Sleeping and driving: not a safe dual-task

Motor vehicle crashes are the #1 cause of death in the US at every age from 2–34, and rank third (behind cancer and heart disease) overall in terms of years of life lost (Subramanian 2007). Though termed ‘accidents,’ there are several well established risk factors. Alcohol use is the most prevalent. Nearly 40% of all fatal crashes involve drivers using alcohol (Subramanian 2006). Another important factor is drowsiness. A recent 100-car naturalistic driving study using instrumented vehicles found that 12% of crashes (and 10% of near crashes) were related to drowsiness (Dingus et al, 2006). Drowsy driving crashes tend to occur late at night when the circadian physiologic sleep pressure is at its height (Mitler et al 1988). They also tend to occur on major highways at higher speeds leading to greater morbidity and they disproportionately involve young men (as do motor vehicle crashes in general) (Pack et al 1995, Strohl 2002, Horne and Reyner 1995). Lack of driving experience, greater physiologic sleep pressure, and greater use of drugs and alcohol are potential explanations. The effects of sleepiness and alcohol interact to produce a particularly dangerous driver (Wilkinson and Colquhoun 1968; Roehrs et al 1994). The lack of people’s insight into their own sleepiness, as evidenced by their frequently being unaware they have entered stage 1 or even stage 2 sleep, further contributes to drowsiness related accidents.

The major cause of drowsy driving is sleep loss, particularly common in young adults (Bonnet and Arand 1995, Levine et al 1988). Even modest amounts of sleep loss or sleep fragmentation can lead to impairments in reaction time, cognitive function, and mood (Durmer and Dinges 2005, Stepanski 2002, Bonnet and Arand 2003). Night shift workers (including house staff officers), and particularly rotating shift workers, are also at increased risk for crashes (Akerstedt 1988, Marcus and Loughlin 1996). Medical problems, principally untreated sleep apnea, will also impair awake function and increase the risk of automobile accidents (Findley 1995, Pack et al 2006.

Given the prevalence of sleep loss in our society, and the risk of driving accidents related to sleep loss, investigators have sought to develop alerting systems to help the drowsy driver recognize their imminent danger. The simplest of these, rumble strips on the road, may reduce off-the–road crashes by 30–50% but are of unclear value in other situations (Strohl 2001). Direct neurophysiologic monitoring of drivers is an attractive idea. In pioneering work, Torsvall and Akerstedt (1987) monitored the EEG of train drivers as they performed through the night. They demonstrated increases in spectral EEG (alpha and theta power) as the night progressed, and bursts of alpha-theta activity correlating with performance lapses such as driving against a red light.

Physiologic markers for drowsiness in the research lab, including those obtained during driving simulation, have often been evaluated, but this has rarely been done in real-world settings. In an article in this month’s issue, Papadelis and colleagues carried out a sophisticated experiment using EEG, eye blink, and heart rate recording while sleep-deprived subjects drove an actual car 200 km (Papadelis et al, 2007). The experimental car had double pedals accessible to the second person in the car. It is assumed that this critical safety precaution was critical for Institutional Review Board approval. The physiological data was analyzed over the course of the entire drive and also immediately prior to driver errors (e.g., inadvertent lane changes).

The analysis of longer time-periods, what the authors refer to as macroscopic event analysis, helped confirm much prior research on drowsiness. Besides analyzing relative band ratios, several measures of interchannel relationships were used: coherence, synchrony, and entropy. In addition to the macroscopic analysis, the authors used the microscopic analysis of short time-periods to observe the important finding of increased bursts of alpha prior to driver errors. This increased alpha, often at a slightly lower frequency than the posterior alpha rhythm, has been previously described as a marker for drowsiness or stage 1 sleep (Santamaria and Chiappa 1987). The authors also observed increased blinking and longer duration blinks just prior to the driver errors. This change in eye blinking with eyes open needs to be clearly differentiated from the decreased eye blinking seen during early drowsiness in subjects who already have their eyes closed.

The authors were meticulous in their analyses. The parameters to differentiate epochs prior to near accidents from other epochs needed to be based on computerized recognition and not based on visual observation. While the authors did record heart rate which was not a useful discriminant, they did not record other physiologic parameters that might change during sleep onset (see (Oken et al. 2006) for review).

The study shows that there are physiologic changes immediately prior to drowsy drivers making errors in actual driving situations. Whether these physiologic changes can be converted into some device used by large number of drivers is uncertain. The authors comment on the practical considerations of using EEG recordings to do this. Although dry electrodes might facilitate such EEG recordings, the authors suggest that eye blink monitoring may be easier to implement and ultimately more useable for wide-scale implementation. Whatever device and analysis technique is used, before there is wide-scale implementation there will need to be much greater numbers of subjects tested, both sleep deprived and not sleep deprived. It is important to gain a better understanding of the false positive and negative rates, as well as the positive predictive value in various populations, especially given the great person to person variation in the pattern of EEG changes in subvigil states.

In the end, any countermeasure to drowsy driving can only serve as a warning system with the intent that the driver will pull off the road and seek a remedy to their condition. Responsible driving is, and will continue to be, the key safety element. Poorly motivated drivers may ignore or turn off a device just like an unwelcome alarm clock. With that caveat, the continued development and application of new driver safety technologies, whether they are physiologically-based devices or sophisticated global positioning and lane change warning systems, will hopefully lead to fewer car accidents.