Driver sleepiness and individual differences in preferences for countermeasures


Anna Anund, Swedish National Road and Transport Research Institute (VTI) SE-581 95 Linköping, Sweden. Tel.: 0046-13-204327; fax: 0046-13-141436; e-mail:


The aim of the present national questionnaire study was to relate the use of sleepiness countermeasures among drivers to possible explanatory factors such as age, sex, education, professional driving, being a shift worker, having experience of sleepy driving, sleep-related crashes, problems with sleep and sleepiness in general and sleep length during working days. Also the attitude to countermeasures related to information or driver support system was studied. A random sample of 3041 persons was drawn from the national register of vehicle owners. The response rate was 62%. The most common countermeasures were to stop to take a walk (54%), turn on the radio/stereo (52%), open a window (47%), drink coffee (45%) and to ask passengers to engage in conversation (35%). Logistic regression analysis showed that counteracting sleepiness with a nap (a presumably efficient method) was practiced by those with experience of sleep-related crashes or of driving during severe sleepiness, as well as by professional drivers, males and drivers aged 46–64 years. The most endorsed means of information to the driver about sleepiness was in-car monitoring of driving performance providing drivers with information on bad or unsafe driving. This preference was related to experience of sleepy driving, not being a professional driver and male gender. Four clusters of behaviours were identified: alertness-enhancing activity while driving (A), stopping the car (S), taking a nap (N) and ingesting coffee or other sources of caffeine (C) (energy drinks, caffeine tablets). The participants were grouped according to their use of any of the four categories of countermeasures. The most common cluster was those who used activity, as well as stopping and drinking caffeine.


Sleep-related crashes have received increasing attention during the latest decade. The National Transportation and Safety Board (US) has pointed out that sleepiness while driving is one of the most important contributing factors for road crashes (NTSB 1999). Epidemiological studies based on self-reports or in-depth crash investigations show much higher figures compared to crash statistics and suggest that about 10–20% of all crashes might be sleep or fatigue related (Horne and Reyner, 1995a,b; Maycock, 1997; Stutts et al., 1999). It has also been demonstrated in post-crash interviews that night driving, prior sleep <5 h, and sleepiness level before the crash are major predictors of the risk of being involved in a road crash (Connor et al., 2001).

Considering the central role of sleepiness knowledge, the use of countermeasures should be an important issue. With respect to ‘strategic’ countermeasures one should obviously avoid night driving and make sure that sufficient amounts of sleep have been obtained before driving. However, more acute countermeasures, while actually driving, are less obvious and there does not seem to exist in the public domain studies of what countermeasures are used. Anecdotally, however, it appears that acts like opening a window, turning on the radio, or taking a break may be common. However, laboratory studies indicate that the three suggested countermeasures, including doing exercise during a break, do not improve alertness (Horne and Reyner, 1996; Lisper and Eriksson, 1980). Horne and Reyner have also demonstrated that caffeine and taking a nap significantly reduce driving impairments, subjective sleepiness and electroencephalographic (EEG) indications of drowsiness (Horne and Reyner, 1996). The dramatic alerting effects of these behaviours have also been demonstrated repeatedly in laboratory studies with other performance measures (Tietzel and Lack, 2002; Wesensten et al., 2005). A matched case–control study indicated that the crash risk was less for drivers who used highway rest stops, drank coffee within the last 2 h or played a radio while driving (Cummings et al., 2001).

Apart from the interest in understanding what countermeasures are used one also needs to know whether different groups use differentially efficient countermeasures as these may be related to long-term risk of accidents. This type of knowledge might aid in identifying vulnerable groups. Age is probably such a factor because of its close relation to accident risk and risk behaviour, the group in focus being young drivers (Galvan et al., 2007; Horne and Reyner, 1999). Gender is another factor, with greater risk attached to males (Akerstedt and Kecklund, 2001). One might also hypothesize that experience of drowsy driving, sleep-related crashes, shift work, professional driving, might lead to use of countermeasures as they might provide some insight into the problem. Higher education or age might carry similar insights but via more indirect routes. No previous studies on this topic seem available, however.

Apart from driver-initiated countermeasures one might also consider various types of information to the driver about fatigue risks in driving. Information in media could be one such way. Public campaigns along the roads might be another one. Information in connection with the annual vehicle safety inspection might be a third. In addition, there has been considerable development in the area of driver support systems, focused on feedback on hazardous driving in terms of impaired lateral control (Brookhuis and de Waard, 1993; Dinges and Mallis, 1998) or focused on the physiological state of the individual (e.g. sleepiness) (Akerstedt and Folkard, 1997; Horne and Reyner, 1999; Wierwille and Ellsworth, 1994). As driver support systems are associated with sizeable investments by society and/or manufacturers it would be of interest to investigate the attitude to such countermeasures, as well as to study whether background factors are related to such attitudes. To the best of our knowledge these issues have not been studied before.

The purpose of the present study was to describe the use of acute countermeasures against sleepiness at the wheel and relate the use of such countermeasures to possible explanatory factors, such as age, sex, education, professional driving, being a shift worker, having experience of sleepy driving, having experience of sleep-related crashes, problems with sleep and sleepiness in general and sleep length during working days. For those groups also the attitude towards countermeasures related to information or driver support system was studied.


A questionnaire was sent to a random sample of 3041 passenger car owners in Sweden. The sample was drawn from the national register of vehicle owners, and stratified, using equal sample sizes, into four age groups: young drivers (18–25 years), young middle-aged (26–45 year), old middle-aged age (46–64 years) and elderly drivers (65 years or more). The questionnaires were distributed during the winter 2002/2003 and two postal reminders were sent out. The response rate was approximately 62%; highest among the elderly drivers (71%) and lowest among the young drivers (52%). The results have been weighted in order to permit generalization. The weights used were the following: starting with the youngest: (18–25) 62; (26–45) 1366; (46–64) 1269 and (65 or older) 438. Hence, one answer from a driver aged 65 or older represents approximately 438 vehicle owners in that age. There were more men than female respondents. This was also the case for the population of vehicle owners (see Table 1 for details).

Table 1.   Background factors – percent, number (n) or mean ± standard deviation (SD)
  1. Total responses (n) was 1880; n = subjects in defined category. Partial missing values not higher than 1.5%.

Sex – female30562
Shift workers13235
Professional drivers – Yes6115
Average sleep quality (1 = very bad; 5 = very good)3.9 ± 1.0 
Poor sleep quality (=1 + 2)7140
Experience of persistent sleepiness during the past 6 months – Yes17317
Experience of severe sleepiness while driving in the past 6 months – Yes38714
Experience of fatigue-related crash/es during the past 10 years – Yes356
Higher level of education (>12 years)46859
Snoring – Yes13246
Sleep duration working day (average hours)7.0 ± 1.1 
Habitual sleep 6 h or less during working days24448

The construction of the questionnaire was based on three discussions with focus groups: one with young drivers, one with professional drivers and one with middle-aged commuters (Anund et al., 2002). Much of the questionnaire concerns sleepiness and definition of ‘sleepy’ or ‘sleepiness’ varies among individuals and disciplines (Bartley and Chute, 1947; Broadbent, 1979; Brown, 1994, 1997; Grandjean, 1979). When questions in the present study were asked concerning experiences of sleepiness while driving the following definition was given to the subjects: “By ‘sleepy’ or ‘sleepiness’ we refer to situations when you as a driver has to make efforts to stay awake while driving”. The questionnaire consisted of 38 questions divided into the following parts: background, health and sleep, what makes drivers become sleepy, experiences of being sleepy while driving, experience of sleep-related crashes, the awareness of sleepiness signals, actually used and potential countermeasures. Most of the response alternatives on the questions about health and sleep were given on a semantic differential scale from 1 (=very bad) to 5 (=very good) with anchored end points.

In order to capture the driver’s knowledge of efficient and lasting countermeasures, the drivers were asked if they ‘normally do anything to reduce their sleepiness or to be more alert while driving’. They were presented a menu of 22 different items and asked to check those alternatives that correspond to what they would usually do. The items presented to the respondents were chosen based on the discussions within the focus groups and from the results from the test questionnaire. No limit was set on the number of items that could be checked. Questions related to the drivers’ attitude to the means of information was structured in the same way.

In order to determine what background factors that may be related to the use of countermeasures a logistic regression was carried out using ‘efficient’ countermeasures as a criterion and ‘less efficient’ countermeasures as the reference. Two different criteria of ‘efficient’ were used. The first one was ‘stop to take a nap’, while excluding those in the reference category that took coffee/red bull or caffeine pills (since they represented the second ‘efficient’ countermeasure. Thus, the reference category only included those who only used took less efficient countermeasures than caffeine or naps. As a second definition of efficient was used drinking coffee/red bull or taking caffeine pills, again excluding those in the reference group that did stop for a nap.

In order to relate use of efficient countermeasures to age, sex, education, professional driving, being a shift worker, having experience of sleepiness while driving, sleepiness-related crashes, persistent sleepiness, the driver’s education level, if they suffer from snoring or reduced sleep quality or sleepless than 6 h during working days, a logistic regression analysis was used. In a first step, a univariate model was used. Secondly, variables with significant odds ratios were, in a second step, included in a multivariate logistic regression (forward stepwise approach). The significance level was set to 0.05. The same analysis approach was used for questions related to information about awareness of driver sleepiness. The criteria for defining exposure and reference groups are found in Table 1 above.

This study was reviewed and accepted by VTI’s internal ethical board. VTI has an ISO 9001 certificate and has its own quality board and procedures related to quality.


The most common countermeasures were to stop to take a walk, turn on the radio/stereo, open a window, drink coffee and to ask passengers to engage in conversation (see Table 2). In order to study how drivers combined countermeasures: two clusters of behaviours were identified: alertness-enhancing activity while driving (A) and stopping the car (S). Based on the laboratory studies of alertness countermeasures like taking a nap (N) and ingesting coffee or other sources of caffeine (C) (energy drinks, caffeine tablets) were broken out of these categories. The participants were then grouped according to their use of any of the four categories of countermeasures. The results are presented in Fig. 1. It shows that the most common cluster were those who used activity, as well as stopping and drinking caffeine. The second most common cluster was those that used activity or stopping only and the third most common cluster was activity only. The three following clusters all included napping in various combinations. Note that all clusters except the last one always included activity.

Table 2.   Drivers’ countermeasures against sleepiness while driving – percent of drivers who indicated the used countermeasure when sleepy
CountermeasureTotal (%)
Stop and go for a short walk54
Turn on the radio/stereo52
Open the window47
Drink coffee45
Ask the passenger to engage in conversation35
Eat candy32
Stop and exercise outside the vehicle28
Stop and rest for a short time – while seated26
Body movements while driving27
Turn up the radio/stereo26
Drink lemonade26
Eat fruit26
Stop and sleep for a short while – remain seated18
Turn on the fan or the AC16
Use nicotine14
Drive slower13
Drive more actively13
Drink an energy drink, e.g. Red Bull6
Drive faster5
Take caffeine pills1
Increase the heat0
Figure 1.

 Countermeasures grouped according to driver’s use of any of the four categories of countermeasures: nap, caffeine, stopping or activity. Some categories with small numbers of endorsements are left out.

Table 3 shows the results from the univariate logistic analysis. When ‘efficient’ was defined as ‘stop for a nap’, young age was associated with a decreased likelihood of using efficient countermeasures, while belonging to the age group 45–64 years was associated with an increased likelihood. The latter was the case also for male gender, being a professional driver, having had experience of sleepy driving, having had a sleep-related crash, being a snorers and sleeping less than 6 h during working days. Being a professional driver showed the highest odds ratio (3.43). The results were almost the same when ‘efficient’ was defined as drinking coffee/red bull or taking caffeine pills. The difference was that having had a sleep-related crash was not significantly related to taking efficient countermeasures or not, instead drivers with snoring problems, drivers with poor sleep quality and drivers that sleep less than 6 h during working days had an increased probability of taking an efficient countermeasure.

Table 3.   Univariate logistic regression – dependent variable: efficient = stop for a nap (n = 303); or efficient = caffeine intake (n = 824)
Model with univariate predictorsEfficient = stop for a nap Efficient = caffeine intake
Odds ratioCI POdds ratioCI P
  1. Odds ratio = Exp(β); CI, confidence interval for odds ratio and P-value = significant level.

 65 or older1.010.68–1.500.971.581.20–2.07<0.01
Gender – male versus female2.832.04–3.93<–2.46<0.01
Higher education versus lower1.280.98–1.660.070.950.79–1.150.60
Professional drivers versus non-professional3.432.05–5.73<0.012.451.60–3.75<0.01
Experience of sleepy driving versus not2.762.11–3.60<0.011.451.19–1.76<0.01
Experience of sleep-related crashes versus not2.802.01–7.19<–1.860.91
Shift workers versus day workers1.250.87–1.810.230.930.70–1.240.64
Persistent sleepiness versus not0.870.60–1.250.451.050.82–1.350.70
Snoring versus not1.701.16–2.50<0.011.711.29–2.26<0.01
Poor sleep quality versus good1.430.88–2.320.151.461.02–2.090.04
Sleep duration <6 h versus more1.741.30–2.32<–1.430.22

Variables with significant odds ratios were in a second step included in a multivariate logistic regression (forward stepwise approach). The multivariate regression against ‘stop for a nap’ resulted in a model containing experience of sleep-related crashes (OR = 3.90; CI=1.91–8.0; P < 0.01), being a professional driver (OR = 2.98; CI = 1.70–5.23; P < 0.01), experience of driving under conditions of severe sleepiness (OR = 2.76; CI=2.04–3.75; P < 0.01), age (OR 46–64 years = 2.66; CI = 1.48–4.76; P < 0.01) and male gender (OR = 2.56; CI = 1.80–3.64; P < 0.01) as significant predictors.

When ‘efficient’ was defined as caffeine intake, the multivariate logistic regression resulted in the following significant predictors: being a professional driver (OR = 2.45; CI = 1.57 − 3.83; P < 0.01), age (26–45 years OR = 1.50; CI = 1.11 − 2.03; P < 0.01; 46–64 years OR = 2.03; CI = 1.50–2.74; P < 0.01; 65 and older OR = 1.97; CI = 1.44 − 2.70; P < 0.01), male gender (OR = 1.83; CI = 1.46 − 2.29; P < 0.01) and experience of driving under conditions of severe sleepiness (OR = 1.53; CI = 1.23–1.91; P < 0.01).

The drivers were also asked about their view of what would be efficient countermeasures in terms of information. This showed that driver support systems, providing drivers with information on bad or unsafe driving, was seen as efficient by about 32% of the drivers. Warning systems based on the drivers’ physiological state was thought to be efficient by 16%, information by road signs was judged to be efficient by 12%, while that for information through media was 9% and that through the annual vehicle safety inspection was 4%.

For each of five information alternatives, a logistic regression analysis was carried out, similar to the previous analyses. For information through road signs, media or the annual vehicle test there was no major difference between predictors. For driver support systems, providing information on bad or unsafe driving, there were differences for almost all independent variables except for education, being a shift worker and sleeping less than 6 h during working days, see Table 4. Driver support system were seen to be more efficient among drivers aged 24–64, but also by males, drivers with experience of sleepy driving or sleep-related crashes, experience of persistent sleepiness, bad sleep quality and problems with snoring. The highest odds ratio was seen for those with experience of sleep related crashes.

Table 4.   Univariate logistic regressions – dependent variable: positive to drivers’ support system providing information about poor/unsafe driving and information about drivers´ physiological state
  Information about poor/unsafe drivingInformation about drivers’ physiological state
Model with univariate predictorsORCIORCI
  1. Odds ratio (OR) = Exp(β); CI, confidence interval for odds ratio and P-value = significant level.

 65 or older 1.240.93–1.671.230.84–1.80
Gender – male versus female0.780.63–0.971.150.87–1.52
Higher education versus lower1.210.99–1.471.180.92–1.51
Professional drivers versus non-professional0.600.38–0.951.370.77–2.44
Experience of sleepy driving versus not1.311.07–1.601.641.27–2.11
Experience of sleep related crashes versus not1.771.03–3.031.370.70–2.68
Shift workers versus day workers0.960.71–1.290.900.61–1.33
Persistent sleepiness versus not1.301.01–1.681.200.87–1.66
Snoring versus not1.341.01–1.781.641.78–2.29
Poor sleep quality versus good1.501.05–2.141.500.98–2.31
Sleep duration <6 h versus more1.140.91–1.431.110.83–1.48

The multivariate regression (forward stepwise approach) entering variables with significant differences in the univariate analysis as predictors, resulted in a model with the effect of male gender (OR = 0.77; CI = 0.62 − 0.95; P = 0.02), being a professional driver (OR = 0.60; CI = 0.38 − 0.96; P = 0.03) and having experience of driving under severe sleepiness (OR = 1.32; CI = 1.08 − 1.63; P < 0.01) as significant predictors. The univariate analysis against ‘Using information from a warning system based on the drivers’ physiological state’ showed significant effects of age 26–45 years, experience of driving under severe sleepiness and being a snorer (see Table 4). The multivariate regression (forward stepwise approach), using the significant predictors from the univariate analysis, resulted in a model with experience of driving in severe sleepiness (OR = 1.60; CI = 1.24–2.06; P < 0.01) and snoring (OR = 1.56; CI = 1.12 − 2.19; P < 0.01).


The most common self-administered countermeasures involved stopping for a short walk, turning on the radio/music player, opening a window and drinking coffee. Thus, one of the two most efficient countermeasures, caffeine intake, was the fourth most common countermeasure, whereas the second most efficient countermeasure, stopping for sleep, was only practiced by 18%. This is a relatively discouraging result but reveals a potential safety benefits within reach by information and education efforts. However, the amount of research into efficient countermeasures is limited and has mostly been carried out in the laboratory (Horne and Reyner, 1996, 1999). Thus, there is a possibility that the more stimulating driving context of the real road might result in other countermeasures also becoming efficient. The matched case–control study showed for instance crash reduction among those using highway rest stops, drinking coffee or playing radio while driving (Cummings et al., 2001).

The pattern of countermeasure use seems rather logical. Most clusters included some sort of activity presumed to be alertness promoting. Very likely, the reason is that activity possible to carry out while still driving is less intrusive than stopping or taking a nap, and will therefore be easier to apply. One would then expect a combination of activity as a first line of defence, followed by stopping, often combined with caffeine intake. Napping would be a less prevalent means, possibly related to exposure to sleepiness as discussed below.

Interestingly, the use of efficient countermeasures clearly differed between groups. Among the most obvious result was the importance of having had a sleep-related crash or experience of sleepiness while driving, for identifying a nap as an efficient countermeasure. This probably explains the relatively low prevalence (18%) observed above – as due to a lack of experience of dangerous sleepiness. Note that being a professional driver or being older (up to the retirement age criterion) add explanatory power. Both seem to represent other aspects of experience that would lead one to use napping. This may be related to increasing caution with increasing knowledge about the dangers of driving. The increase in preference for efficient countermeasures with increasing age is interesting from the point of view of the lower risk of sleepiness-related accidents in higher age groups (Akerstedt and Kecklund, 2001; Pack et al., 1995).

Also, being a male was associated with a preference for a nap. The reason for this is not immediately obvious. However, the results from the pre-discussion within focus groups (Anund et al., 2002) suggested that females can be afraid of stopping for reasons of personal safety. Interestingly, having disturbed sleep, being a snorer or being a shift worker were not related to identifying a nap as important, despite the fact that such individuals should have sample experience of sleepiness (Philip and Akerstedt, 2006).

When caffeine intake was used as the dependent variable, being a male, being a professional driver, being aged over 25 years and having experience of driving under severe sleepiness were again significant predictors. Having had a sleepiness-related crash did not enter into the regression. The reason for the latter is unclear, but clearly caffeine intake is more common as a countermeasure than taking a nap and caffeine intake is also very widespread and a relatively easily administered countermeasure. Thus, caffeine intake countermeasures may demand less experience of critical situations, e.g. sleep-related crashes to be applied as a countermeasure.

With regard to information to drivers about sleepiness, the most accepted one clearly was feedback of driving behaviour. This is similar to what has been found in a field study on this topic (Dinges et al., 2005). More passive information (signs, other information) was less credible as countermeasures. Clearly, again, experience of sleepy driving was an important predictor. However, professional drivers and women were not positive. The former may believe that their professional skill and experience does not need external information on driving performance, but the reason for the negative attitude of women is more difficult to understand. One might perhaps suspect a lower interest in technical gadgets among women compared to men, and therefore trust in such devices might be lower in women. This remains to be demonstrated, however.

Another feedback to the driver could be information about the drivers’ physiological state (alertness level). For this system, there was less difference between driver groups. Thus, a positive view was seen among those with experience of sleepy driving or drivers among those with problems with snoring, but also for drivers aged 26–45 years. The reason for the difference in preference between systems based on impaired driving behaviour and drivers’ physiological state is not clear.

The practical implications of the present results are several. Thus, it seems that exposing learner drivers to sleepy driving might be a way of raising awareness. Safe rest areas might increase the willingness of women to stop for a break. Information addressed to learning drivers about effective/not effective countermeasures and how to be prepared before leaving might influence the drivers to plan and do lasting countermeasures.

It should be kept in mind that the present results are based on the drivers’ own reports. This means a certain risk of poor recall or of social desirability. Still, the questionnaire was based on the impression from focus groups and the correspondence between those impressions and questionnaires was rather good. Another limitation is that the countermeasures applied will depend on the context. There will probably be a difference due to time on task, circadian, environmental or other external factors. The response rate differed between the age groups and the lowest response rate was observed among young drivers. This is normally the case in questionnaires in Sweden. However, also here the results for the young group were similar to the impression from the discussion with focus groups with young drivers.

In the present study, 10 predictors were used in two separate univariate analyses. Assuming independence, one might expect one predictor to show a significant beta weight. On the other hand, the results show many more significant results and interpretable patterns. Thus, there seems no need to correct for the number of variables used but rather to caution that one of the significant beta weights might be spurious.

In conclusion, the present study showed that efficient countermeasures like taking a nap or stopping for coffee was more common among those with experience of sleepy driving, professional drivers, males and higher age groups. Sleepiness monitoring was seen as useful by 32% and mainly by those with experience of sleepy driving, males and non-professional drivers. It is suggested that different groups may require different approaches in terms of approaches to change behaviour.


This study was supported by the Swedish National Road Administration and the EU project SENSATION.