<p>DMP-2016.</p>

This study explores the demographic and behavioral determinants of road traffic accident (RTA) severity in the context of the Dhaka metropolitan area, Bangladesh. Road crash data recorded by the Dhaka Metropolitan Police (DMP) were analyzed through ordered logistic regression and generalized ordered logistic regression. The results were interpreted using log odds ratios, odds ratios, predicted probabilities, and marginal effects. The findings reveal that young and middle-aged drivers exhibit significantly higher odds of severe crashes compared to underage drivers. Young-aged drivers are 14 percentage points more likely to cause fatal crashes when compared to old aged drivers in our ordered logit model. In addition, male drivers show higher odds of severe crashes than females. Factors such as overloading of vehicles, alcohol consumption while driving, and over-speeding were identified as the major contributors to increasing crash severity. Alcohol consumption had an odds ratio of 1.223 in the ordered logit model, and it had odds ratios of 2.418, 1.722, and 1.086 for the thresholds of motor collision, simple injury, and grievous injury, respectively, in the generalized ordered logit model. In contrast, the use of seatbelts, vehicle fitness maintenance, and drivers’ licensing shows mitigating effects on crash severity, with significant odds ratios < 1 in both the ordered logit and generalized ordered logit models. From the ordered logit model, we found that seat belt use, fitness certificate, and license decrease the likelihood of fatal crash by 10.7 percentage points, 8.2 percentage points, and 28.2 percentage points, respectively, whereas overspeed increases the likelihood of fatal crash by 13.5 percentage points. The results were reflected in the generalized ordered logit model, too. This research provides valuable insights for policymakers to design and implement effective policies and transport planning, including demographic driving regulations and behavioral control mechanisms to reduce road crash severity.本研究以孟加拉国达卡都会区为研究场景，探究道路交通事故（Road Traffic Accident，简称RTA）严重程度的人口统计学与行为学影响因素。本研究采用有序逻辑回归（ordered logistic regression）与广义有序逻辑回归（generalized ordered logistic regression）方法，对达卡大都会警察局（Dhaka Metropolitan Police，简称DMP）记录的道路碰撞事故数据展开分析，并通过对数优势比、优势比、预测概率与边际效应对研究结果进行阐释。研究结果显示，与未成年驾驶员相比，青年与中年驾驶员发生严重碰撞事故的优势比显著更高。在有序logit模型中，青年驾驶员引发致命碰撞事故的概率较老年驾驶员高出14个百分点。此外，男性驾驶员发生严重碰撞事故的优势比高于女性驾驶员。车辆超载、驾驶时饮酒与超速行驶等因素，被认定为加剧事故严重程度的主要诱因。在有序logit模型中，饮酒的优势比为1.223；而在广义有序logit模型中，针对机动车碰撞、轻微受伤与重伤三个阈值，饮酒的优势比分别为2.418、1.722与1.086。与之相反，安全带使用、车辆维保与驾驶员持证等因素对事故严重程度具有缓解作用，在有序logit与广义有序logit模型中，其优势比均显著小于1。基于有序logit模型，研究发现安全带使用、车辆维保合格证明与驾驶员持证，分别可使致命碰撞事故的发生概率降低10.7个百分点、8.2个百分点与28.2个百分点；而超速行驶则会使致命碰撞事故的发生概率提升13.5个百分点。上述结果在广义有序logit模型中同样得到验证。本研究可为政策制定者设计并推行有效的政策与交通规划方案提供重要参考，包括制定人口统计学相关的驾驶规制与行为管控机制，以降低道路碰撞事故的严重程度。