Injury severity score (ISS) is commonly used in trauma registries to describe injury severity and to predict outcomes in trauma patients regardless of injury mechanism. This study examined the correlation between ISS and mortality in adult trauma patients presenting to emergency departments in the United States with different mechanisms of injury.
A retrospective observational study was conducted using the 2014 Nationwide Emergency Department Sample. Patients’ characteristics were stratified by mortality. Receiver operating characteristic (ROC) curves were generated for death against ISS for each mechanism of injury. A logistic regression model was conducted for each mechanism of injury to determine whether ISS (≥16 vs <16) is a predictor of mortality.
The study sample consisted of 16,147,058 weighted adult trauma patients. Median age was 46 years. Slightly over half were females (51.9%). Falls, motor vehicle accidents and being struck by or against, were the most commonly reported mechanisms of injury (44.6%, 18.1%, and 15.3%, respectively). The overall mortality in the study population was 0.4%. The area under the ROC curve was highest in injuries sustained in accidents involving machinery (0.947; 95% confidence intervals [CI], 0.896-0.998), followed by motor vehicle traffic (MVA) (0.788; 95% CI, 0.775-0.801) and cutting or piercing (0.746; 95% CI, 0.701-0.791). Deceased patients were accurately identified by ISS 65.2% in injury by machinery, 47.7% in injury involving MVA, 39.7% in injury by firearm and 31.4% in injury by assault. After adjusting for confounders, the multivariate models in which ISS was the main independent factor performed best in predicting mortality from firearm and machinery mechanism of injuries.
Although the ROC curve analysis demonstrated a moderate or high discriminatory ability to identify deceased patients in 6 out of twelve mechanisms, and the multivariate analysis revealed that ISS was a significant predictor of mortality in 9 out of 12 injury mechanisms, the sensitivities of all logistic regression models were poor. The ISS ≥ 16 threshold alone therefore should not be used to identify patients with high-mortality risk. The mortality risk assessment should be done individually and be based on clinical evaluation.