OPTIMIZING EMERGENCY DETECTION OF BLUNT TRAUMATIC AORTIC INJURY.

Abstract

Blunt traumatic injury to the thoracic aorta is a common and deadly consequence of motor vehicle collisions and falls from significant heights. A relatively simple detection algorithm for traumatic aortic injury has been developed at Harborview Medical Center. This algorithm relies on the initial chest radiograph (CXR) to inform the decision of whether to proceed to chest computed tomographic angiogram (chest CTA) or catheter angiography (CA). In an attempt to improve this algorithm, Blackmore et al developed a clinical prediction rule to determine the probability of aortic injury based on a number of clinically evident risk factors. The probability of injury can then further inform the decision of whether to proceed to more sensitive and specific, yet more costly, invasive and time-intensive imaging modalities. The purpose of the present study was to (1) evaluate the current detection algorithm and (2) validate the previously developed clinical prediction rule. A 3-year single-institution retrospective case-control study with 25 cases of aortic injury and 181 controls was conducted. Chart review was performed for each subject, and the effectiveness and efficiency of the current imaging algorithm were determined. The detection algorithm correctly identified 96% of all cases of aortic injury. A definitive diagnosis was established within 12 hours for 88% of the patients. However, only 1 in 400 CXRs lead to a positive diagnosis of aortic injury, suggesting that much energy is invested in screening an injury that is relatively rare. Compared with the previous clinical prediction rule only four of the seven composite risk factors were significant predictors of aortic injury. The probability of injury ranged from 0 with no risk factors to 1.8% with three or four risk factors. Since this is not a clinically useful range of probabilities, the previously developed clinical prediction rule cannot be used to stratify patients according to probability of injury and thus cannot be used to inform the current detection algorithm. Any future clinical prediction rule that is developed must either consider different clinical factors or include a greater number of cases to increase the power of the study in the hope of elucidating a greater number of significant predictors. Until that time, the CXR-based detection algorithm appears to be a highly effective, albeit marginally efficient, method for detecting traumatic aortic injury.