BACKGROUND: Traumatic blunt cerebrovascular injury (BCVI) may portend catastrophic complications if untreated. Who should be screened for BCVI is controversial. The purpose of this study was to develop and validate a prediction score (pBCVI) to identify those at sufficient risk to warrant dedicated screening. METHODS: We conducted a cohort study using data for years 2002-2007 from the National Trauma Data Bank. Blunt trauma patients aged 16 years and older were randomly divided into two groups for score creation and validation. Final prediction model included age, sex, Trauma Mortality Prediction Model p(death), traumatic intracranial hemorrhage, cerebellar/brain stem injury, malar/maxillary fracture, mandible fracture, cervical spine fracture, cervical spinal cord injury, thoracic spinal cord injury, and chest Abbreviated Injury Scale ≥3. pBCVI was evaluated using receiver operating characteristic curve area and the Hosmer-Lemeshow statistic. The Youden Index estimated an optimal cut-point (J) of the pBCVI. RESULTS: The cohort numbered 1,398,310 patients, including 2,125 with BCVI. The overall incidence of BCVI was 0.15%. Cervical spine fracture had the strongest association with BCVI (odds ratio 4.82, p < 0.001). The receiver operating characteristic curve for pBCVI was 0.93 and the Hosmer-Lemeshow statistic was 206.3, p < 0.01. The optimal cut-point (J) of pBCVI was 0.0013 (sensitivity 0.91, specificity 0.82) and would miss 186 (8.8%) injuries in our cohort. To identify all BCVI using this model, an unrealistic 96% of the cohort would require screening. CONCLUSIONS: A model based on a pattern of other injuries cannot be used as a stand-alone instrument to determine screening for BCVI. "Optimal" model cut-points are not ideal for all injuries. Clinical suspicion that integrates energy of mechanism and associated injuries remains essential to effectively screen for BCVI and minimize patient risk for a catastrophic missed injury.
BACKGROUND:Traumatic blunt cerebrovascular injury (BCVI) may portend catastrophic complications if untreated. Who should be screened for BCVI is controversial. The purpose of this study was to develop and validate a prediction score (pBCVI) to identify those at sufficient risk to warrant dedicated screening. METHODS: We conducted a cohort study using data for years 2002-2007 from the National Trauma Data Bank. Blunt traumapatients aged 16 years and older were randomly divided into two groups for score creation and validation. Final prediction model included age, sex, Trauma Mortality Prediction Model p(death), traumatic intracranial hemorrhage, cerebellar/brain stem injury, malar/maxillary fracture, mandible fracture, cervical spine fracture, cervical spinal cord injury, thoracic spinal cord injury, and chest Abbreviated Injury Scale ≥3. pBCVI was evaluated using receiver operating characteristic curve area and the Hosmer-Lemeshow statistic. The Youden Index estimated an optimal cut-point (J) of the pBCVI. RESULTS: The cohort numbered 1,398,310 patients, including 2,125 with BCVI. The overall incidence of BCVI was 0.15%. Cervical spine fracture had the strongest association with BCVI (odds ratio 4.82, p < 0.001). The receiver operating characteristic curve for pBCVI was 0.93 and the Hosmer-Lemeshow statistic was 206.3, p < 0.01. The optimal cut-point (J) of pBCVI was 0.0013 (sensitivity 0.91, specificity 0.82) and would miss 186 (8.8%) injuries in our cohort. To identify all BCVI using this model, an unrealistic 96% of the cohort would require screening. CONCLUSIONS: A model based on a pattern of other injuries cannot be used as a stand-alone instrument to determine screening for BCVI. "Optimal" model cut-points are not ideal for all injuries. Clinical suspicion that integrates energy of mechanism and associated injuries remains essential to effectively screen for BCVI and minimize patient risk for a catastrophic missed injury.