Table 1.
Performance of time-to-event models.
| Model | Time-dependent concordance index, mean (95% CI) | Integrated Brier score, mean (95% CI) |
| Cox proportional hazards | 0.781 (0.771-0.791) | 0.098 (0.095-0.100) |
| Random survival forest | 0.786 (0.776-0.796) | 0.096 (0.094-0.099) |
| Gradient boosting machine | 0.787 (0.775-0.799) | 0.100 (0.097-0.102) |
| DeepSurv | 0.787 (0.776-0.798) | 0.097 (0.095-0.100) |
| DeepHit | 0.789 (0.779-0.799) | 0.108 (0.101-0.114) |
| Cox-Timea | 0.787 (0.776-0.798) | 0.097 (0.095-0.099) |
| Neural multitask logistic regression | 0.786 (0.776-0.796) | 0.098 (0.086-0.101) |
aNeural net-extended time-dependent Cox.