Comparison of the performance of the ML model, the Cox-derived score, and the CIBMTR model
| . | ML model . | ML model . | Cox-derived score . | Cox-derived score . |
|---|---|---|---|---|
| 4 groups . | 4 groups . | |||
| OS risk score for mortality prediction (Harrell C-index) . | ||||
| Training set (n = 1773)∗ | 0.603 | 0.596 | 0.594 | 0.589 |
| Test set (n = 566)∗ | 0.612 | 0.608 | 0.587 | 0.580 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 0.608 | 0.599 | 0.557 | |
| Test set (n = 618)† | 0.654 | 0.650 | 0.581 | |
| OS risk score for mortality prediction (AIC) | ||||
| ML model | ML model | Cox-derived score | Cox-derived score | |
| 4 groups | 4 groups | |||
| Training set (n = 1773)∗ | 9 944 | 9 954 | 9945 | 9 954 |
| Test set (n = 566)∗ | 2 757 | 2 759 | 2765 | 2 767 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 12 573 | 12 588 | 12 647 | |
| Test set (n = 618)† | 3 318 | 3 322 | 3 374 | |
| OS risk score for NRM prediction (AIC) | ||||
| ML model | ML model | Cox-derived score | Cox-derived score | |
| 4 groups | 4 groups | |||
| Training set (n = 1763)∗ | 7 210 | 7 210 | 7208 | 7 214 |
| Test set (n = 566)∗ | 2 006 | 2 002 | 2008 | 2 006 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 8 738 | 8 738 | 8 762 | |
| Test set (n = 618)† | 2 154 | 2 160 | 2 182 | |
| . | ML model . | ML model . | Cox-derived score . | Cox-derived score . |
|---|---|---|---|---|
| 4 groups . | 4 groups . | |||
| OS risk score for mortality prediction (Harrell C-index) . | ||||
| Training set (n = 1773)∗ | 0.603 | 0.596 | 0.594 | 0.589 |
| Test set (n = 566)∗ | 0.612 | 0.608 | 0.587 | 0.580 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 0.608 | 0.599 | 0.557 | |
| Test set (n = 618)† | 0.654 | 0.650 | 0.581 | |
| OS risk score for mortality prediction (AIC) | ||||
| ML model | ML model | Cox-derived score | Cox-derived score | |
| 4 groups | 4 groups | |||
| Training set (n = 1773)∗ | 9 944 | 9 954 | 9945 | 9 954 |
| Test set (n = 566)∗ | 2 757 | 2 759 | 2765 | 2 767 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 12 573 | 12 588 | 12 647 | |
| Test set (n = 618)† | 3 318 | 3 322 | 3 374 | |
| OS risk score for NRM prediction (AIC) | ||||
| ML model | ML model | Cox-derived score | Cox-derived score | |
| 4 groups | 4 groups | |||
| Training set (n = 1763)∗ | 7 210 | 7 210 | 7208 | 7 214 |
| Test set (n = 566)∗ | 2 006 | 2 002 | 2008 | 2 006 |
| ML model | ML model | CIBMTR model | ||
| 4 groups | 3 groups | |||
| Training set (n = 1925)† | 8 738 | 8 738 | 8 762 | |
| Test set (n = 618)† | 2 154 | 2 160 | 2 182 | |
Analyses performed on the subset of patients with complete data on the variables included in the Cox-derived score and the CIBMTR model to minimize biases.
Higher C-index values reflect better model performance in ranking predictions whereas lower AIC values indicate a better fit to the data.
Cox-derived score.
CIBMTR model.