In an effort to better understand thoracic trauma in frontal impacts, seventy-one frontal impact sled tests were conducted using post-mortem human subjects in the driver’s position. Various contemporary automotive restraint systems were used in these tests. The post-mortem subjects were instrumented with accelerometers and chest bands to characterize their mechanical response during the impact. The resulting injury from the impact was determined through radiography and detailed autopsy and its severity was coded according to the Abbreviated Injury Scale.

The measured mechanical responses were analyzed using statistical procedures. In particular, linear logistic regression was used to develop models which associate the measured mechanical parameters to the observed thoracic injury response. Univariate and multivariate models were developed taking into consideration potential confounders and effect modifiers. The risk factors used in the models were normalized with respect to the size and weight of the specimen. The gender and age of specimen at time of death were found not to be confounders in this data set. A linear combination of the 3-msec clip value of maximum resultant spine acceleration and maximum normalized chest deflection from an array of five measurements provided the best goodness of fit measure. This linear combination was found to have significantly better injury predictive ability, for thoracic trauma in human subjects under any restraint environment, than other existing injury criteria such as VC_max, chest deflection, or chest acceleration alone.