Computational methodology to predict injury risk for motor vehicle crash victims: A framework for improving Advanced Automatic Crash Notification systems

Abstract

Advanced Automatic Crash Notification (AACN) systems, capable of predicting post-crash injury severity and subsequent automatic transfer of injury assessment data to emergency medical services, may significantly improve the timeliness, appropriateness, and efficacy of care provided. The estimation of injury severity based on statistical field data, as incorporated in current AACN systems, lack specificity and accuracy to identify the risk of life-threatening conditions. To enhance the existing AACN framework, the goal of the current study was to develop a computational methodology to predict risk of injury in specific body regions based on specific characteristics of the crash, occupant and vehicle. The computational technique involved multibody models of the vehicle and the occupant to simulate the case-specific occupant dynamics and subsequently predict the injury risk using established physical metrics. To demonstrate the computational-based injury prediction methodology, three frontal crash cases involving adult drivers in passenger cars were extracted from the US National Automotive Sampling System Crashworthiness Data System. The representative vehicle model, anthropometrically scaled model of the occupant and kinematic information related to the crash cases, selected at different severities, were used for the blinded verification of injury risk estimations in five different body regions. When compared to existing statistical algorithms, the current computational methodology is a significant improvement toward post-crash injury prediction specifically tailored to individual attributes of the crash. Variations in the initial posture of the driver, analyzed as a pre-crash variable, were shown to have a significant effect on the injury risk.