Comparing model performance of the injury prediction models
| Model performance metric | Original model (M1) | Original model with linearity assumed for continuous predictors (M2) | Original model with continuous predictors and non-linear transformations (M3) | Original model with further included predictors that are dichotomised (M4) | Predictor selection based on rationale from literature and clinical reasoning and kept continuous with non-linear transformations (M5) |
| Discrimination† | 0.63 (0.60 to 0.66) | 0.89 (0.87 to 0.91) | 0.90 (0.88 to 0.92) | 0.63 (0.60 to 0.66) | 0.90 (0.87 to 0.93) |
| Calibration in the Large | −0.11 (−0.22 to 0.00) | −0.02 (−0.17 to 0.13) | 0.03 (−0.13 to 0.19) | −0.13 (−0.25 to –0.01) | 0.04 (−0.12 to 0.20) |
| Calibration slope | 0.84 (0.61 to 1.07) | 0.97 (0.86 to 1.08) | 0.92 (0.76 to 1.10) | 0.81 (0.59 to 1.03) | 0.87 (0.72 to 1.02) |
| R2 | 0.07 (0.05 to 0.09) | 0.56 (0.48 to 0.64) | 0.64 (0.58 to 0.70) | 0.08 (0.05 to 0.11) | 0.63 (0.56 to 70) |
| Brier Score | 0.21 | 0.12 | 0.11 | 0.21 | 0.12 |
Discrimination, calibration in the large, calibration slope and R2 are reported with 95% CIs.
*All model performance is reported following 2000 bootstraps.
†Discrimination is reported as area under the curve where 0.50=no discrimination and 1.00=perfect discrimination.