Predictors for the selection of patients for abdominal CT after blunt trauma: a proposal for a diagnostic algorithm.

Objective: To select parameters that can predict which patients should receive abdominal computed tomography (CT) after high-energy blunt trauma.

Summary Background Data: Abdominal CT accurately detects injuries of the abdomen, pelvis, and lumbar spine, but has important disadvantages. More evidence for an appropriate patient selection for CT is required.

Methods: A prospective observational study was performed on consecutive adult high-energy blunt trauma patients. All patients received primary and secondary surveys according to the advanced trauma life support, sonography (focused assessment with sonography for trauma [FAST]), conventional radiography (CR) of the chest, pelvis, and spine and routine abdominal CT. Parameters from prehospital information, physical examination, laboratory investigations, FAST, and CR were prospectively recorded for all patients. Independent predictors for the presence of > or =1 injuries on abdominal CT were determined using a multivariate logistic regression analysis.

Results: A total of 1040 patients were included, 309 had injuries on abdominal CT. Nine parameters were independent predictors for injuries on CT: abnormal CR of the pelvis (odds ratio [OR], 46.8), lumbar spine (OR, 16.2), and chest (OR, 2.37), abnormal FAST (OR, 26.7), abnormalities in physical examination of the abdomen/pelvis (OR, 2.41) or lumbar spine (OR 2.53), base excess <-3 (OR, 2.39), systolic blood pressure <90 mm Hg (OR, 3.81), and long bone fractures (OR, 1.61). The prediction model based on these predictors resulted in a R of 0.60, a sensitivity of 97%, and a specificity of 33%. A diagnostic algorithm was subsequently proposed, which could reduce CT usage with 22% as compared with a routine use.

Conclusions: Based on parameters from physical examination, laboratory, FAST, and CR, we created a prediction model with a high sensitivity to select patients for abdominal CT after blunt trauma. A diagnostic algorithm was proposed.