Authors: Lily H. Kim; Jennifer Quon, MD; Marco Lee, MD, PhD; Tene Cage, MD; Lan Pham; Anthony Holguin, MA; Harminder Singh, MD (Mountain View, CA)
Gunshot wound (GSW) remains one of the most lethal forms of head trauma. The lack of clear guidelines for managing civilian gunshot wounds complicates the surgical decision making process. We present a clinically applicable decision-tree model based on 15-year data from our level 1 trauma center.
We retrospectively reviewed 95 consecutive patients who presented with cerebral GSWs between 2003 and 2018. Patients were divided into two cohorts based on survival status. Clinical information captured in our trauma database, EMR, and relevant imaging scans was reviewed for each patient. A decision-tree model was constructed based on variables showing statistically significant differences (P=0.05) between two groups on chi-square test.
After excluding patients who died at arrival and/or did not undergo brain imaging, 54 patients with radiologically confirmed intracranial injury were included. Survivors (53.7%) and non-survivors (46.3%) had similar average age and gender distribution. Non-survivors were significantly more likely to have self-inflicted (P=0.049) and perforating (entry and exit wound), as opposed to penetrating (entry wound only), injuries (P=0.02). Bi-hemispheric and posterior fossa involvement, cerebral herniation, and intraventricular hemorrhage (IVH) were the radiologic features more commonly present in patients who expired. Based on the decision-tree, in patients with Glasgow Coma Scale (GCS) >8, penetrating injury limited to a single hemisphere predicted survival. When initial GCS was 8 or lower, all patients with absent pupillary response failed to survive. Among patients with pupillary response, lack of 1) posterior fossa involvement, 2) cerebral herniation, 3) bi-hemispheric injury, and 4) IVH, was associated with survival.
We present a decision-tree model to help neurosurgeons identify surgical candidates with favorable prognosis based on readily available clinical and radiological information in a time-sensitive setting. Further validation of the model in a large patient setting is recommended.