An independent, external validation and component analysis of the Surviving Penetrating Injury to the Brain score for civilian cranial gunshot injuries.

Objective: The Surviving Penetrating Injury to the Brain (SPIN) score utilizes clinical variables to estimate in-hospital and 6-month mortality for patients with civilian cranial gunshot wounds (cGSWs) and demonstrated good discrimination (area under the receiver operating characteristic curve [AUC] 0.880) in an initial validation study. The goal of this study was to provide an external, independent validation of the SPIN score for in-hospital and 6-month mortality.
Methods: To accomplish this, the authors retrospectively reviewed 6 years of data from their institutional trauma registry. Variables used to determine SPIN score were collected, including sex, transfer status, injury motive, pupillary reactivity, motor component of the Glasgow Coma Scale (mGCS), Injury Severity Score (ISS), and international normalized ratio (INR) at admission. Multivariable logistic regression analysis identified variables associated with mortality. The authors compared AUC between models by using a nonparametric test for equality.
Results: Of the 108 patients who met the inclusion criteria, 101 had all SPIN score components available. The SPIN model had an AUC of 0.962. The AUC for continuous mGCS score alone (0.932) did not differ significantly from the AUC for the full SPIN model (p = 0.26). The AUC for continuous mGCS score (0.932) was significantly higher compared to categorical mGCS score (0.891, p = 0.005). Use of only mGCS score resulted in fewer exclusions due to missing data. No additional variable included in the predictive model alongside continuous mGCS score was a significant predictor of inpatient mortality, 6-month mortality, or increased model discrimination.
Conclusions: Given these findings, continuous 6-point mGCS score may be sufficient as a generalizable predictor of inpatient and 6-month mortality in patients with cGSW, demonstrating excellent discrimination and reduced bias due to missing data.