Table 4 Quality criteria for the considered scoring systems in the development dataset (TR-DGU 2010 and 2011), and for RISC II in the validation dataset (TR-DGU 2012)
Discrimination | Precision | Calibration | |
|---|---|---|---|
AUC of ROC curve (95% CI) | observed and predicted mortality | HL goodness-of-fit statistic | |
All patients, development dataset (n = 30,866) | |||
Observed mortality | 11.5% | ||
RISC II with 20% missing values per variable | 0.943 (0.939-0.946) | 11.4% | 55.2 |
RISC II with all available data | 0.947 (0.944-0.951) | 11.6% | 55.3 |
Patients with RISC, development dataset (n = 26,041) | |||
Observed mortality | 11.4% | ||
RISC II | 0.947 (0.943-0.951) | 11.6% | 50.8 |
RISC | 0.934 (0.929-0.938) | 13.5% | 233.8 |
Patients with TRISS, development dataset (n = 17,411) | |||
Observed mortality | 10.7% | ||
RISC II | 0.953 (0,949-0,957) | 11.0% | 38.3 |
RISC | 0.939 (0.933-0.944) | 12.9% | 178.6 |
TRISS | 0.917 (0.911-0.924) | 13.5% | 554.4 |
NISS | 0.849 (0.839-0.858) | ||
ISS | 0.821 (0.811-0.831) | ||
All patients, validation dataset (n = 21,918) | |||
Observed mortality | 10.9% | ||
RISC II | 0.951 (0.947-0.954) | 11.3% | 50.3 |