SLEEPINESS

Night Driving, Season, and the Risk of Highway Accidents

Torbjörn Åkerstedt PhD,1 Göran Kecklund PhD,2 and Lars-Göran Hörte PhD3

1National institute for Psychosocial Factors and Heath and Department of Public Health Sciences, Karolinska Institute, Stockholm, Sweden;
2National institute for Psychosocial Factors and Heath, Stockholm, Sweden; 3Department of Public Health Sciences, Karolinska institute, Stockholm,
Sweden

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

Abstract: Official accident and traffic density statistics on Swedish high- related accidents approximately doubled the nighttime peak for total acci-
ways were used to compute the relative risk (Odds Ratio - OR) of being dents. During the winter, the peak of total accidents occurred at 03:00h
injured or killed in a traffic accident at different times of day. After remov- (OR=3.8, Ci=3.5-4.0), five hours before sunrise, whereas the summer
ing accidents due to alcohol 10344 accidents remained for computations, peak occurred at 04:00h (OR=11.6, Ci=11.3-11.9), shortly after the early
and the period 10:00h—11:00h was used as the reference point. The summer sunrise and with consistently higher nighttime risk than for winter
highest total risk was seen at 0400h (OR=5.7, Confidence interval = 5.6- driving. It was concluded that early morning driving is several times more
5.8), with an OR of 11.4 (Ci=10.3-12.5) for fatal accidents at the same dangerous than driving during the forenoon. Apart from alcohol the effect
point. The same pattern was exhibited by single vehicle, head-on, and seems related to sleepiness, but not to darkness.
"other" (e.g., turning off the road) accidents, whereas overtaking and rear- Key words: Driving, accidents; highway; car; time of day; light; sleepi-
end accidents did not show clear 24 hour patterns. Retaining alcohol- ness; fatigue; circadian; sleep loss

INTRODUCTION particular importance to try to estimate the exact time of the peak,
as well as the actual risk of being on the road at different times
OFFICIAL ROAD SAFETY STATISTICS USUALLY SHOW of day. The concept of risk probably has a higher relevance for
MUCH LOWER ACCIDENT RATES AT NIGHT COMPARED the individual contemplating driving at night than the traditional
TO DAYTIME, WHEN PEAK LEVELS ARE REACHED IN accident rate per mile driven. There is also some evidence that
THE MORNING AND AFTERNOON.1 This pattern is very sim- the rate of accidents, based on miles driven, may be a confound
ilar to the 24 hour pattern of traffic density, suggesting a rather since mileage is correlated with experience. Thus, a large number
logical relation, with accidents being dependent on the number of of miles per year may actually be associated with a reduced risk
cars on the road. This should encourage night driving for higher of an accident.5
safety, and, to some extent, such a trend is already in progress. It was also the intention to investigate whether darkness may
On the other hand, there appears to be a higher overall risk at be involved in the higher accident rate at night. For this purpose
night in terms of accident per mile driven.2,3,4,5,6,7 It is not clear, we used the accident patterns around the winter and summer sol-
however, whether the higher rate is uniform during the night or stices, since these two times of year differ greatly in daylight pat-
whether there exists a clear peak or, for that matter, whether there tern at 58°-61°N, from which latitudes most of the present data
exists a trough of sharply reduced risk during the day. were obtained.
Much of the increased risk at night has been attributed to dark-
ness or alcohol. An alternative hypothesis involves sleepiness, METHODS
since sleep-related accidents show a pronounced peak around
03:00—05:00 in the morning.8,9,10,11 On the other hand, the term Data on police reported accidents (involving an injury to the
"Sleep related" is, a diffuse and largely subjective concept, clear- driver) on five major highways (E20, E4, E6, E18 and E66) dur-
ly biased towards single vehicle accidents, especially in the early ing five years (1987—1991) were obtained from Statistics
morning. An alternative approach may be to reverse the logic and Sweden. This yielded 12,535 injuries due to road accidents, but
look at all types of accidents (head-on, overtaking, etc.), regard- 10344 remained after excluding injuries to passengers and acci-
less of suspicion of sleepiness, while excluding alcohol and while dents involving alcohol. The latter was based on one item in the
using seasonal differences in lighting to control for the darkness accident report form, and checking of this item is normally based
effect. Even if snow and ice would add to the risk of winter driv- on a positive blood sample (›0.05% blood alcohol level), usually
ing, the effect should be evenly distributed across the 24-hours. subsequent to a positive breathalyzer test. The latter is normally
The purpose of the present study was to provide a detailed 24h carried out if there is any suspicion of alcohol, based on erratic
pattern for the risk (alcohol excluded) of being involved in any driving, smell, intoxicated behavior by the driver or passengers,
type of highway accident, resulting in injury or death. It was of etc. Fatal accidents virtually always involve an autopsy with
screening for alcohol and other drugs (amhpetamine, marijuana,
sedatives, and similar substances).
Accepted for publication March 2001 Most of the highways were four-lane type motorways,
Address correspondence to: Torbjörn Åkerstedt PhD, IPM & Public Health although a significant portion had only two lanes. The recorded
Sciences, Karolinska Institute, Box 230, 17177 Stockholm, Sweden; accidents were characterized as: injuries due to single vehicle
Tel: 46 8 7286945; Fax: 46 8 320521; E-mail:[email protected] accidents, to rear-end accidents, to overtaking/change of lane

SLEEP, Vol. 24, No. 4, 2001 401 Night Driving—Akerstedt et al

 Accidents Exposure Prop
2000 3.0 1.0

Ot
2.5 0.8
1500
2.0
0.6 Ho

1000 1.5 Re
Ov
0.4

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

1.0
500
0.2
0.5
Si

0 0.0 0.0
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time of Day Time of Day

Figure 1—Left: Exposure (traffic density) and accidents involving an injury to the driver, or death, plotted at different times of day in one-hour intervals.
Right: Proportion of different types of accidents occurring at different times of day. Each field represents the proportion of the total of each type at each time of day.
Ot=other accidents, Ho=head-on accidents, Re=rear-end accidents, Ov=accidents involving overtaking or change of lane, Single=single vehicle accidents

accidents, to head-on accidents, and those due "other" accidents. month, that is, for May + June and November + December (based
The latter largely involved accidents with vehicles turning off the on the exposure values for these two periods). Two months were
road, meeting another vehicle at intersections or running into ani- needed to obtain a sufficient number of cases. In order to explic-
mals (mainly moose, deer, and reindeer). itly compare driving during the short summer nights with long
Traffic flow measures (vehicles per hour and day) were winter nights, the odds ratio for summer driving was computed
obtained from the National Road Administration for each day of for each time of day, using the winter period as the "unexposed"
the year of 1990 at seven measurement points on four of the high- condition. Finally, a time of day analysis was carried out also for
ways. E66 was excluded because of measurement problems. The weekday and weekend driving separately, followed by a compu-
results were multiplied by five to form an estimate of the five tation of weekend risk, using weekday as the unexposed condi-
years 1987—1991. The traffic flow changed very little across tion. Weekend was defined as the time between 00:00h on
these years and should not have any effect on the subsequent Saturday to 24:00h on Sunday.
analyses.
All data were first organized in one-hour bins. A risk ratio was RESULTS
then computed as the total number of accidents/number of vehi-
cles/hour (absolute risk per hour). To compute the relative risk, as Figure 1 shows a bimodal distribution of exposure and acci-
the odds ratio (OR), at different times of day, the bin between dents with clear morning and afternoon peaks for both types of
10:00h and 11:00h was used as reference (labeled "11:00h"). This measures. The two patterns are very similar. After removal of all
reference point was preferred to, for example, a mean across the injuries to passengers and all accidents which involved alcohol,
24 hours or across the daylight hours since it may be easier to 10344 accidents resulting in injury or death remained, of which
communicate the implication of risk if an easily identifiable time 6.1 % (629) were fatal. Out of the 10,344 accidents, the 2,604
of lowest risk is used as the reference. The time between 10:00h single vehicle ones accounted for 25.2%, the 2,622 rear end acci-
and 11:00h was particularly suitable since it represented a rela- dents for 25.3%, the 1285 head-on accidents for 9.5%, the 987
tively stable plateau after the morning rush hour. overtaking accidents for 7.0%, and the 3,406 "other" accidents
For each risk ratio a 95% confidence interval was computed as for 32.9%. Figure 1 also shows that single vehicle accidents dom-
2
OR1±v√1.96/Chi .12 These analyses were carried out for the total inated the night hours. In addition, 683 accidents involved suspi-
number of accidents involving an injury to the driver, as well as cion of alcohol, accounting for 6.19% of the total (10344 + 683).
for: single vehicle, head-on, overtaking, rear end, and "other" All computations of risk below are based on the 10344 accidents
accidents. For most ratios the number of accidents were well not involving alcohol.
above 20, but for night time accidents (02:00h - 06:00h) overtak- Figure 2 shows that, compared with driving between 10:00
ing and rear end accidents fell to between five and ten, and for and 11:00 am, the relative risk of having an injury was five-fold
overtaking, the level fell below five for this period, which makes at 04:00 in the morning and well above unity from 23:00h to
the estimates highly unreliable. 06:00h. Incidentally, the corresponding absolute risk per driver
In addition, the risk at different times of day was computed for on the road around 04:00 (i.e., at 03:00—04:00h) was approxi-
the month of the summer and winter solstice plus the preceding mately 0.026% and the daytime values oscillated around

SLEEP, Vol. 24, No. 4, 2001 402 Night Driving—Akerstedt et al

 0 4 8 12 16 20 24 0 4 8 12 16 20 24
13 13
12 Total 12 Injuries
11 Single 11 Dead
10 10
Odds Ratio

9 9
8 8
7 7
6 6

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

5 5
4 4
3 3
2 2
1 1
0 0
7 7
Other Head on
6 Rear End 6 Overtaking

5 5
Odds Ratio

4 4

3 3

2 2

1 1

0 0
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time of Day Time of Day

Figure 2—Odds ratio (95% Confidence interval) for being involved in a highway accident at different times of day, resulting in injury or death. The reference time is
10:00h—11:00h, with OR=1 (dashed, horizontal line). The different types of accidents (including total) all involve injury or death as outcomes. In the upper right panel
total accidents have been separated in the categories "injuries" and "dead.”

0.0006% per hour. The relative risk of other accidents (turning, collision with an
Fatal accidents showed approximately the same relative risk animal) was also clearly increased during the night (four-fold at
pattern as the total number of accidents, but the increase in risk 04:00h), decreased during the forenoon, increased slightly during
was eleven-fold around 04:00h and there was a high relative risk the afternoon, and started to rise after 21:00h. The number of
throughout the night. The percentage of fatal accidents (of the accidents varied between 7 and 100 per hour, with the lowest lev-
total number) at different times of day varied between 3.4 and els during the early morning.
10.4%, with the highest value around 06:00h. Accidents related to overtaking did not show any clear time of
With regard to type of accident, the relative risk of having a day pattern and had odds ratios close to unity. In addition, the
single vehicle accident at 04:00h was increased twelve-fold and number of accidents between 01:00 and 06:00h was less than
stayed high all the time between 24:00h and 06:00h. The rest of five, which makes the estimates highly unreliable.
the day it did not deviate from unity. The number of accidents Rear end accidents showed a nighttime increase in relative
varied between 63 and 183, with the lowest figures for the night risk, but also clear morning and afternoon-evening peaks, reach-
hours. ing a four-fold risk level around 17:00h. The number of accidents
Also the risk of head-on accidents was increased during the varied between 9 and 904 per hour, with the lowest levels in the
night hours (with a six-fold value around 03:00h), but remained early morning.
at high levels during most of the day. The number of accidents The relative risk of the alcohol-related accidents is not pre-
varied between 9 and 168, with low values of 9—20 during the sented since the extremely low daytime levels make the night-
early morning hours. time risk almost astronomical at 04:00h. The Odds Ratio reached

SLEEP, Vol. 24, No. 4, 2001 403 Night Driving—Akerstedt et al

 0 4 8 12 16 20 24 0 4 8 12 16 20 24
12 9
11 Winter Weekday
8
10 Summer Weekend
7
9
8 6
7
5
6
4

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

5
4 3
3
2
2
1 1

0 0
4 9
Summer/Winter Weekend/Weekday
8

3 7

5
2
4

3
1
2

1
0 0
0 4 8 12 16 20 24 0 4 8 12 16 20 24
Time of Day Time of Day

Figure 3—Upper panels: Odds ratio (95% Confidence interval) for being involved in a highway accident at different times of day, resulting in injury or death. The val-
ues are given for winter (November + December), summer (May + June), weekdays and weekends. The reference time is 10:00h—11:00h (OR=1; dashed, horizon-
tal line).
Lower panels: Odds ratio (95% confidence interval) for being involved in a highway accident at different times of day, resulting in injury or death. Left panel: for the
summer, using winter values as reference (OR=1). Right panel: for weekends, using weekdays as reference (OR=1)

136, for example. The percent of total accidents (based on 10,344 high as the winter risk at 04:00—06:00h in the morning, where-
+ 683) with alcohol involved was above 20% between 01:00h as the daytime and evening risks were clearly reduced in the sum-
and 06:00h, varied around 2—3% during most of the day, and mer. During the beginning of December the times of sunrise and
started to increase above 5% after 20:00h. sunset are 08:19h and 14:54h respectively, and during the begin-
The number of alcohol-related fatal accidents was 73, which ning of June the times are 0347h and 2145h. The number of acci-
is too small a number for a proper risk analysis with respect to dents per hour ranged between 27 to 228 and 13 to 255 for sum-
time of day, but the percentage of the total number of fatal acci- mer and winter, respectively, with the lowest values around
dents (629 without alcohol plus 73 with alcohol = 702) was 03:00h—05:00h.
10.4%. The percentage of fatal accidents (of 702) was 42.3% Figure 3 also shows that the night-time peak in relative risk
from midnight to 03:00h, 25% from 03:00h to 06:00h, 7.5% from was present in both weekend and weekday driving, although the
06:00h to 09:00h, below 5% up to 18:00h, 15.5% between 18:00h former extended long into the morning hours. In fact, it exceed-
and 21:00h, and 23.6% between 21:00h and 24:00h. ed that of weekday driving during the whole night, up to 08:00h
Figure 3 shows that the (total) accident risk during the sum- or 09:00h in the morning, with a very clear peak at 06:00h.
mer (May + June) had a very high peak at 04:00h, whereas that Weekend driving accounted for 31.5% of all injuries and deaths
of winter driving (November + December) occurred at 03:00h with a peak of 72% at 03:00h and a bottom around 07:00h to
and was considerably lower. The summer risk was about twice as 09:00h, with 13%-15%. The number of accidents per hour varied

SLEEP, Vol. 24, No. 4, 2001 404 Night Driving—Akerstedt et al

between 37 and 992 with the lowest value around 0300h. poral pattern differed—the time at high risk was extended into
the late morning. The reason for this is not clear but it is likely
DISCUSSION that the factors behind sleepiness (extended time awake, short
prior sleep, etc) may be more pronounced in connection with
The rate of accidents (involving injuries or death) on main weekend driving. Possibly also the duration of driving may be
highways showed a clear time-of-day pattern with peaks in the extended in order to reach distant goals outside normal weekday
morning and late afternoon, and the same pattern was seen for reach. In addition, drugs and perhaps after effects of alcohol con-
traffic density. However, the actual risk of injury (controlling for sumption may be involved. In any case, the weekend effect needs
exposure) instead showed a sharp peak around 04:00h, approxi- to be especially considered since a significant proportion of total
mately five times the forenoon levels. Obviously, then, the night- road accident injuries and deaths occur then.
time accidents did not fall from daytime values to the same extent The present study focused on accidents in which alcohol
as did traffic flow.

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

intoxication was controlled for. But it should be emphasized that
When fatal accidents only were considered, the peak reached the time of day pattern of accident frequency is strongly influ-
11 times forenoon levels. The pattern was similar for head-on enced by alcohol, particularly during the early night hours—it
collisions and "other" accidents, while it was even more pro- accounts for up to 20% of the accidents at night. And there is
nounced for single vehicle accidents. Accident risk involving some risk that the elimination of alcohol-related accidents may
overtaking did not show any clear pattern, whereas rear-end acci- not have been completely successful since investigators in injury
dents showed a tendency towards an increase in the afternoon. cases may have failed to perceive signs of alcohol involvement.
The increased nighttime risk clearly agrees with previous On the other hand, alcohol-related fatal accidents should have
work.2,3,4,5,6,7 The exact position of the peak had very narrow con- been completely eliminated based on the compulsory autopsies.
fidence intervals and even if there is a lack of similar studies to Consequently, since the risk pattern of fatal accidents is very
compare with, the position of the peak time clearly tallies with close to that of all accidents with injury or death, it seems rea-
previous data on the rate of fatigue-related accidents.8,9,10 The sonable to conclude that the latter pattern reflects accidents not
same early morning peak is also seen when professional drivers involving alcohol. Other drugs could be involved in nighttime
are questioned about when fatigue-related accidents/incidents accidents, but there are no reliable data available. It should be
tend to occur,11,13 and also appear in in-depth investigations of emphasized that also the 73 fatal accidents that did involve alco-
accidents.14 Also single vehicle accidents, used as proxy for hol tended to occur during the night hours. However, they clear-
sleep-related accidents, show a clear early morning peak,15,16 as ly did not account for a majority of the nighttime fatalities.
do sleep intrusions into electroencephalographic (EEG) record- Suicidal behavior is another possible cause of nighttime acci-
ings of truck drivers,17,18 and train drivers.19 dents. The number of such accidents is rather small,25 however,
Sleepiness as a cause of the increased nighttime risk is also and seems to involve mainly head-on accidents, which account-
supported by laboratory studies showing pronounced sleepiness ed for less than 10% of the total number in the present study.
(including sleep intrusions into the EEG) towards the end of a Driving in darkness was hypothesized as one potential cause
night awake.20 The main causative factors here seem to be the of nighttime accidents, because of poor visibility and perhaps
combination of being active at the time of the circadian trough reduced stimulation. At the latitude of Stockholm (around 59°N),
and having spent a long time awake.21 This may be exacerbated the sun was below the horizon between 14:54h and 08:19h dur-
by drivers having slept too little,22 and possibly also to accumu- ing the beginning of December, and the accident peak occurred
lated sleep loss.23 Still, it should be emphasized that the present around 03:00h, five hours before sunrise. However, the early
data do not contain sufficient information for drawing conclu- morning peak was twice as high during early June, when the sun
sions about the extent of fatigue involvement in night-time acci- was below the horizon between 21:45h and 03:47h, and the peak,
dents. In addition, one shouldn't disregard other causes of sleepi- thus, occurred after sunrise, at 04:00h. This does not support a
ness at the wheel, such as sleep apnea and other sleep distur- notion that darkness would be a major cause of the increased
bances,24 which could exacerbate the other fatigue-inducing accident risk at night. Interestingly, in the summer the daytime
aspects of night driving. risk was clearly below that of winter driving, presumably because
Another factor, partly related to sleepiness, may be monotony, of the effects of ice and snow during the winter. Thus, it seems
since the reduced traffic density at night might be likely to induce likely that the increased late night risk during the summer
such a state. The present study shows that the morning drop in involves an underestimation of the risk if snow and ice could be
overall in risk occurs simultaneously with the increase in traffic. controlled for.
On the other hand, the risk during the weekend drops from a high It can also be speculated that driving behavior might be dif-
level at 06:00h to much lower levels at 07:00h and 08:00, despite ferent at night. For example, the duration of the drive, might be
the fact that traffic density remains low or has risen only longer, perhaps caused by a desire to reach ones destination at
marginally. Thus, monotony may not be a strong influence. any cost. It is also conceivable that less traffic may tempt the
Even if traffic density is not the same as mileage, one should driver to increase speed. A third possibility may be that night
also consider the observation that the rate of accidents, based on driving could attract more incompetent or careless drivers. None
miles driven, may be subject to confounding since mileage is cor- of these possibilities have any empirical support, however.
related with experience. Thus, a large number of miles per year Interestingly, accidents due to overtaking seem to have no
may actually be associated with a reduced risk of an accident.5 nocturnal orientation. One may assume that reduced alertness is
The weekend risk was also considerably higher than the week- virtually absent in drivers overtaking another vehicle. Thus, a
day risk, as has been demonstrated in most of the studies cited nocturnal peak may be less expected, particularly if one also con-
above. However, in the present data we also found that the tem- siders the fact that the need for overtaking might be less with

SLEEP, Vol. 24, No. 4, 2001 405 Night Driving—Akerstedt et al

fewer vehicles on the road. Rear end accidents could arise in sit- Verkehrsmed 1955;1:19-30.
uations of reduced alertness, but it seems more likely that such 14. Summala H, Mikkola T. Fatal accidents among car and truck
accidents would occur in dense (daytime) traffic, as suggested by drivers: effects of fatigue, age and alcohol consumption. Hum Factors
the morning and afternoon peaks in risk. 1994;36:315-326.
15. Harris W. Fatigue, circadian rhythm and truck accidents. In: Mackie
In the present paper we have identified the peak risk hours
RR, eds Vigilance. New York: Plenum Press, 1977:133-147.
during the 24-hour span, but one need also to consider the trough 16. Hamelin P. Lorry driver's time habits in work and their involvement
hours—mainly during the forenoon hours around 10:00h to in traffic accidents. Ergonomics 1987;30:1323-1333.
11:00h. We have no obvious explanation for this timing but it 17. Kecklund G, Åkerstedt T. Sleepiness in long distance truck driving:
seems reasonably to assume that an optimal traffic density, the an ambulatory EEG study of night driving. Ergonomics 1993;36:1007-
rising metabolism/alertness, the lower probability of having dis- 1017.
turbed night sleep or an extended time awake26 all may con- 18. Mitler MM, Miller JC, Lipsitz JJ, Walsh JK, Wylie CD. The sleep
tribute. At present we may only speculate, but this time of day, of long-haul truck drivers. N Engl J Med 1997;337:755-761.
with low accident risk, is worthy of study in its own right. 19. Torsvall L, Åkerstedt T. Sleepiness on the job: continuously mea-

Downloaded from https://academic.oup.com/sleep/article/24/4/401/2750016 by St. Lawrence College user on 21 November 2023

sured EEG changes in train drivers. Electroencephalogr Clin
Another interpretation of the somewhat elevated risk after the
Neurophysiol 1987;66:502-511.
forenoon is the influence of a circasemidian (12 hour) rhythm in 20. Åkerstedt T, Gillberg M. Subjective and objective sleepiness in the
alertness. A moderate afternoon peak in sleepiness has been active individual. Int J Neurosci 1990;52:29-37.
demonstrated in a number of laboratory studies27 there is a clear 21. Dijk DJ, Duffy JF, Czeisler CA. Circadian and sleep-wake depen-
afternoon increase in sleep-related road accidents.8,10,28 dent aspects of subjective alertness and cognitive performance. J Sleep
It should be emphasized that the present data may only be gen- Res 1992;1:112-117.
eralized to highway driving, which accounts for 55% of the total 22. Jewett ME, Dijk D-J, Kronauer RE, Dinges DF. Dose-response
amount of accidents and 66% of the fatal ones, according to the relationship between sleep duration and human psychomotor vigilance
and subjective alertness. Sleep 1999;22:171-179.
official statistics.29 Thus, the excess nighttime risk still accounts
23. Dinges DF, Pack F, Williams K, et al. Cumulative sleepiness, mood
for a moderate proportion of the total number of accidents.
disturbance, and psychomotor vigilance performance decrements during
However, with increasing night driving this proportion is likely a week of sleep restricted to 4-5 hours per night. Sleep 1997;20:267-277.
to increase. 24. Lyznicki JM, Doege TC, Davis RM, Williams MA. Sleepiness,
In summary, the results demonstrate that the risk (excluding driving, and motor vehicle crashes. JAMA 1998;279:1908-1913.
alcohol) of being injured in a highway accident is increased five- 25. Ohberg A, Penttila A, Lonnqvist J. Driver suicides. Br J of Psychiat
fold in the early morning, probably because of sleepiness induced 1997;171:468-472.
by the circadian system. This should have implications for traffic 26. Philip P, Taillard J, Gilleminault C, Quera Salva MA, Bioulac B,
policy as well as for individual decisions to drive at night. Ohayon M. Long distance driving and self-induced sleep deprivation
among automobile drivers. Sleep 1999;22:475-480.
Clearly, there is a need for detailed field studies of the prevalence
27. Lavie P. Ultrashort sleep-waking schedule. III. "Gates" and "for-
of fatigue/sleepiness on the highways.
bidden zones" for sleep. Electroencephalog Clin Neurophysiol
1986;63:414-425.
REFERENCES 28. Pack AI, Pack AM, Rodgman E, Cucchiara A, Dinges DF, Schwab
CW. Characteristics of crashes attributed to the driver having fallen
1. NCSA/NHTSA. Traffic Safety Facts. National Center for Statistics
asleep. Accid Anal & Prev 1995;27:769-775.
and Analysis / National Highway Traffic Safety Administration, 1998.
29.SCB/SIKA. Vägtrafikskador 1998. Statiistiska Centralbyrån, 1999.
2. Williams AF, Karpf RS. Deaths of teenagers as passengers in motor
vehicles. Accid Anal & Prev 1983;15:49-54.
3. Williams AF. Nighttime driving and fatal crash involvement of
teenagers. Accid Anal & Prev 1985;17:1-5.
4. Williams AF, Lund AK. Night driving curfews in New York and
Louisiana: Results of a questionnaire survey. Accid Anal & Prev
1985;17:461-466.
5. Massie DL, Campbell KK, Williams AF. Traffic accident involve-
ment rates by driver age and gender. Accid Anal & Prev 1995;27:73-87.
6. Massie DL, Green PE, Campbell KL. Crash involvement rates by
driver gender and the role of average annual mileage. Accid Anal & Prev
1997;29(5):675-685.
7. Doherty ST, Andrey JC, MacGregor C. The situational risks of
young drivers: the influence of passengers, time of day and day of week
on accident rates. Accid Anal & Prev 1998;30:45-52.
8. Langlois PH, Smolensky MH, Hsi BP, Weir FW. Temporal patterns
of reported single-vehicle car and truck accidents in Texas, USA during
1980-1983. Chronobiol Int 1985;2:131-146.
9. Lavie P, Wollman M, Pollack I. Frequency of sleep related traffic
accidents and hour of the day. Sleep Res 1987;16:275.
10. Horne JA, Reyner LA. Sleep related vehicle accidents. Br Med J
1995;310:565-567.
11. Maycock G. Sleepiness and driving: the experience of UK car
drivers. J Sleep Res 1996;5:229-237.
12. Rothman KJ, eds. Modern epidemiology. Boston/Toronto: Little,
Brown and Company, 1986.
13. Prokop O, Prokop L. Ermüdung und Einschlafen am Steuer. Zbl

SLEEP, Vol. 24, No. 4, 2001 406 Night Driving—Akerstedt et al