Cross-validation analysis of Models 1 and 2
| Performance of mTRISS Models 1 and 2 with coefficients derived from QEH dataset | ||||
|---|---|---|---|---|
| Model 1* | Model 2† | |||
| ROC analysis | 3-fold (average) | Full dataset | 3-fold (average) | Full dataset |
| Sensitivity % | 97.6 | 97.5 | 97.3 | 97.3 |
| Specificity % | 51.3 | 52.0 | 48.8 | 48.9 |
| Accuracy % | 91.6 | 91.6 | 91.0 | 91.1 |
| Area under the ROC (95% CI) | 0.931 (0.916 to 0.947) | 0.932 (0.923 to 0.940) | 0.919 (0.901 to 0.936) | 0.920 (0.910 to 0.930) |
The performance on PS in Model 1 and Model 2 based on the coefficients derived from the threefold cross-validation and that derived from the full dataset were compared using the sensitivity, specificity, accuracy and area under the ROC curve.
*Age=scale variable.
†Age into two groups (0–54, 55+) that is, MTOS TRISS with application of QEH dataset-derived coefficients.
Full dataset, performance measures derived from full local dataset of QEH; MTOS, Major Trauma Outcome Study; ROC, Receiver operator characteristic; QEH, The Queen Elizabeth Hospital; TRISS, Trauma and Injury Severity Score; threefold (average), average performance measures derived from threefold cross-validation statistics.