Comparison of the Integrated Calibration Index and the median of the absolute difference between the predicted survival probabilities and smoothed survival frequencies for each prediction model
. | Integrated calibration index (EC50) . | ||||
---|---|---|---|---|---|
Risk category . | GRFS . | OS . | Relapse . | aGVHD . | cGVHD . |
Cox-PH | 0.139 (0.151) | 0.283 (0.248) | 0.055 (0.029) | 0.218 (0.212) | 0.263 (0.208) |
Random Survival Forest | 0.142 (0.147) | 0.365 (0.372) | 0.048 (0.029) | 0.173 (0.178) | 0.345 (0.346) |
XGBoost | 0.027 (0.007) | 0.393 (0.381) | 0.176 (0.163) | 0.265 (0.264) | 0.306 (0.265) |
Gradient Boosting | 0.050 (0.047) | 0.438 (0.449) | 0.159 (0.129) | 0.254 (0.256) | 0.309 (0.275) |
Component-wise Gradient Boosting | 0.061 (0.068) | 0.397 (0.395) | 0.171 (0.145) | 0.261 (0.264) | 0.324 (0.318) |
Dynamic DeepHit | 0.054 (0.059) | 0.405 (0.409) | 0.152 (0.153) | 0.106 (0.108) | 0.319 (0.320) |
Stacked Ensemble Model | 0.023 (0.017) | 0.210 (0.194) | 0.044 (0.018) | 0.017 (0.018) | 0.258 (0.226) |
. | Integrated calibration index (EC50) . | ||||
---|---|---|---|---|---|
Risk category . | GRFS . | OS . | Relapse . | aGVHD . | cGVHD . |
Cox-PH | 0.139 (0.151) | 0.283 (0.248) | 0.055 (0.029) | 0.218 (0.212) | 0.263 (0.208) |
Random Survival Forest | 0.142 (0.147) | 0.365 (0.372) | 0.048 (0.029) | 0.173 (0.178) | 0.345 (0.346) |
XGBoost | 0.027 (0.007) | 0.393 (0.381) | 0.176 (0.163) | 0.265 (0.264) | 0.306 (0.265) |
Gradient Boosting | 0.050 (0.047) | 0.438 (0.449) | 0.159 (0.129) | 0.254 (0.256) | 0.309 (0.275) |
Component-wise Gradient Boosting | 0.061 (0.068) | 0.397 (0.395) | 0.171 (0.145) | 0.261 (0.264) | 0.324 (0.318) |
Dynamic DeepHit | 0.054 (0.059) | 0.405 (0.409) | 0.152 (0.153) | 0.106 (0.108) | 0.319 (0.320) |
Stacked Ensemble Model | 0.023 (0.017) | 0.210 (0.194) | 0.044 (0.018) | 0.017 (0.018) | 0.258 (0.226) |
EC50, the median of the absolute difference between the predicted survival probabilities and smoothed survival frequencies.