Creating and validating a neurosurgical intervention rule-out tool for patients with mild traumatic brain injury and isolated subdural hematoma: a 5-year, six-center retrospective cohort study.

Objective: Because there is no reliable method on admission to predict whether a patient will require neurosurgical intervention in the future, the general approach remains to treat each patient with mild traumatic brain injury (mTBI) and subdural hematoma (SDH) as if they will require such an intervention. Consequently, there is a growing population of patients with mTBI and SDH that is overtriaged despite having a low probability of needing neurosurgical intervention. This study aimed to train and validate a predictive rule-out tool for neurosurgical intervention in patients with mTBI and SDH.

Methods: This was a retrospective cohort study of all trauma patients admitted to six level I trauma centers in three states. Patients were included if they met the following criteria: admitted between 2016 and 2020, ≥ 18 years of age, ICD-10 diagnosis of isolated SDH, initial head imaging available, initial Glasgow Coma Scale score of 13-15, and arrived within 48 hours of injury. Exclusion criteria included skull fracture, intracranial hemorrhage other than an SDH, and no neurosurgical consultation. Prediction variables included 34 demographic, clinical, and radiographic variables. The study outcome was neurosurgical intervention within 48 hours of hospital admission. Seventy-five percent of the data were used for training, and 25% for testing. Multivariable logistic regression with fivefold cross-validation was used on the training set to identify covariates with the highest specificity while holding sensitivity at 100%. Results were validated on the testing set.

Results: In total, 1000 patients were in the training set and 333 in the testing set. The overall neurosurgical intervention rate was 8.8%. For the fivefold cross-validation process, three variables were selected that maximized specificity while holding sensitivity at 100%: maximum hematoma thickness, initial Glasgow Coma Scale score, and preinjury antithrombotic use (sensitivity 100%, specificity 56%, area under the receiver operating characteristic curve 0.94). With a cutoff probability of neurosurgical intervention set at 1.88%, the final model was validated to predict neurosurgical intervention with a sensitivity of 100% (95% CI 88.4%-100%) and specificity of 55.1% (95% CI 49.3%-60.8%).

Conclusions: In this study, the largest of its kind to date, the authors successfully developed and validated a new tool for ruling out the necessity of neurosurgical intervention in patients with mTBI and isolated SDH. By successfully identifying more than half of patients who are unlikely to require neurosurgery within the first 2 days of admission, this tool can be used to improve treatment efficiency and provide patients and clinicians with valuable prognostic information.