Multinomial and ordered logistic regressions to investigate factors associated with accidents at junctions in United Kingdom
Introduction: The safety of vehicle occupant is a major concern in United Kingdom. Despite the considerable number of fatalities and injuries in traffic crashes, a limited number of studies focused on crash severity at junctions in the UK.
Methods: This study investigates and identifies the influential factors determining the severity of injuries in traffic crashes at junctions in the UK by using crash data in the first half of 2023. An unordered multinomial logit model and an ordered logit model are developed for this purpose.
Results: Crashes at give way / uncontrolled junctions, on dark streets (where streetlights present or unknown, compared with Daylight), with higher class and number of second road, higher speed limit, which has no human control in crossing within 50 meters are associated with a higher probability of Severe and Fatal injuries. Crashes whose road surface conditions is flood (surface water over 3cm deep), at junctions more than four arms (not RAB) / Crossroads / Other junctions can improve probability of Fatal injuries compared with wet streets and T or staggered junction respectively. Crashes at Crossroads / Slip Road junction, which is Control only by Stop sign, can increase probability of Serious injuries compared with T or staggered junction and Give way or uncontrolled junctions respectively. Concerning other conditions, no significant relationship is found between injury severity and road type / day of week / weather conditions / special conditions at site / carriageway hazards / first road class. Comparison of the model results illustrated that the MNL model is slightly better fitted on the data than the ordered logit model, and they both contain the same 12 significant predictors (the OL best model has 1 more: first road number), but the conclusions inferred from these two models are generally similar. Besides, the OL best model violates the parallel regression assumption, hence this model can lead to incorrect or misleading results.
Practical Application: To reduce the injuries of crashes, we recommend improving lighting conditions and sidewalk design, implementing speed reduction strategies, introducing more human control at pedestrian crossings, cleaning up standing water at complex junctions, setting more dynamic controls (human with automatic traffic signal) at junctions, installing warning signs to drivers.
