Identifying interactions among factors related to death occurred at the scene of traffic accidents: Application of "logic regression" method
Milad Jamali-Dolatabad,Homayoun Sadeghi-Bazargani,Saman Salemi,Parvin Sarbakhsh
DOI: https://doi.org/10.1016/j.heliyon.2024.e32469
IF: 3.776
2024-06-05
Heliyon
Abstract:Aim: Traffic accidents are caused by several interacting risk factors. This study aimed to investigate the interactions among risk factors associated with death at the accident scene (DATAS) as an indicator of the crash severity, for pedestrians, passengers, and drivers by adopting "Logic Regression" as a novel approach in the traffic field. Method: A case-control study was designed based on the police data from the Road Traffic Injury Registry in northwest of Iran during 2014-2016. For each of the pedestrians, passengers, and drivers' datasets, logic regression with "logit" link function was fitted and interactions were identified using Annealing algorithm. Model selection was performed using the cross-validation and the null model randomization procedure. Results: regarding pedestrians, "The occurrence of the accident outside a city in a situation where there was insufficient light" (OR = 6.87, P-value<0.001) and "the age over 65 years" (OR = 2.97, P-value<0.001) increased the chance of DATAS. "Accidents happening in residential inner-city areas with a light vehicle, and presence of the pedestrians in the safe zone or on the non-separate two-way road" combination lowered the chance of DATAS (OR = 0.14, P-value<0.001). For passengers, "Accidents happening in outside the city or overturn of the vehicle" combination (OR = 8.55, P-value<0.001), and "accidents happening on defective roads" (OR = 2.18, P-value<0.001) increased the odds of DATAS; When "driver was not injured or the vehicle was two-wheeled", chance of DATAS decreased for passengers (OR = 0.25, p-value<0.001). The odds of DATAS were higher for "drivers who had a head-on accident, or drove a two-wheeler vehicle, or overturned the vehicle" (OR = 4.03, P-value<0.001). "Accident on the roads other than runway or the absence of a multi-car accident or an accident in a non-residential area" (OR = 6.04, P-value<0.001), as well "the accident which occurred outside the city or on defective roads, and the drivers were male" had a higher risk of DATAS for drivers (OR = 5.40, P-value<0.001). Conclusion: By focusing on identifying interaction effects among risk factors associated with DATAS through logic regression, this study contributes to the understanding of the complex nature of traffic accidents and the potential for reducing their occurrence rate or severity. According to the results, the simultaneous presence of some risk factors such as the quality of roads, skill of drivers, physical ability of pedestrians, and compliance with traffic rules play an important role in the severity of the accident. The revealed interactions have practical significance and can play a significant role in the problem-solving process and facilitate breaking the chain of combinations among the risk factors. Therefore, practical suggestions of this study are to control at least one of the risk factors present in each of the identified combinations in order to break the combination to reduce the severity of accidents. This may have, in turn, help the policy-makers, road users, and healthcare professionals to promote road safety through prioritizing interventions focusing on effect size of simultaneous coexistence of crash severity determinants and not just the main effects of single risk factors or their simple two-way interactions.