An in-depth study analyzing thousands of drivers provides a new look at what behaviors drivers are involved in before a collision.

The Insurance Institute for Highway Safety says it has been difficult for researchers to conclusively determine what a driver was doing in the moments before a crash. IIHS conducted a long-term naturalistic driving study, which monitors drivers' behaviors over a set period of time, to get a better idea of what factors contribute to accidents.

The study concludes that drivers who manipulate a cell phone, eat or drink, or interact with an in-vehicle system while behind the wheel will increase their risk of a crash. However, researchers also noted that it is important to consider the severity of the crash when determining the effect of distracting behaviors.

Most of the collisions observed in the study were low-risk tire strikes, such as a vehicle hitting a curb. IIHS says 42 percent of the 1,465 collisions observed over the three-year period fell into this category.

Another 41 percent of collisions were classified as minor. Ten percent caused enough damage that the accident was reported to police, and 7 percent fell into the "most severe" category. These crashes included airbag deployment, a significant change of speed at impact, or injuries to the occupants.

"This is a huge trove of valuable data that is adding to our understanding of driver distraction," says David Kidd, a senior research scientist at IIHS and the lead author of the study. "However, in all the hours of driving that were monitored, there were no fatal crashes, and most of the crashes that did occur didn't involve serious injuries. That makes it hard to draw conclusions about the crashes we're most interested in — the ones that kill and injure people."

Researchers looked at video taken prior to the crashes as well as random six-second segments of "normal" driving to see if drivers were engaged in any secondary behaviors. From this information, researchers could calculate the odds that a secondary behavior would lead to a collision. In particular, the study focused on using a cell phone, eating or drinking, using an in-vehicle system, or adjusting the radio or temperature.

Drivers significantly increased the chance of a crash when interacting with an in-vehicle system or manipulating a cell phone, such as using one to send a text message or checking an app. Eating, drinking, or talking on a cell phone also increased a driver's chance of a collision, but IIHS says the increase was not statistically significant.

However, every secondary behavior carried a significant increase in the risk of a collision when low-risk tire strikes were included. The odds of a collision were five times higher than normal driving when a driver was manipulating a cell phone.

"Our understanding of the role cell phones play in crashes continues to evolve," said Anne McCartt, senior vice president for research at IIHS and a co-author of the study. "Although this study shows that manipulating a cell phone is more risky than some other secondary behaviors, it's important to remember that drivers are distracted in many other ways, and putting down the phone does not mean a driver is paying attention to the road. An approach that addresses all kinds of distraction, instead of focusing specifically on cell phones, will be most successful in improving safety."

More than 3,000 drivers participated in the study, and their behavior was monitored between 2010 and 2013. Video footage recorded in the vehicle allowed researchers to see a driver's behavior in the moments before any crashes.

The study is a more extensive follow-up to the first Strategic Highway Research Program, in which 105 drivers were monitored between 2003 and 2004. Due to the limited sample size and study period, few crashes occurred and researchers had to include "near crashes" such as hard braking in their analysis.

The more recent research, known as SHRP2, was conducted in association with the American Association of State Highway and Transportation Officials, the Federal Highway Administration, and the National Research Council. IIHS notes that the study did not account for driver characteristics or environmental factors, and it plans to include these variables in future research.