Investigation of contributing factors to extremely severe traffic crashes using survival theory

ABSTRACT

This study aimed to investigate the contributing factors to serious casualty crashes in China. Crashes with deaths greater than 10 people are defined as serious casualty crashes in China. The serious casualty crash data were collected from 2009 to 2014. The random forest analysis was first conducted to select the candidate variables that affect the risks of serious casualty crashes. The Bayesian random parameters accelerated failure time (AFT) model was then developed to link the probability of the serious casualty crash with road geometric conditions, pavement conditions, environmental characteristics, collision characteristics, vehicle conditions, and driver characteristics. The AFT model estimation results indicate that overload driving, country road, northwest china region, turnover crash, private car, snowy or icy road surface and sight distance conditions have significant fixed effects on the likelihood of serious casualty crashes. In addition to these fixed-parameter variables, freeway, clear weather conditions, coach drivers, and upgrade horizontal curve affect the likelihood of serious casualty crashes with varying magnitude across observations. One of the important findings is that the serious casualty crash likelihood does not always decrease with an increase in the driving experience (number of years driven). Before the inflection point of 7 years, the serious casualty crash likelihood increases as the driving experience grows. The results of this study can help to develop effective countermeasures and policy initiatives for the prevention of serious casualty crashes.

Highlights

• This study aimed to identify contributing factors to serious casualty crashes in China.

• Accelerated failure time (AFT) models were developed based on variables selected by random forest.

• The random-parameter AFT model provides the best fitness to the duration data.

• The serious casualty crash probability was linked with geometric, pavement, environmental, collision, vehicle, and driver characteristics.

• The serious casualty crash risks do not always decrease as driving experience grows.

KEYWORDS:

• Serious casualty crashes
• Bayesian survival analysis
• traffic safety
• accelerated failure time model
• random-parameter regression
• duration until-crash occurrence

Acknowledgements

The authors would like to thank the editor and the reviewers for their constructive comments and valuable suggestions to improve the quality of the article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was supported by the National Natural Science Foundation of China [grant number 51508093], [grant number 5151101143].