Pedestrian injury severity in motor vehicle crashes: An integrated spatio-temporal modeling approach

This study employs an integrated spatio-temporal modeling approach to untangle the crashed injury correlates that may vary across the space and time domain. Specifically, this study employs Geographically and Temporally Weighted Ordinal Logistic Regression (GTWOLR) to examine the correlates of pedestrian injury severity in motor vehicle crashes. The method leverages the space- and time-referenced crash data and powerful computational tools. This study performed non-stationarity tests to verify whether the local correlates of pedestrian injury severity have a significant spatio-temporal variation. Results showed that some variables passed the tests, indicating they have a significantly varying spatio-temporal relationship with the pedestrian injury severity. These factors include the pedestrian age, pedestrian position, crash location, motorist age and gender, driving under the influence (DUI), motor vehicle type and crash time in a day. The spatio-temporally varying correlates of pedestrian injury severity are valuable for researchers and practitioners to localize pedestrian safety improvement solutions in North Carolina. For example, in near future, special attention may be paid to DUI crashes in the city of Charlotte and Asheville, because in such areas DUI-involved crashes are even more likely to cause severe pedestrian injuries that in other areas. More implications are discussed in the paper.
Source: Accident Analysis and Prevention - Category: Accident Prevention Source Type: research