Drowsy and Distracted Drivers
Section III
General Description of the Problem
Driver distraction is typically defined in terms of an object or event drawing one's attention from the driving task. It is this presence of a triggering event that distinguishes a distracted driver from other inattentive drivers. The research literature identifies four ways in which persons can be distracted while driving. They can be distracted visually—for example, when they look away from the roadway to locate a CD or tend to a crying baby. They can also be distracted audibly—for example, by a honking car or by children fighting in the back seat of the car. When drivers manipulate radio controls, reach to open the glove compartment, or dial a cell phone number, they are being physically distracted from the driving task. And finally, when they engage in a conversation, whether with a passenger or with the person on the other end of a cell phone connection, they are in danger of being distracted cognitively. Cell phone use, an activity that has garnered considerable attention from the highway safety community, the media, and state and local lawmakers, has the potential for distracting drivers in all four of these areas. Cell phones also represent just one of many wireless technologies increasingly available to drivers in their vehicles.
Unlike driver distraction, driver drowsiness or fatigue involves no triggering event, but instead is characterized by a progressive withdrawal of attention from the road and traffic demands. Drowsiness is the inevitable result of inadequate sleep. Physical fatigue, on the other hand, can occur in drivers who may be tired from hard work or stress, or who may have been driving for a prolonged period of time. For both drowsy and fatigued drivers, however, the effects are the same: decreased driving performance and an increased risk of crash involvement. Therefore, for the purposes of this report, and generally in traffic safety, the terms "drowsiness" and "fatigue" are used interchangeably to mean tiredness. The ultimate level of drowsiness and fatigue is falling asleep at the wheel, although driving performance and safety may be significantly affected by lesser levels of fatigue or drowsiness.
Both distracted and fatigued driving crashes are thought to be underreported on police crash files, since there may be no evidence of driver distraction or fatigue at the scene of a crash. Moreover, drivers may be reluctant to admit distraction or fatigue if they believe it will increase their likelihood of being charged in a crash. Although most state crash report forms contain a code for identifying drowsy and/or fatigued drivers, only about a third contain codes for identifying drivers who were distracted at the time of their crash.
Human alertness level ranges from high to very low (i.e., unconscious) depending on such factors as amount of prior sleep and time of day (Wylie et al., 1996). A poor night's sleep may reduce a driver's performance in subtle ways that he or she may not understand or perceive. Driver attention to the driving task also varies from high to low. Many distracting actions and thoughts are fleeting and can occur almost continuously during driving. At a highway speed of 65 mph, a vehicle is traveling almost 100 feet per second. A glance at roadside scenery or reaching toward a console cup holder can be the difference between timely perception and response to a crash threat and a serious crash.
In addition, a major effect of sleep deprivation is a reduction in vigilance and attention to visual and other stimuli (Dinges et al., 1998; Balkin et al., 2000). Severe sleep deprivation is associated with long and frequent lapses of attention, but even mild sleep deprivation results in some loss of vigilance. Much driver inattention has its roots in drowsiness caused by sleep deprivation and natural time-of-day variations in alertness associated with circadian rhythms. This interaction between drowsiness and inattention is shown as an overlap between the two "icebergs" in Exhibit I-1.
In June 2003, NHTSA released an updated edition of the Model Minimum Uniform Crash Criteria Guideline (NHTSA, 2003). The publication recommended the addition of a new data element to state crash report forms to collect information on driver distraction at the time of the crash. Recommended codes included not distracted, electronic communication devices (cell phone, pager, etc.), other electronic devices (navigation device, palm pilot, etc.), other inside the vehicle, other outside the vehicle, and unknown. The addition of this data element was deemed important for documenting emerging highway safety issues. However, there is still a need for increased training of law enforcement in identifying distraction and drowsiness as contributing factors to crashes.
In the absence of definitive crash data, there is ample evidence of the prevalence of driver distraction and fatigue and their importance for driving safety from survey data as well as from controlled research studies. According to the National Sleep Foundation's annual Sleep in America survey, 37 percent of drivers fell sleep or nodded off while driving in the past year (NSF, 2005). And according to a recent NHTSA survey, 14 percent of drivers involved in a crash in the past five years attribute their crash to being distracted and 3 percent attribute it to drowsy driving (Royal, 2003).
Research results also confirm the increased risks associated with distracted or drowsy driving. A frequently cited study published in the New England Journal of Medicine concluded that the risk of a crash is over four times greater when a driver uses a cell phone (Redelmeier and Tibshirani, 1997). And in a study published by the AAA Foundation for Traffic Safety, more than half of the drivers involved in sleep-related crashes had slept less than 6 hours the night before their crash, compared to less than 10 percent of drivers in a control sample of crashes (Stutts et al., 1999).
Specific Attributes of the Problem
Overall Magnitude and Scope
The CDS, a part of NHTSA's National Accident Sampling System,1 collects detailed data on an annual probability sample of approximately 5,000 police-reported traffic crashes involving at least one passenger vehicle that has been towed from the crash scene. Trained professional crash investigation teams collect information from the scene of the crash, from an examination of the crash-involved vehicles, from interviews with the crash victims and other witnesses, and from available medical records. Even so, the CDS is far from an ideal source of information on precrash driver factors, since its focus is on vehicle crashworthiness, and crashes are typically investigated well after their occurrence.
Beginning in 1995, a variable describing the attention status of the driver—Driver's Distraction/Inattention to Driving—was added to the CDS data collection protocol. An analysis of 2000–2003 CDS data, weighted to reflect all passenger car crashes in the United States, reveals that 6.6 percent of drivers were distracted at the time of their crash, 2.2 percent were sleepy or asleep, and an additional 5.8 percent "looked but didn't see" (See Exhibit III-1. Supporting tables for this and other figures in this section based on the CDS data are contained in Appendix 1). These three categories together total 14.6 percent of crash-involved drivers. This number does not take into account the fact that for nearly half (46.4 percent) the cases, the driver's attention status at the time of the crash was coded as unknown. Thus, the CDS data almost certainly underestimate the true magnitude of the problem.
EXHIBIT III-1
Distribution of Driver Attention Status Based on Weighted 2000–2003 CDS Data
The above numbers are based on all
crash-involved drivers. The
percentage of crashes involving an
inattentive driver is still higher, since in multi-vehicle crashes it is
frequently the case that only the at-fault driver is distracted or fatigued.
According to the same 2000–2003 CDS data,
- 11.6 percent of crashes involve one or more distracted drivers (the same for both single-vehicle and multi-vehicle crashes),
- 3.9 percent involve one or more drivers who were sleepy or had fallen asleep (9.1 percent of single-vehicle crashes and 1.3 percent of multi-vehicle crashes), and
- 10.2 percent involve one or more drivers who "look but don't see" (0.7 percent of single-vehicle and 14.8 percent of multi-vehicle crashes).
Overall, the percentage of crashes with one or more drivers who were identified as inattentive (i.e., either distracted or fatigued or "looking but not seeing") was 25.5 percent. Again, the actual number is likely greater, since information on driver attention status was unknown or missing for many of the crash-involved vehicles.
The CDS data also provide information on the specific sources of driver distraction. Exhibit III-2 shows the sources of distraction for those 6.6 percent of drivers identified as distracted at the time of their crash. The most frequently cited distraction was an object, person or event outside the vehicle. Examples here include other cars and drivers on the roadway, pedestrians, work zones, accident scenes, and general "rubbernecking" (i.e., looking at scenery or landmarks). "Other occupant in vehicle" was cited nearly as often, with frequent reference to infants and young children. Further down the list of distractions were objects brought into the vehicle (which might include portable electronic devices, but also purses and packages), moving objects in the vehicle (e.g., packages or items that fall from the seat, pets, and flying insects), and cell phones. Adjusting the radio or other audio and eating/drinking are each cited in less than 3 percent of the cases, and adjusting vehicle or climate controls and smoking are each in about 1 percent of cases. In the remaining 34.8 percent of cases, the specific source of distraction was either unidentified or simply coded as "other."
EXHIBIT III-2
Specific Sources by Percentage of Driver Distraction Identified in the Weighted 2000–2003 CDS Data
These numbers differ slightly from those contained in an earlier report that examined 1995–1999 CDS data. In that report, outside object/person/event was still the number one identified distraction category at 29 percent, but other occupants in vehicle and adjusting radio/cassette/CD were tied for the numbers two and three positions at 11 percent each. Cell phones appeared further down the list at 1.5 percent (Stutts et al., 2001).
Here, an additional cautionary note is in order regarding the difficulties in collecting reliable data on specific sources of driver distraction. In particular, in the absence of any direct evidence at the scene of a crash, drivers may be much more likely to admit some distractions (e.g., being distracted by a young child or by a passing vehicle) than others (e.g., talking on a cell phone, reading a newspaper) to an investigating officer.
Driver Age and Injury Severity
While younger drivers under the age of 20 are especially likely to be distracted at the time of their crash, all age groups are affected (Exhibit III-3) (Appendix 1 shows more detailed tables on the CDS data). Drivers in the 20–29 age group have the highest percentage of "sleepy/asleep" crashes, while the oldest age groups (60–69 and 70+) are overrepresented in "looked but didn't see" crashes. Clearly, no age group is immune to the problem of inattentive driving.
Exhibit III-3
Distribution of Driver Attention Status within Categories of Driver Age, Based on Weighted 2000–2003 CDS Data
Information with respect to injury status of the crash-involved drivers is shown in Exhibit III-4. Compared to attentive drivers, distracted drivers are 50 percent more likely to be seriously injured or killed in their crash, while drivers who have fallen asleep are 2.3 times more likely to be seriously injured or killed. NHTSA has conservatively estimated that drowsy driving is responsible for 1,500 deaths per year (Knipling and Wang, 1994, 1995). (The higher percentage of fatal injuries for drivers with unknown attention status reflects the difficulty of determining attention status for drivers killed in crashes.)
Exhibit III-4
Percentage of Crashes Involving Serious or Fatal Injury to the Driver, Based on Weighted 2000–2003 CDS Data
Crash Characteristics
Exhibit III-5 provides information on how the crashes of inattentive drivers differ from those of attentive drivers (with supporting tables again contained in Appendix 1). Distracted drivers are somewhat more likely than attentive drivers to be involved in non-collision (i.e., single-vehicle) and rear-end crashes. These two manners of collision together account for nearly 70 percent of distracted and attentive drivers' crashes, with most of the remainder being angle collisions. For crashes where the driver "looked but didn't see," the manner of collision was overwhelmingly an angle collision, reflecting the fact that these crashes primarily occur at roadway or driveway intersections. In contrast, 78 percent of sleepy or asleep drivers are involved in non-collision crashes, with most of the remainder (15 percent) rear-end crashes.
Exhibit III-5
Manner of Collision by Driver Attention Status, Based on Weighted 2000–2003 CDS Data
Exhibit III-6
Crash Type Category by Driver Attention Status, Based on Weighted 2000–2003 CDS Data
This information is confirmed in Exhibit III-6, which shows that distracted driver crashes are primarily categorized as single-vehicle or "STSD" (same travelway, same direction) crashes, looked but didn't see crashes are primarily categorized as turning crashes, and sleepy/asleep crashes are primarily categorized as single-vehicle events.
Information on the time of day during which attentive and inattentive drivers are involved in crashes is presented in Exhibit III-7. The overinvolvement of sleepy/asleep drivers in nighttime crashes is especially notable: over half (52 percent) of all drowsy driving crashes occur at nighttime, between the hours of 10 p.m. and 6 a.m., with nearly 40 percent occurring between 2 a.m. and 6 a.m. Compared to attentive drivers, the crashes of distracted drivers are also somewhat more likely to occur in the evenings and at nighttime. In contrast, crashes where the driver "looked but didn't see" are more likely to occur during the morning hours, a finding that likely reflects the greater proportion of older drivers in these types of crashes.
Interestingly, these results for drowsy driving crashes by time of day contrast with what survey data reveal about the problem. In the NHTSA/Gallup survey referenced earlier, almost three-fourths of the reported instances of nodding off while driving occurred between 6 a.m. and midnight (Royal, 2003). Thus, while crash report data indicate that drowsy driving is primarily a nighttime problem, survey data suggest that it is also a daytime problem. The discrepancy is likely tied to daytime-nighttime differences in exposure, the difficulty that law enforcement officers have in identifying sleepiness as a factor in crashes, and a reliance on a restricted set of "indicators," such as a single-vehicle, running off the roadway, at nighttime, and not involving alcohol. The discrepancy also suggests a significant underreporting of drowsy driving crashes in police and most other crash investigation data, including the CDS.
Exhibit III-7
Time of Day of Crash by Driver Attention Status, Based on Weighted 2000–2003 CDS Data
Roadway Characteristics
Information on how the attention status of crash-involved drivers varies for different roadway characteristics is summarized in Exhibit III-8. Results are presented separately for single- and multi-vehicle crashes.
Compared to attentive drivers, the crashes of distracted drivers are somewhat less likely to occur on higher-speed roadways, on multi-lane (three or more lane) roadways, at a curve in the road, and at a roadway intersection. This is especially true with respect to single-vehicle distracted driver crashes. "Looked but didn't see" crashes, not surprisingly, are much more likely than attentive driver crashes to occur at intersections and are less likely to occur on higher-speed and multi-lane roadways and at a curve in the road.
Even though drowsy drivers are overrepresented in crashes on high-speed roadways (55 percent of their total), they are underrepresented in crashes occurring on multi-lane roadways, especially with regard to single-vehicle crashes. This supports the earlier description of the "typical" drowsy driving crash as involving a single vehicle departing a high-speed, two-lane roadway.
Exhibit III-8
Driver Attention Status by Roadway Characteristics, Based on Weighted 2000–2003 CDS Data
Commercial Vehicle Crashes
In addition to these results from the CDS that are based on passenger vehicles, drowsy driving has also been identified as a problem for commercial vehicle operators, especially long-haul truck drivers. This is primarily due to the more frequent nighttime driving, extended driving times, and irregular sleep schedules that characterize long-haul trucking operations. An estimated 1 percent of all large-truck crashes, 3–6 percent of fatal heavy-truck crashes, and 15–33 percent of fatal-to-the-truck-occupant-only crashes have been attributed to driver fatigue as a primary factor (Knipling and Shelton, 1999). These percentages are based on crash investigations and thus are probably conservative because they do not capture the subtle negative effects that everyday fatigue has on driver performance and crash risk.
The Federal Motor Carrier Safety Administration (FMCSA) and NHTSA have performed a Large Truck Crash Causation Study (LTCCS) (Craft and Blower, 2004) to identify critical factors contributing to crashes involving large trucks. Of 287 two-vehicle crashes involving a large truck and another vehicle (typically a light vehicle) in a preliminary and unweighted LTCCS dataset, 87 had a "critical reason" assigned to the truck, and 200 had a critical reason assigned to the other vehicle. Of the 87 cases where the critical reason was assigned to the truck, 3 percent involved truck driver "non-performance" (a category that includes drowsiness, fatigue, and illness), and 46 percent involved truck driver recognition errors, including driver inattention, distraction, and poor surveillance. Of the 200 cases with the critical reason assigned to the other vehicle, 11 percent involved driver non-performance and 34 percent involved driver recognition errors as critical reasons. In addition to these critical reason designations, there were many other cases where fatigue and/or non-recognition driver errors (poor surveillance, internal distraction, external distraction, or other inattention) were cited. Although driver fatigue is often associated with drivers of large trucks, in the LTCCS fatigue was actually coded more often to the other vehicle driver (both as the critical reason and as a related factor) in crashes involving both trucks and other vehicles. A final report on the LTCCS should be available later in 2005.
State Data
States vary in the extent to which they collect data on the attention status of drivers involved in crashes. Whereas all but six states responding to a recent survey indicated that their crash report form includes a category for identifying sleep- or fatigue-related crashes, not all forms include places for identifying both fall asleep and other fatigue-related crashes. In some states only one of the categories is identified, and in others they are combined (NSF, 1998). Only 17 states collect information on the role of distraction in traffic crashes, and many of these identify only a few major sources of distraction, such as cell phone use (Sundeen, 2003).
While recent interest in cell phones and other technologies has spurred a number of states to modify their crash report forms to include more information on driver distractions, and in particular cell phone use, the reliability of the resulting data has not been demonstrated. In its recent update on state legislative activities related to cell phone use, the National Conference of State Legislatures summarized published data from seven states (California, Florida, Michigan, Minnesota, Oklahoma, Pennsylvania, and Tennessee) regarding crashes attributed to driver inattention and driver cell phone use. The reported percentage of crashes involving inattention ranged from a low of 0.6 percent to a high of 29.9 percent (Sundeen, 2003), another clear indication of the difficulties in collecting reliable data on driver attention status at the time of a crash.
Neither the National Automotive Sampling System (NASS) General Estimates System (GES), based on a nationally representative probability sample of all police-reported crashes, nor FARS, based on a census of all reported fatal crashes, typically reports state-level data on the prevalence of crashes due to driver inattention or fatigue.
Given the known limitations of routinely reported police crash data, and in particular the underreporting of distracted and drowsy driving as contributing factors to crashes, states are encouraged to undertake special data collection activities to better estimate the magnitude of the problem and to identify the most relevant target populations, target locations, and countermeasures for addressing these problems.
A good example of this type of effort is a pilot study of distracted and drowsy driving carried out in Virginia (Glaze and Ellis, 2003). The study was a collaborative effort that involved completion of a special survey form for crashes involving one or more inattentive drivers. Data were collected over a 6-month period in 2002 by troopers and police officers in a sampling of Virginia counties and cities. Results showed that 17 percent of the identified cases involved drowsiness or fatigue; 13 percent involved a driver being distracted by something outside the vehicle; 10 percent involved looking at scenery or landmarks ("rubbernecking"); and 9 percent involved other passengers or children in the vehicle. These results are quite similar to those reported earlier for the 2000–2003 CDS data, except for a higher level of drowsy driving incidents. This difference, however, may be at least partially explained by a higher percentage of single-vehicle crashes (half of the total reported) in the Virginia data.
Improving Data
Prospective data collection activities such as that undertaken in Virginia (described above) can not only yield useful information and serve as a basis for programmatic activities, but also contribute to increased awareness of the problem of distracted and drowsy driving by law enforcement officials and improved reporting by officers responsible for investigating crashes. Other techniques may also be needed to improve available data for addressing the problem of inattentive driving.
As noted earlier, the 2003 revision of the Model Minimum Uniform Crash Criteria (MMUCC) recommends the addition of a new data element on state crash report forms to collect information on driver distraction at the time of a crash (in addition to the data element for driver physical condition, which includes codes for fatigue and loss of unconsciousness or fell asleep). Although many states have added this data element to their crash report forms, there is as yet no documented evidence that such information can be reliably collected and reported by officers who investigate crashes. And indeed, the high level of "missing" and "unknown" data for the driver attention status variable in the CDS data suggests that reliable data collection may be a problem. One or more special studies examining the reliability of reported data in states that have adopted the MMUCC, and perhaps more importantly, approaches for improving data quality, may be needed.
Other approaches to improving the quality of available data on the role of driver inattention in traffic crashes may also prove useful. For example, the Utah Department of Transportation follows a four-step approach for identifying and treating high-crash locations. The approach involves (1) querying the Central Accident Records System, a computerized database of all reportable crashes in the state; (2) using geographic information systems (GIS) to spatially map different types of crashes; (3) visually inspecting crash locations via an Internet-based photo logging system; and (4) conducting onsite reviews to further pinpoint potential safety projects. Following this approach, the department estimates that the percentage of fatal crashes in the state due to drowsy driving alone exceeds 11 percent (see Appendix 2).
In the end, it must be recognized that available data on distracted and drowsy driving will likely never be as accurate or complete as the data on other important aspects of driver behavior. Unlike the use of seat belts, driver attention status cannot be so easily categorized as "yes" or "no," and it certainly cannot be measured and quantified, as with the case of blood alcohol level. Underreporting of the role of driver inattention in crashes will likely remain a problem. However, sufficient evidence exists, from both crash data and other sources, to warrant increased attention to the problem.
1 Although NHTSA's Fatality Analysis Reporting System, or FARS, data also record information on driver-related factors in fatal crashes, driver inattention is believed to be seriously underreported because it is not contained in most states' crash report forms. In 2002, 2.9 percent of drivers involved in fatal crashes were identified as asleep or fatigued, and 6.5 percent were identified as inattentive (NHTSA, 2004).