Analysis of the Elderly Pedestrian Injury Severity in Urban Traffic Accidents in Spain using Machine Learning Techniques

Abstract

Walking is an essential activity for a healthy lifestyle. In urbanized areas, the risk of suffering a traffic accident as a pedestrian is a matter of concern, and this situation has been widely studied. Elderly pedestrians are vulnerable road users due to their fragility and loss of physical faculties, and the probability of elderly pedestrians being killed or seriously injured in a traffic accident is higher than for the rest of the pedestrians. Furthermore, the elderly population is expected to increase in the coming decades. Hence, the study of these situations is necessary to build safer, healthier, and more sustainable cities. Little research has been devoted to the study of the relationship between injury severity elderly pedestrians and the characteristics of the accident location. The objective of this study is to investigate the influence of accident and built environment factors on the severity of elderly pedestrian urban traffic accidents in Spain. For this purpose, logistic regression, and random forest models together with data resampling techniques were used to analyze the injury severity of vehicle collisions suffered by elderly pedestrians in Spain from 2016 to 2019, and its link with accident and built environment features. As expected, the random forest outperformed the logistic regression performance for both samples, but the logistic regression results are easier to interpret. Results showed the influence of both accident and built environment variables in the injury severity, being accidents in more populated cities less severe for all pedestrians. In addition, the age of the pedestrian, which indicates the need to study this age group in a more disaggregated manner, and accidents in dark spots artificially lighted, which suggests that those locations are not properly lighted for the elderly, are key factors for the injury level of elderly pedestrian traffic accidents.