A penalized-likelihood approach to characterizing bridge-related crashes in New Jersey

Abstract

Objective

A roadway departure crash is one in which a vehicle crosses an edge line, a centerline, or otherwise leaves the traveled way. These crashes that involve run-off-road and cross-median/centerline head-on collisions tend to be more severe than other crash types. According to the NHTSA Fatality Analysis Reporting System database, a total of 7,833 people perished in crashes involving fixed roadside objects in 2017, accounting for 21 percent of the total number of fatalities in the United States. Several previous studies have reported that rural bridge-related crashes result in more fatalities due to their being mostly the fixed-object crash type. As such, further in-depth investigation of this type of crash is necessary. Due to the lack of a comprehensive database that includes bridge-related crashes and bridge characteristics, identifying the key factors contributing to this type of crash is a challenging task that is addressed in this paper.

Method

Study team gathered and compiled five years (2011–2015) of crash data from the New Jersey crash database and the characteristics of bridges from the Long-Term Bridge Performance portal. A Firth’s penalized-likelihood logistic regression model was developed to examine the impact of explanatory variables on crash severity.

Results

Based on the five years (2011–2015) of crash data, significant factors (i.e., driver age, weather conditions, surface conditions, lighting conditions, speed limit, roadway characteristics, and direction of traffic) were identified that affect the severity of bridge-related crashes in Middlesex County, New Jersey.

Conclusion

This model is an appropriate tool for predicting the impact of all the confounding variables on the probability of bridge-related crashes while also considering the rareness of the event. Based on the obtained odds ratio, the various effects of the identified variables are discussed, and recommendations made regarding countermeasures policymakers can establish to reduce the number of these crashes in New Jersey.

Keywords:

• Bridge-related crash
• fixed-object
• odds ratio
• Firth’s logistic model