Abstract
Introduction: Due to the relative rarity of crashes, researchers use traffic offenses, police records, public complaints, and In-Vehicle Data Recorder (IVDR) data as proxies for assessing crash risk. In this study, a unique IVDR system, called Vision-Based Technology [(VBT), (Mobileye Inc.)] was used to monitor perilous naturalistic driving events, such as insufficient distance from other vehicles and pedestrian or bicycle rider near-misses. The study aimed to test the convergent validity of VBT as an indicator of crash involvement risk. Methods: Data from 61 professional drivers working for a large bus company were analyzed (16 of 77 in the original data cohort were excluded for insufficient VBT data). Data included: recorded VBT data, objective data collected from official records (crash records provided by the bus company, and public complaints of reckless driving), self-report data regarding crash involvement, and police tickets. The correlation between VBT, objective and self-reported data was analyzed. Binary-logistic regression modeling (BLM) was used to calculate the odds ratio (OR) for participants involved in a car crash. Results: Correlations were found between the total VBT risk score and official crash records, public complaints, and self-reports of crash involvement. The BLM correctly classified 90% of those who were involved in a crash (sensitivity) and 60% of those who were “crash-free” (specificity). The VBT total risk score was the only significant contributing factor to crash risk, and for each point of increase, the odds of being involved in a crash increased by a factor of 1.55. Conclusions: It is the first study to provide empirical evidence validating the VBT as an indicator of crash involvement and driver safety among professional bus drivers. Practical Applications: VBT technology can provide researchers and clinicians a better understanding of bus drivers' risky driving behaviors- a valuable contribution to road safety interventions for this target group.
Original language | English |
---|---|
Pages (from-to) | 402-408 |
Number of pages | 7 |
Journal | Journal of Safety Research |
Volume | 82 |
DOIs | |
State | Published - 1 Sep 2022 |
Externally published | Yes |
Keywords
- Binary-logistic regression model
- Crash involvement
- In vehicle data recorder
- Naturalistic driving
- Professional drivers
ASJC Scopus subject areas
- Building and Construction
- Safety, Risk, Reliability and Quality