Abnormal sleep duration has recently been recognized as an independent risk factor for heart disease (Ayas et al. 2003) and all-cause mortality (Kripke et al. 2002; Patel et al. 2004), but the causal pathways are uncertain. Furthermore, excess mortality does not seem to be due to a single disease pathway (Patel et al. 2004). A variety of evidence led us to hypothesize that some of the observed excess mortality might be linked to motor vehicle crashes. In a random sample of the New Zealand middle-aged population, sleep and sleepiness measures predict self-reported motor vehicle crash involvement as strongly as traditional risk factors such as age, gender, and time on the road (P. H. Gander, N. S. Marshall, R. B. Harris and P. Reid, unpublished observations). A case–control study of injury crashes in the Auckland Region of New Zealand has estimated that crash risk could be reduced by 19% if people avoided driving when sleepy, or after <5 h of sleep, or between 2 and 5 am (Connor et al. 2002).
As part of a fatigue management programme at BP, all 960 Australasian light vehicle drivers attending fatigue management training sessions completed a questionnaire on their sleep habits, sleepiness, driving behaviour, and 3-year crash involvement while driving (Maycock 1996). The questionnaire is available from the corresponding author. Data were double entered and visually checked for inconsistencies and outliers. Analyses were undertaken in SAS (v.8; SAS Institute, Cary, NC, USA). Descending logistic regression modelling was used to determine independent risk factors for crash involvement.
The respondents were mostly male (69%) and their average age was 38.1 yrs (SD = 9.8 years). As a group, they were sleepier (24% scored >10 on the Epworth Sleepiness Scale) than a random sample of New Zealand adults aged 30–60 years, of whom 15% scored >10 (Harris, 2003). Among the BP employees, 57% reported feeling close to falling asleep whilst driving in the past year, and 16% reported being the driver in at least one crash in the past 3 years. Drivers felt that fatigue (either their own or that of someone else or both) played a role in 23–39% of these crashes.
Self-reported usual sleep duration on a night off averaged 7.8 h (SD = 1.8 h). Reported sleep durations fell into the following approximate quartiles: <7.0 h sleep per night (28%); 7.0–7.99 h (27%); 8.0–8.49 h (30%); and ≥ 8.5 h (14%). People who reported sleeping <7.0 h (21.3% crash involvement), or ≥ 8.0 h (19.1%), were more likely to report crashes than those who reported sleeping 7.0–7.99 h (13.0%) (Fig. 1).
A total of 766 respondents (135 who had crashed) answered all the necessary questions to be included in the logistic regression modelling (have crashed/have not crashed). Table 1 shows the independent variables included in the model.
|Variable||Type||Adjusted OR||95% CI||P|
|Gender||Dichotomous: females versus males||1.03||0.64–1.63||0.917|
|Age||Quartiles: 18–30 years, 31–36 years, 37–44 years, ≥ 45 years||0.77 (1.33)||0.65–0.92||0.004|
|Annual distance driven||Quartiles: 1–9000 km, 9001–19 000 km, 19 001–30 600 km, ≥ 36 001 km||1.01||0.84–1.22||0.921|
|Abnormal sleep duration||Dichotomous: outside 7–7.99 h versus 7–7.99 h||1.62||1.03–2.55||0.036|
|Epworth question 8||Dichotomous: any chance versus no chance of dozing in a car while stopped in traffic||1.76||1.17–2.67||0.007|
Overall model fit was significantly different from a zero slope (Wald chi-square 20.4, P = 0.001). There were no significant issues with co-linearity in this model with all correlations below r = 0.4. No interactions between variables were significant.
Among BP light vehicle drivers in Australasia, abnormal sleep duration and abnormal sleepiness in a motor vehicle are at least as important independent predictors of self-reported crash involvement as the traditionally recognized risk factors of age, gender and driving exposure.
Excessive daytime sleepiness, as measured by the Epworth Sleepiness Scale, was more prevalent in this employed population than in the general population, possibly due to the fact that the questionnaire was answered immediately after fatigue management training. The proportion of crashes in which fatigue was considered to have played a role was comparable or greater than has been reported for other groups (Garbarino et al. 2001; Maycock 1996; Mondini et al. 2000).
This study does not demonstrate a relationship with mortality. However, the link between abnormal sleep duration and crash risk link may offer another pathway in the excess mortality observed with abnormal sleep duration (Kripke et al. 2002; Patel et al. 2004). Both Kripke et al. (6.5–7.4 h) and Patel et al. (7 h, tick box) found that the lowest mortality rates were independently associated with the group reporting sleep below the standard recommended roughly 8-h duration from recent performance-based literature (Belenky et al. 2003; Van Dongen et al. 2002).
All studies purporting to show relationships between abnormal sleep and morbidity or mortality, that are based on self-report, have the problem that subjective reports of sleep duration do not reliably reflect objectively measured sleep duration, at least on an episode-by-episode basis. The current research also has the limitation that the data are all retrospective, and sleep may have changed across the 3-year period for crash reporting. Nevertheless, the notion that some mid-range ‘optimal dose’ of sleep is somehow protective against adverse outcomes, including motor vehicle crashes, raises interesting questions. While the link between chronic sleep restriction and adverse outcomes is well established (Belenky et al. 2003; Spiegel et al. 2002; Van Dongen et al. 2002), it is not clear why sleeping longer than population norms should be deleterious. One possibility is that excessive sleep duration is associated with a range of other co-morbidities that in turn mediate the adverse outcomes. How that hypothesis might relate to elevated crash risk is unclear.