Development and Validation of a Novel Classification System and Prognostic Model for Open Traumatic Brain Injury: A Multicenter Retrospective Study.

INTRODUCTION

Open traumatic brain injury (OTBI) is associated with high mortality and morbidity; however, the classification of these injuries and the determination of patient prognosis remain uncertain, hindering the selection of optimal treatment strategies. This study aimed to develop and validate a novel OTBI classification system and a prognostic model for poor prognosis.

METHODS

This retrospective study included patients with isolated OTBI who received treatment at three large medical centers in China between January 2020 and June 2022 as the training set. Data on patients with OTBI collected at the Fuzong Clinical Medical College of Fujian Medical University between July 2022 and June 2023 were used as the validation set. Clinical parameters, including clinical data at admission, radiological and laboratory findings, details of surgical methods, and prognosis were collected. Prognosis was assessed through a dichotomized Glasgow Outcome Scale (GOS). A novel OTBI classification was proposed, categorizing patients based on a combination of intracranial hematoma and midline shift observed on imaging, and logistic regression analyses were performed to identify risk factors associated with poor prognosis and to investigate the association between the novel OTBI classification and prognosis. Finally, a nomogram suitable for clinical application was established and validated.

RESULTS

Multivariable logistic regression analysis identified OTBI classification type C (p < 0.001), a Glasgow Coma Scale score (GCS) ≤ 8 (p < 0.001), subarachnoid hemorrhage (SAH) (p = 0.004), subdural hematoma (SDH) (p = 0.011), and coagulopathy (p = 0.020) as independent risk factors for poor prognosis. The addition of the OTBI classification to a model containing all the other identified prognostic factors improved the predictive ability of the model (Z = 1.983; p = 0.047). In the validation set, the model achieved an area under the curve (AUC) of 0.917 [95% confidence interval (CI) = 0.864-0.970]. The calibration curve closely approximated the ideal curve, indicating strong predictive performance of the model.

CONCLUSIONS

The implementation of our proposed OTBI classification system and its use alongside the other prognostic factors identified here may improve the prediction of patient prognosis and aid in the selection of the most suitable treatment strategies.