Crash involvement of drivers with multiple crashes

Area of Science:

• Road safety
• Traffic accident analysis
• Predictive modeling

Background:

• Licensing agencies aim to identify high-risk drivers to prevent future accidents.
• Kentucky data reveals a significant number of drivers are repeat offenders in crashes.
• Existing methods may not adequately predict a driver's likelihood of future at-fault crashes.

Purpose of the Study:

• To develop a predictive model for estimating a driver's likelihood of being at fault in a future crash.
• To identify key factors contributing to a driver's responsibility in traffic accidents.

Main Methods:

• Utilized multiple logistic regression techniques on Kentucky licensed driver data.
• Included predictors such as previous crashes (at-fault and not-at-fault), citations, time gaps between crashes, crash type, and demographics.
• Analyzed driver license suspensions and traffic school referrals as contributing factors.

Main Results:

• The crash prediction model achieved up to 74.56% accuracy in classifying at-fault drivers.
• Previous at-fault crash involvements, license suspensions, and traffic school referrals are strong predictors of future at-fault crashes.
• Factors increasing crash likelihood include young/old age, male gender, multiple citation types, and recent crash involvement.

Conclusions:

• The developed model effectively identifies drivers at higher risk of future at-fault crashes.
• Key predictors like prior at-fault incidents and specific demographic/citation patterns inform risk assessment.
• This enables proactive monitoring of driver behavior by licensing agencies for enhanced road safety.