Citation

Hasanat-E-Rabbi S, Raihan MA, Mahmud SMS, Hoque MS. IATSS Res. 2022; 46(2): 269-280.

Copyright

(Copyright © 2022, International Association of Traffic and Safety Sciences, Publisher Elsevier Publishing)

DOI

10.1016/j.iatssr.2022.01.002

PMID

unavailable

Abstract

Pedestrian fatality and injury is one of the most concerning issues around the globe. The predictors for such mishaps have been investigated in the developed countries through econometric models and are proven useful techniques. Such studies in the context of developing countries, especially for urban cities, are however still very scarce. Using five years reported pedestrian crash data, this study looks into the performance of three statistical models - Multinomial Logit (MNL), Ordered Logit (OL) and Partial Proportional Odds (PPO) model while examining the impact of various attributes related to pedestrian crashes severity outcomes for Dhaka metropolitan city in Bangladesh. The comparative analysis reveals that the performance of the PPO model is relatively better for the available dataset in terms of identifying critical risk factors. Undivided roadway, heavy vehicles, unfit vehicles, adult drivers with no seat belt use, young and older pedestrians, pedestrian road crossing action are found to be associated with higher probability of fatal injuries. In contrast, one-way traffic movement, daytime, motorcycles and mid-aged pedestrians decrease the likelihood of fatal injury. Based on these identified risk factors, a combined 3-E approach has been suggested to reduce the severity levels of pedestrian in the event of crash occurrence.

Language: en

Keywords

Developing country; Injury severity; Multinomial logit; Ordered logit; Partial proportional odds; Pedestrian