Abstract
Previous studies have demonstrated that patients with traumatic brain injury (TBI) who also have progressive hemorrhagic injury (PHI), have a higher risk of clinical deterioration and worse outcomes than do TBI patients without PHI. Therefore, the early prediction of PHI occurrence is useful to evaluate the status of patients with TBI and to improve outcomes. The objective of this study was to develop and validate a prognostic model that uses information available at admission to determine the likelihood of PHI after TBI. Retrospectively collected data were used to develop a PHI prognostic model with a logistic regression analysis. The prediction model was validated in 114 patients from a separate hospital. Eight independent prognostic factors were identified: age ≥ 57 years (5 points), intra-axial bleeding/brain contusion (4 points), midline shift ≥ 5 mm (6 points), platelet (PLT) count<100×10⁹/L (10 points), PLT count ≥ 100 but <150×10⁹/L (4 points), prothrombin time>14 sec (7 points), D-dimer ≥ 5 mg/L (12 points), and glucose ≥ 10 mmol/L (10 points). Each patient was assigned a number of points proportional to the regression coefficient. We calculated risk scores for each patient and defined three risk groups: low risk (0-13 points), intermediate risk (14-22 points), and high risk (23-54 points). In the development cohort, the PHI rates after TBI for these three groups were 10.3%, 47.3%, and 85.2%, respectively. In the validation cohort, the corresponding PHI rates were 10.9%, 47.3%, and 86.9%. The C-statistic for the point system was 0.864 (p=0.509 by the Hosmer-Lemeshow test) in the development cohort, and 0.862 (p=0.589 by the Hosmer-Lemeshow test) in the validation cohort. In conclusion, a relatively simple risk score using admission predictors accurately predicted the risk for PHI after TBI.