Statistical validity and clinical merits of a new civilian gunshot injury classification

Despite the high prevalence of civilian gunshot injuries (GSIs) in the United States, no universally accepted classification currently exists. Recently, two of us (ZG, RWL) proposed a GSI classification based on energy transferred, vital structure damage, wound characteristics, fracture, and degree of contamination. This classification has not been validated in a clinical setting.

We determined the feasibility, internal consistency, and predictive accuracy of this classification.

We reviewed the medical records of 216 patients with 264 GSIs treated at a Level I trauma center. Feasibility was determined by the investigators' ability to retrospectively complete the classification system based on patient information routinely collected in medical records. Internal consistency was determined using Cronbach's coefficient alpha. Predictive accuracy was constructed and interpreted in a receiver operating characteristic (ROC) curve using all the classification components to predict GSI severity. The clinical management/outcome (deceased, hospitalization versus nonadmission, and surgical versus nonsurgical treatment) was used as a proxy measure of GSI severity.

We were able to apply the classification to 82% of charts we reviewed. The classification components appeared to be internally consistent (Cronbach's alpha was 0.69 and was increased to 0.78 after exclusion of contamination). Each component was associated with clinical management. GSI classified as high energy, worse vital structure, and high contamination had higher rates of surgery (84%, 84%, and 100%, respectively). The area under the ROC curve was 0.80, suggesting the classification can accurately describe GSI severity.

Our results suggest this new civilian GSI classification is statistically valid and has clinical merits warranting further investigation in the setting of a prospective trial.