Drivers not paying attention — such as answering a phone call, a text message, or being distracted by a passenger — for any length of time are 29 times more likely to be involved in a collision or near-collision in a highway work zone, according to new research by the University of Missouri (MU).

The results from the study could provide recommendations on “behavioral countermeasures” to state transportation agencies and the Federal Highway Administration, which are implementing countermeasures to help decrease injuries and fatalities in highway work zones. They could also be used when developing new technology, such as driverless vehicles.

“Prior to our study, researchers analyzed data on work zone safety by looking at one checkbox among 70-80 different fields on a police officer’s crash report to see if the crash occurred inside a work zone,” said Praveen Edara, Professor of Civil and Environmental Engineering at the MU College of Engineering. “Unfortunately, crash reports do not include detailed information about driver behavior prior to a crash. What’s unique about our research project is that we used naturalistic driving study data that provides information about how driver, vehicle, roadway and environmental factors contribute to a crash. In other words, we reconstructed a driver’s actions and the surrounding environment prior to the crash from a firsthand account.”

The study used data from the Transportation Research Board’s second Strategic Highway Research Program’s Naturalistic Driving Study. During 2006 – 2015, researchers collected data from more than 3,000 drivers traveling more than 50 million miles. With this information, researchers were able to see a detailed firsthand account of a driver’s interaction with the vehicle, roadway and surrounding environment.

The study, Risk Factors in Work Zone Safety Events: A Naturalistic Driving Study Analysis, is published in the National Academies of Sciences, Engineering and Medicine’s Transportation Research Record: Journal of the Transportation Research Board.

Identification of crash risk factors and enhancing safety at work zones is a major priority for transportation agencies. There is a critical need for collecting comprehensive data related to work zone safety. The naturalistic driving study (NDS) data offers a rare opportunity for a first-hand view of crashes and near-crashes (CNC) that occur in and around work zones. NDS includes information related to driver behavior and various non-driving related tasks performed while driving. Thus, the impact of driver behavior on crash risk along with infrastructure and traffic variables can be assessed. This study: (1) investigated risk factors associated with safety critical events occurring in a work zone; (2) developed a binary logistic regression model to estimate crash risk in work zones; and (3) quantified risk for different factors using matched case-control design and odds ratios (OR). The predictive ability of the model was evaluated by developing receiver operating characteristic curves for training and validation datasets. The results indicate that performing a non-driving related secondary task for more than 6 seconds increases the CNC risk by 5.46 times. Driver inattention was found to be the most critical behavioral factor contributing to CNC risk with an odds ratio of 29.06. In addition, traffic conditions corresponding to Level of Service (LOS) D exhibited the highest level of CNC risk in work zones. This study represents one of the first efforts to closely examine work zone events in the Transportation Research Board’s second Strategic Highway Research Program (SHRP 2) NDS data to better understand factors contributing to increased crash risk in work zones.

Work zone safety continues to be a high priority area for state transportation agencies. Federal Highway Administration (FHWA) reports that a crash occurs in a work zone every 5.4 minutes (per 2015 data) (1). In 2015, 96,626 crashes occurred in work zones across the US resulting in 25,485 injuries and 642 fatalities. Work zone crashes are not only a problem for the traveling public, they are a serious concern for highway workers who are injured or killed by errant vehicles. According to data from the U.S. Department of Transportation, 20,000 people experience work-related injuries in road construction zones each year. Of these, approximately 12% can be traced back to vehicle crashes or other transportation-related incidents. A better understanding of the causal factors related to work zone crashes is essential to designing effective countermeasures.

Traditional safety research relied on using statistical modeling approaches to capture the effect of various road, traffic, and environmental characteristics. Poisson and Negative Binomial models are well accepted approaches to model count data such as crashes (25). However, the effect of individual driver behavior is typically not included in such approaches due to the difficulty in measurement. The Transportation Research Board’s second Strategic Highway Research Program (SHRP 2) conducted a large naturalistic driving study (NDS) to investigate the role of driver and other factors in crash and near-crash (CNC) events (6). Over 3,000 drivers participated in the comprehensive driving experiment from six sites in Florida, Indiana, New York, North Carolina, Pennsylvania, and Washington. Nearly 50 million vehicle miles of data were recorded from trips made by these drivers. These unprecedented data allow for the investigation of the role of driver behavior in traffic safety, including under work zone conditions. The interaction of the driver with the vehicle, roadway, and the environment is captured in detail. Such highly detailed data enable a more accurate determination of the causes of CNC than the typical post-crash investigation using law enforcement data. The NDS data addresses a need that is not met by traditional data sources used in work zone safety research.

This study was conducted to better understand the contributing factors of safety critical events in work zones using NDS data. This understanding will enable us to design optimal countermeasures and also proactive warning systems to alert drivers of impending hazardous conditions near work zones. The study has three key objectives: (1) identify factors associated with risk of an individual driver being involved in a safety critical event in a work zone; (2) develop a logistic regression model to predict this risk; and (3) quantify risk for different factors using matched case-control design and odds ratio (OR).