with slope and intercept) | Event probability (time-dependent) | Plot of the event probability values predicted by a model against the observed event probabilities |
| Overall | Brier score | Time-dependent | Mean squared error of the predicted risks |
AUC = area under the curve, ROC = receiver operating characteristic