Validation of Injury Prediction Model of Japan and URGENCY of United States

Abstract

Advanced Automatic Collision Notification (AACN) is a notification system that uses data about previous accidents to predict the extent of passenger injuries. The purpose of this study is to compare injury prediction models: the Honda-Nihon University model of Japan and the URGENCY model of the United States. These models are verified using 116 cases of Japanese in-depth accident data collected by the Nihon University and Chiba Hokusoh Hospital of the Nippon Medical School. Receiver operating characteristic (ROC) curves were generated by evaluating serious injury risks from the in-depth accident data, and the best models were selected to maximize the area under the curve (AUC) of ROC. In this result, the full model of the Honda-Nihon University model was the best model for predicting injuries during accidents in Japan, (AUC = 0.7699).