A data-driven procedure is presented to estimate the costs and the number of fatalities attributable to specific types of injuries. It continues Martin and Eppinger’s work presented at the 2003 ESV conference. The procedure examines a crash victim’s entire injury record in the process. All possible injuries are denoted by unique codes as described in the AIS Injury Coding Manual. The two most serious injuries – denoted as the primary injury and the secondary injury – are chosen from the injury record and are used to characterize a victim’s entire set of injuries. When the mortality rate of the primary injury code is combined with that of the secondary injury, an overall fatality risk is obtained. Fatalities attributable to specific injuries may then be determined by considering the effect that a specific injury or set of injuries has on fatality risk. Attributable costs are estimated in a similar manner. Ultimately, this process – which singles out specific injuries – provides a means to determine the types of injuries NHTSA should strive to prevent, and to determine the capabilities needed of a crash dummy to ascertain whether such injuries are sustainable in a crash test.