Examining driver injury severity in intersection-related crashes using cluster analysis and hierarchical Bayesian...

Area of Science:

• Traffic Safety Research
• Transportation Engineering
• Accident Analysis

Background:

• Intersection crashes present complex traffic environments, increasing accident likelihood.
• Driver behavior and injury severity patterns in these crashes require detailed examination.

Purpose of the Study:

• To investigate driver injury severity in intersection-related crashes using a hybrid modeling approach.
• To identify key contributing factors influencing injury outcomes across different environmental conditions.

Main Methods:

• Utilized K-means cluster analysis to group crash data based on weather and roadway conditions.
• Developed hierarchical Bayesian random intercept models for injury severity analysis within clusters.
• Compared the hybrid approach with a standard multinomial logistic model.

Main Results:

• Identified significant crash-level (e.g., time, weather, light, area, road grade) and vehicle/driver-level (e.g., traffic controls, vehicle action, seatbelt use, impairment) factors.
• Revealed significant cross-level interactions, such as left turns at night and impairment in dark conditions.
• The hybrid approach demonstrated superior suitability and effectiveness compared to traditional models.

Conclusions:

• The study provides a nuanced understanding of driver injury severity in intersection crashes.
• Findings offer valuable insights for developing targeted countermeasures to enhance traffic safety.