Table 2.
Performance of various multivariable logistic regression models to predict 6-month post-injury outcomes
Model | Favorable vs unfavorable outcome | Alive vs dead | ||||
---|---|---|---|---|---|---|
AUC (95% CI) | AIC | Nagelkerke R2 | AUC (95% CI) | AIC | Nagelkerke R2 | |
TBI-IMPACT | 0.86 (0.76–0.96) | 66.00 | 0.51 | 0.78 (0.66–0.90) | 68.37 | 0.27 |
TBI-IMPACT + % Time CPP < 60 mmHg + % Time CPP > 70 mmHg | 0.86 (0.76–0.96) | 68.81 | 0.53 | 0.81 (0.69–0.93) | 69.44 | 0.33 |
TBI-IMPACT + % Time CPP < 60 mmHg + % Time CPP > 70 mmHg + % Time PbtO2 < 20 mmHg | 0.87 (0.77–0.96) | 69.83 | 0.54 | 0.81 (0.70–0.93) | 71.15 | 0.34 |
TBI-IMPACT + % Time CPP < LLR | 0.86 (0.76–0.96) | 65.81 | 0.53 | 0.80 (0.68–0.93) | 65.85 | 0.32 |
TBI-IMPACT + % Time CPP > ULR | 0.86 (0.77–0.96) | 66.66 | 0.52 | 0.81 (0.69–0.92) | 67.40 | 0.28 |
TBI-IMPACT + % Time CPP < LLR + % Time CPP > ULR | 0.87 (0.77–0.96) | 67.77 | 0.53 | 0.80 (0.68–0.93) | 67.48 | 0.33 |
The TBI-IMPACT model included age, admission GCS, admission pupil exam, and Marshall score
AIC Akaike Information Criterion, AUC Area under the curve, CPP cerebral perfusion pressure, GCS Glasgow Coma Scale, LLR lower limit of reactivity, PbtO2 brain tissue oxygenation, ULR upper limit of reactivity