EXECUTIVE SUMMARY
December
1996
Disclaimer
- Please Read
INTRODUCTION
The Driver Fatigue and Alertness Study
(DFAS) was the largest and most comprehensive over-the-road study ever
conducted on driver fatigue and alertness in North America. It
provides extensive information on the alertness, driving performance,
and physiological and subjective states of commercial motor vehicle (CMV)
drivers as they perform real-life, revenue-generating trips. This
Executive Summary overviews the objectives, methods, principal
findings, and safety implications of this landmark 7-year study.
STUDY
CHRONOLOGY AND PARTICIPANTS
The DFAS was initiated in 1989 by the
Federal Highway Administration's (FHWA) Office of Motor Carriers (OMC)
in response to a Congressional directive contained in the Truck and
Bus Safety and Regulatory Reform Act of 1988. Field data collection
was conducted in 1993 and the project was completed in 1996. The
overall cost of the study was US$4.45 million.
The DFAS was both a public-private and
an international partnership. In addition to the funding provided by
the FHWA, the Trucking Research Institute (TRI) of the American
Trucking Associations Foundation and Transport Canada funded a
significant portion of the data collection and analysis effort. The
TRI, the National Private Truck Council, the International Brotherhood
of Teamsters, and the Owner-Operator Independent Drivers Association
provided considerable input in public forums. These organizations, as
well as the Canadian Trucking Association and the Private Motor Truck
Council of Canada, helped recruit motor carriers and drivers and
provided technical and operational support to the research effort.
Over-the-road data were collected both in the U.S. and Canada.
Numerous organizations and individuals shared their views and
suggestions regarding the study with the project team during
publicly-announced consultation sessions and/or individual
discussions.
Essex Corporation, Columbia, Maryland,
was the principal research organization conducting this study.
Supporting organizations included the Scripps Clinic and Research
Foundation of La Jolla, California; Miller Ergonomics of Imperial
Beach, California; Deaconess Hospital, St. Louis, Missouri; and the
Sleep Disorders Centre of Metropolitan Toronto, Toronto, Ontario,
Canada. Three motor carriers provided drivers, vehicles, and the real
world less-than-truckload (LTL) operational setting for the study.
PROJECT
PUBLICATIONS
Four principal publications have been
produced to report the findings of this study:
-
A 6-page "
Highlights
"
provides background and a top-level overview of the study and its
findings.
-
A 15-page
Executive Summary
[this document].
-
A 59-page
Technical Summary
provides a moderately-detailed account of the study. This and the
above two documents are currently available at no charge from:
-
FMCSA, Office of Research and Technology, 400 Virginia Ave., SW, Suite 600, Washington, DC 20024, telephone: 202-385-2388, fax 202-385-2422.
-
Transportation Development Centre,
Transport Canada, 800 Rne Lvesque Blvd. W., 6th Floor, Montral,
Qubec H3B 1X9 Canada, fax: (514) 283-7158, e-mail:
tdccdt@tc.gc.ca
.
-
An approximately 500-page
Technical
Report
provides comprehensive and detailed information
and data. The Technical Report is expected to be published in
late 1996 or early 1997. It may be purchased from the National
Technical Information Service, Springfield, VA 22161, telephone
(703) 487-4650, e-mail: (prepaid orders)
orders@ntis.fedworld.gov
;
(inquiries)
info@ntis.fedworld.gov
BACKGROUND
Driver fatigue is a safety issue of
special concern to CMV transportation. Under current U.S. Federal
hours-of-service (HOS) regulations, CMV drivers may drive up to 10
hours after a mandatory 8-hour off-duty period. In Canada, the maximum
driving time is 13 hours. Many CMVs often run at night, and drivers
sometimes have irregular and unpredictable work schedules. Most of
their mileage is compiled during long trips on Interstate and other
limited-access highways. Because of the CMVs' high annual mileage
exposure (often 5-10 times that of passenger vehicles) and other
factors, commercial drivers' risk of being involved in a
fatigue-related crash is far greater than that of non-commercial
drivers -- even though CMV drivers represent a relatively small
proportion of all drivers involved in fatigue-related crashes. In
addition, many other crash causation factors, such as alcohol use,
speeding, and other unsafe driving acts, are generally less common in
crashes involving commercial drivers. Thus, fatigue is a relatively
larger concern for these CMV drivers and their vehicles.
HISTORY
OF DOT'S FATIGUE FOCUS
The maximum amount of time that CMV
drivers, operating in U.S. interstate commerce, may drive their
vehicles is specified in Title 49, Code of Federal Regulations, in
Part 395. (In Canada, it is the federal "Commercial Vehicle
Drivers Hours of Service Regulations, 1994"; SOR/DORS/94-716, 15
November 1994.) The U.S. regulations were originally developed in 1935
by the Interstate Commerce Commission (ICC) to counteract perceived
unsafe driver scheduling practices. In 1938, the ICC requested the
U.S. Public Health Service to conduct an investigation into CMV HOS in
interstate commerce. This was the first scientific study to address
driver fatigue as related to HOS. The Public Health Service study
supported the need for regulatory limitation of HOS to help ensure
highway safety. In 1967, the ICC's responsibilities concerning CMV
driver and vehicle safety were transferred to the Bureau of Motor
Carrier Safety (now OMC) of the FHWA, an agency within the then
newly-created U.S. Department of Transportation (DOT).
The DOT conducted three studies on CMV
driver fatigue between the 1970s and the present. None resulted in
changes being made to the Federal HOS regulations. In 1988, the
Congress directed the DOT to conduct research to determine the
relationships among HOS regulations, driver fatigue, and the frequency
of serious accidents involving CMVs. Also in 1988, the FHWA hosted a
Symposium on Truck and Bus Driver Fatigue, which brought together
experts from the motor carrier industry, the scientific and medical
communities, law enforcement, and public policy. The DFAS was
initiated in response to this Congressional directive and its design
was based upon Symposium recommendations. The study began in 1989.
In the 1990s, driver fatigue has
continued to be a major industry and public safety concern. The 1995
FHWA-sponsored Truck and Bus Safety Summit, attended by over 200
national leaders in CMV and highway safety, including a large
contingent of drivers, identified driver fatigue as the top priority
CMV safety issue. Accordingly, the fatigue issue dominates current
FHWA-sponsored human factors research on CMV driving safety.
STUDY
OBJECTIVES
The primary goal of the Driver Fatigue
and Alertness Study was to observe and measure the development and
progression of driver fatigue and loss of alertness, and to develop
countermeasures to address it, through a field study undertaken within
the framework of a realistic driving environment. To accomplish this
goal, several objectives were established:
-
To establish measurable
relationships between CMV driver activities and physiological and
psychological indicators of fatigue and reduced alertness.
-
To identify and evaluate
effectiveness of those alertness-enhancing measures that legally
may be used by CMV drivers. Approximately 500 drivers were
surveyed in 4 locations (west coast, east coast, midwest,
southeast). The FHWA will release the results of this work
separately from the main study report, probably in the late fall
of 1996.
-
To provide a scientifically valid
basis to determine the potential for revisiting the current HOS
requirements, which have been essentially unchanged for more than
50 years.
Secondary goals of the research were to
investigate the potential for utilizing elements of the vehicle- and
driver-based measurements in the development of a system for
monitoring or predicting changes in driver alertness; to identify an
effective subset of data types to improve the efficiency and economy
of conducting future fatigue research in field settings; and to
provide a data set that could be used for validating future fatigue
research using driving simulators.
METHODOLOGY
The methodology and conduct of the DFAS
reflected the research objectives described above. The study
investigated, in an operational context, a number of work-related
factors thought to be related to the development of fatigue and loss
of alertness and degraded performance in CMV drivers. These factors
included:
-
the amount of time spent driving
during a work period,
-
the number of consecutive days of
driving,
-
the time-of-day when driving took
place,
-
the number of hours spent in
principal sleep periods, and
Subjects
Eighty (80) properly qualified male CMV
drivers between the ages of 25 and 65 served as subjects in this
study. Drivers had to have at least one year of experience driving
Class 8 (33,001 pounds and over) tractor trailer combination vehicles
and they had to be medically qualified and free from controlled
substances and alcohol.
Design
The study employed a between-subjects
design involving four driving schedule conditions. Four different
groups of 20 subjects drove in the following schedule conditions,
selected to represent four contrasting driving schedules in terms of
fatigue-related factors such as time-on-task, schedule regularity, and
day versus night driving:
-
Condition 1: 10-hour
"baseline" daytime (C1: 10-hour daytime)
:
10-driving-hour turnaround route, starting at about the same time
(10:00) each morning for 5 consecutive days.
-
Condition 2: 10-hour
"operational," or rotating (C2: 10-hour rotating)
:
10-driving-hour turnaround route, starting about 3 hours earlier
each day for 5 days. The first trip began at about 10:00.
-
Condition 3: 13-hour
nighttime start (C3: 13-hour nightstart)
: 13-driving-hour
turnaround route, starting at about the same time (23:00 on
average) each night for 4 consecutive nights.
-
Condition 4: 13-hour daytime
start (C4: 13-hour daystart)
: 13-driving-hour turnaround
route, starting about the same time each day (13:00 on average)
for 4 consecutive days.
Altogether, there were 360 trips and
about 4,000 hours of driving, distributed more or less evenly across
the four conditions. Conditions 1 and 2 took place in the U.S. between
the cities of St. Louis and Kansas City, Missouri. Conditions 3 and 4
took place in Canada between the cities of Montreal, Quebec and
Toronto, Ontario. The study design was developed to comply with
existing U.S. and Canadian hours-of-service regulations.
The four schedules provided different
amounts of time off between trips. Condition 1 provided about 11 hours
off, while the other three conditions provided about 8 hours off.
Vehicles and Instrumentation
Conventional Class 8 truck tractors
from each participating motor carrier were outfitted with on-board
monitoring equipment and a data acquisition computer. Tractors
included both single-drive-axle and tandem-drive-axle designs. Trailer
configurations included both single semitrailers (45', 48', and 53')
and twin 28' trailers. All participating drivers were completely
familiar with their assigned vehicles.
Driver and Driving Measures
Numerous measures were taken of
drivers' physiology, alertness, and performance during driving and of
their physiology during off-duty sleep. Many data elements were
collected simultaneously, and all data were time-stamped to aid in
analysis. Measures collected for each subject included:
-
Lane tracking (collected using a
device that measured the tractor's lateral position relative to
lane markings)
-
Steering wheel movement
-
Driving speed and distance
monitoring (to aid in data analysis)
-
Performance on three surrogate tests
of tasks related to safe driving performance. Drivers took the
tests before starting their runs, after they reached the
turnaround point halfway through their trip (and, during the two
10-hour conditions, before the return trip commenced to study the
effects of a break), and after the run was completed. The tests
were self-administered while the vehicle was stopped via a CRT
display mounted in the truck cab. Each administration of the set
of tests took about 18 minutes. The surrogate tests were:
-
Code Substitution (a cognitive
test involving number/letter substitution)
-
Critical Tracking Test (a test of
hand-eye coordination, requiring a pointer moving in an
unpredictable manner to be kept at the center of a display)
-
Simple Response Vigilance Test (a
test of vigilance and reaction time).
-
Continuous video monitoring
-
Face video (to permit judgments
of alertness based upon eyelid droop and facial expression and
muscle tone; an infrared illuminator was used to permit night
monitoring)
-
Road video (forward-looking
video recording to permit reconstruction of driving and
traffic events).
-
Physiological measures
-
Polysomnography (PSG) during sleep
-
Electroencephalogram (EEG) using
clinical-type scalp electrodes
-
Electrooculogram (EOG); electrodes
placed at left and right outer canthi (corner of the eyes)
-
Electromyogram (EMG); electrodes
placed on chin
-
Respiratory airflow (nasal sensor)
-
Respiratory effort (sensors on
chest)
-
Oxygen saturation of arterial
blood (finger probe).
-
PSG during driving (EEG and EOG
only).
-
Body temperature during waking
hours (obtained using an infrared ear probe)
-
Electrocardiography (ECG) during
driving and sleep.
-
Driver-supplied information
-
Pre-participation questionnaire on
sleep habits
-
Daily log (stops, meals,
noteworthy driving events, etc.)
-
Stanford Sleepiness Scale rating
(a self-assessment of fatigue and mood).
-
Tractor cab environment
(temperature, relative humidity, 8-hour concentrations of carbon
monoxide and nitrogen dioxide)
Data Analysis
The study developed a massive database
which covers more than 200,000 miles of driving. It includes some
4,000 hours of video data, 9,000 hours of physiological recordings,
and 700 megabytes of real-time truck computer records. Close to a year
was needed to clean the raw field data and to enter them into a
complete project data base. Standard protocols were used for
converting raw data into meaningful metrics; for example, the PSG
sleep data were scored manually using standard clinical criteria to
assign sleep stages. Statistical analysis focused on comparisons of
group means to evaluate the effects of driving schedule (and related
factors such as hours of sleep) for a variety of dependent measures of
driver alertness and performance (as listed above). In addition,
instances of drowsiness during driving were identified and analyzed.
Initial comprehensive reviews were done by two research teams working
independently to assess the physiological and driving-performance
data. The results of these reviews were then compared to clearly
document these events.
Strengths and Limitations of
the Methodology
The strengths of the methodology of the
DFAS were in its naturalistic approach (i.e., data collected during
revenue-generating runs), the enormous volume of data collected, and
in the comprehensiveness of the measurements. The limitations of the
study relate primarily to the lack of full control over the full range
of conditions affecting alertness and fatigue and the inability to
isolate some factors due to unavoidable confounding of variables, a
consequence of the naturalistic approach to the study. For example, a
comparison of Condition 1 (10-hour daytime) and Condition 3 (13-hour
nightstart) shows that the conditions differ in several important
ways: number of continuous hours of driving, percent of night driving,
and number of hours off-duty. In addition, the combination of the
inherent variability of the real world environment and the
between-subjects design meant that there was more uncontrolled
variability ("noise") in the data than would be found in a
study performed within a laboratory setting or a study-controlled
driving environment, or one employing a within-subjects design. The
study's analysis methodology included assessments of the consistency
and inconsistency of results from different measures.
RESULTS
AND DISCUSSION
Project findings are reported below as
they relate to major issues.
Time-of-Day of Driving
The strongest and most consistent
factor influencing driver fatigue and alertness in this study was time
of day. Drowsiness, as observed in video recordings of the driver's
face, was markedly greater during night driving than during daytime
driving. Peak drowsiness occurred during the 8 hours from late evening
until dawn.
Night driving (e.g., from midnight to
dawn) was associated with worse performance on four important criteria
(proportion of video-drowsy analysis periods, average lane tracking
standard deviation, incremental differences in Code Substitution test
scores between the outbound and inbound segments of a trip, and
average physiologically-measured total sleep obtained during the
principal sleep period prior to a trip). Time of day was a much better
predictor of decreased driving performance than hours of driving
(time-on-task) or the cumulative number of trips made.
Duration of Driving
Hours of driving (time-on-task) was not
a strong or consistent predictor of observed fatigue. Most notably,
there was no difference in the amount (prevalence) of drowsiness
observed in video records of comparable daytime segments of the
10-hour and the 13-hour trips. Nighttime segments could not be
similarly analyzed because the study design did not provide for this
comparison.
Lane tracking performance was better in
the 10-hour than the 13-hour conditions. The reasons for this are not
completely clear because of confounding factors associated with
different routes and vehicles.
In the surrogate tests, cognitive
performance (via Code Substitution) was better in the 10-hour
conditions. Vigilance and reaction time (via Simple Response Vigilance
Test) were better in the 13-hour conditions (probably because of loss
of display contrast associated with greater amounts of sunlight in the
10-hour conditions). Hand-eye coordination (via Critical Tracking
Test) did not show condition-related variation.
There was little correlation between
Stanford Sleepiness Scale self-ratings and objective performance test
scores.
However, self-ratings of fatigue
level on the Stanford Sleepiness Scale correlated positively with
time-on-task, indicating that drivers may have the feeling of
increasing fatigue with increasing time-on-task even if there are no
strong performance changes.
Cumulative Fatigue Across Days
There was some evidence of cumulative
fatigue across days of driving. For example, performance on the Simple
Response Vigilance Test declined during the last days of all four
conditions. Also, drivers tended to rate themselves as more fatigued
across multiple trips. However, cumulative number of trips was neither
a strong nor consistent predictor of fatigue across different
measures. Although more apparent drowsiness was noted in video
recordings made in the last two trips of Condition 2 (10-hour
rotating), those trips were, on the average, driven at night (see
statement above concerning night driving). The Stanford Sleepiness
Scale self-ratings of sleepiness increased as drivers progressed
through successive trips within Condition 2, but the trends were
unclear in Condition 3 (13-hour nightstart) and Condition 4 (13-hour
daystart).
Daily Principal Sleep Periods
Overall, drivers obtained about 2 hours
less time in bed and 2.5 hours less actual sleep than their reported
"ideal" daily amount of sleep. The drivers reported an
average "ideal" 7.2 hours per principal sleep period on a
questionnaire completed before their first sleep at the sleep lab; the
average observed time in bed over the course of the study was 5.2
hours. (Although the drivers reported what they considered to be their
"ideal" sleep time, they were not asked, and it is not
known, whether they usually obtained this stated amount.) For the four
conditions, the average times in bed and clinically-measured sleep
times were:
-
Condition 1 (10-hour daytime): 5.8
hours in bed, 5.4 hours asleep.
-
Condition 2 (10-hour rotating): 5.1
hours in bed, 4.8 hours asleep.
-
Condition 3 (13-hour nightstart):
4.4 hours in bed, 3.8 hours asleep.
-
Condition 4 (13-hour daystart): 5.5
hours in bed, 5.1 hours asleep.
The observed shortfall could have been
due, in part, to a reduction in free time due to requirements of the
study protocol. The study setting also created an opportunity for
socializing with other drivers that might not exist in normal driving.
In addition, some drivers did not always organize their off-duty time
wisely to obtain the maximum possible sleep. Time-in-bed was lowest
for the three Conditions (2, 3, and 4) that permitted the least
off-duty time (about 8.6 to 8.9 hours on average, excluding time
required for the study protocol). Nevertheless, even in Condition 1,
which permitted about 10.7 hours off-duty between trips, the average
time-in-bed and time asleep were only 5.8 and 5.4 hours, respectively.
The lower ratio of sleep time to time
in bed for Condition C3 (13-hour nightstart) may reflect circadian
disruptions of sleep pattern in comparison with the other conditions.
This condition was the only one that consistently required drivers to
sleep during the daytime.
Quantity and Quality of Sleep
Obtained
The quantity of sleep obtained by the
subjects in their principal sleep periods was low. As noted above,
drivers obtained an average of about 2 hours less sleep than their
daily "ideal" requirements. The average time-in-bed during
the principal sleep period (i.e., not including naps, which are
addressed below) was 5.2 hours versus a self-reported daily
"ideal" of 7.2 hours. The shortest average time-in-bed (4.4
hours) was associated with Condition 3 (13-hour nightstart); these
drivers had about 8.6 off-duty hours daily beginning at about noon.
All of the drivers obtained efficient,
normally-structured sleep as judged by formal clinical criteria. Of
the 5.2 average hours in bed, the drivers were actually asleep for an
average of 4.8 hours. The average sleep efficiency (sleep
time/time-in-bed) was 0.92; levels above 0.90 are often observed in
people who have no trouble sleeping and in people who are sleep
deprived. The average amount of time awake after sleep commenced was
25 minutes; this value is considered low relative to values in the
normal range (less than 60 minutes for adult men) and is also
consistent with reduced time in bed and with sleep deprivation.
The requirements of the study may have
contributed somewhat to driver sleep deprivation, but the overall
effect appears to be due to a combination of insufficient opportunity
for sleep, and the failure of drivers to place a high enough priority
on obtaining sufficient sleep.
Drowsiness During Driving
Video ratings were much more sensitive
for detecting drowsiness while driving than were polysomnographic (PSG)
measures. The 4,000 hours of video recordings were systematically
sampled at 30-minute intervals. Drowsy episodes discovered were judged
in 6-minute periods from 30 minutes before to 30 minutes after their
occurrence. Approximately 4.9% of the sampled face video segments were
scored as drowsy based on trained reviewers' assessment of such
factors as eye movement, eyelid position, yawns, stretches, and
startles. The proportions of video data scored drowsy were much
greater at night than during the day or evening. Fourteen percent of
drivers accounted for 54% of all observed drowsiness episodes.
All EEG and EOG data were analyzed. PSG
analysis indicated that there were two trips, involving different
drivers (an incidence of about 0.6 % of observed trips and about 2.5 %
of observed drivers), that included a number of intermittent episodes
that were identified as PSG-Drowsy Driving. These periods amounted to
just over 19 minutes out of the 244,667 minutes of driving analyzed
(0.008%). During these periods, the drivers' data presented EEG and
EOG patterns that would have been consistent with clinical criteria
for Stage 1 sleep (the initial, shallowest, sleep stage) if the
drivers had been in bed in a dark room. Face-video records during
these periods also showed driver drowsiness. The EEG measurement may
have revealed a worse (by comparison with the face-video judgments)
and infrequently-occurring condition. However, these differences may
be reflective of the relative sensitivities of the two methods of
detecting drowsy driving.
A comparison of steering and lane
tracking performance for video-rated drowsy versus non-drowsy epochs
indicated that drowsiness was associated with more erratic steering
(greater steering wheel angle variability) and poorer lane tracking
(increased standard deviation of lane position), both of which have
obvious implications for driving safety.
Not surprisingly, there was a negative
correlation between the length of the principal sleep period and
amount of drowsiness during the next driving trip (e.g., more sleep
leads to less drowsiness). However, it was not possible to estimate
the "normal" level of drowsiness during driving since there
were no conditions where all drivers obtained adequate sleep.
Although there were video, PSG, and
driving performance indications of driver drowsiness, there were no
crashes during the study.
Napping
Of the 80 drivers, 35 (44%) took at
least one nap during a duty cycle that contained clinically-scorable
sleep. Drivers who elected to nap increased their sleep obtained in
principal sleep periods by an average of 27 minutes which amounted to
an 11% increase in average daily sleep time. Drowsiness, as evident in
face video recordings, was often a precursor to the driver deciding to
take a nap. Thus it appeared that this behavior was replacement or
compensatory napping, taken in response to self-perceived sleepiness.
Because 45 of the drivers did not nap,
and there were only 63 naps taken over the 360 trips in the study, no
analyses were performed to determine whether these driver naps
resulted in post-nap improvement in alertness and performance. This is
one of many important questions which might be addressed by future
analysis of the data collected in the DFAS.
Effects of Mid-Trip Breaks
In the 10-hour conditions (Conditions 1
and 2), drivers self-administered the surrogate performance tests both
at the beginning and the end of their mid-trip turnaround break. The
only test demonstrating improved post-break performance was the Code
Substitution test. The other performance tests failed to show a
statistically-significant recovery effect.
Driver Self-Awareness of
Fatigue
There was little correlation between
driver subjective self-ratings of alertness/sleepiness and concurrent
objective performance measures. It appears that drivers are not very
good at assessing their own levels of alertness; there was a tendency
for drivers to rate themselves as more alert than the performance
tests indicated.
On the other hand, there was a positive
correlation between self-ratings of fatigue and both the number of
hours of driving within a trip and the cumulative number of trips
made. Perhaps these factors affected the experience of fatigue,
reflecting increasing stress or compensatory effort rather than
objective performance. Or, perhaps drivers were basing their
self-ratings in part on a logical expectation that these factors would
increase fatigue and they would thus be led to respond in kind as they
selected their Stanford Sleepiness Scale rating. If the latter
explanation were true, drivers would in effect be saying to
themselves, "If I've been driving for a long time, then I must be
tired."
Self-ratings did
not
correlate
significantly with trip segments ranked according to percent of night
driving, even though performance measures showed significantly reduced
performance at night than during the day. If the
"expectation" explanation of driving self-ratings in the
previous paragraph is correct, a disturbing corollary would be that
drivers had no expectation that night driving would be associated with
reduced performance, when in fact these performance reductions are
significant.
Individual Differences in
Driver Susceptibility to Drowsiness
There were large individual differences
among drivers in levels of alertness and performance. For example,
there was a wide variation in the total number of episodes judged
drowsy in the video records. Thirty-six percent (36%) of the drivers
were never judged drowsy; of the remainder, 77% (49% of the total)
were judged drowsy 10 or fewer times, and 23% (15% of the total) were
judged drowsy more than 10 times. Among the drivers with more than 10
drowsiness episodes, the number of drowsy episodes ranged from 12 to
40, with an average of 22 episodes during their 4-5 day participation
period.
A further illustration of the wide
individual differences among drivers is the fact that 11 of the 80
drivers (14%) accounted for 54% of all observed drowsiness episodes.
This study did not track the subjects
over extended periods of time to determine if the same drivers showing
frequent drowsiness during the week of the study would show frequent
drowsiness weeks or months later. Thus, it cannot be discerned whether
the observed individual differences were reflective of driver
traits
(i.e., long-term, stable individual differences in
physiology and/or performance) or of driver
states
(short-term differences related to recent sleep or other transient
events). Of course, both traits and states may be operative. Future
research should address the trait versus state issue because it has
implications for the potential effectiveness of improved driver
selection, scheduling, and training as fatigue countermeasures.
Comparisons Among Driving
Schedules
In general, differences in driver
alertness among the four driving schedules (study conditions), as
observed and measured by various driving performance and physiological
means, appeared to reflect differences in amount of night driving,
rather than other factors such as differences in continuous driving
time. There was no difference in the amount of drowsiness observed in
the video data during comparable (i.e., daytime) trip segments of the
10-hour and 13-hour trips. Night-driving segments could not be
similarly compared on a trip-by-trip basis because only the last two
trips of Condition 2 (10-hour rotating) were driven through the night.
Condition 3 (13-hour nightstart) was
associated with the shortest sleep latencies (time required to fall
asleep after going to bed), further indicating that these drivers
received less sleep than they needed. Further, the small amount of
sleep obtained by Condition 3 drivers may have exacerbated the
degraded performance observed at night in the study, since those
drivers performed the greatest proportional amount of night driving.
Sleep Apnea
Although this study was not designed to
determine a population prevalence, analysis of subject sleep revealed
that two of the 80 drivers (2.5%) had clinically-diagnosable apnea, a
sleep disorder characterized by breathing cessations. The driving
performance of these two individuals was not statistically different
from that of other comparable drivers in the study.
Age and Fatigue
No significant relationships were found
between driver age and fatigue. There were no consistent differences
between older and younger drivers in terms of observed drowsiness,
frequency of naps, self-ratings, or driving performance. Older drivers
performed more poorly on the Code Substitution test than younger
drivers, but this effect was not fatigue-related. In order to control
for this effect, the Code Substitution data were grouped by driver age
so that the general age difference in performance did not confound
other comparisons.
Study Findings Concerning
Countermeasures
Although the DFAS was not designed
specifically to support the development of technological
countermeasures, the findings of the study are supportive of their
feasibility. Of the surrogate performance tests employed, the Simple
Response Vigilance Test demonstrated the most promise in detecting
fatigue as it might develop during the course of a trip or
cumulatively across trips. Although the sensitivity of this test to
ambient light level, as found in this study, must be reduced,
surrogate tests might be used as part of fitness-for-duty testing
approaches to detecting driver fatigue.
Changes in driving performance,
measured by increased variability in steering and lane tracking, were
shown to be correlated with drowsiness as judged in video observations
of the driver's face. The correlation between drowsiness and degraded
driving performance supports the concept of continuous monitoring of
driver performance to detect fatigue. A related and complementary
approach to performance monitoring is to directly measure
psychophysiological changes such as the eyelid droop seen in face
videos of drowsy drivers or various PSG indices of reduced alertness.
The DOT and other agencies and organizations are sponsoring a wide
range of research on technological fatigue-detection and prevention
countermeasures. Fitness-for-duty (readiness-to-perform) testing and
continuous driver monitoring approaches are both being assessed. At
the same time, driving performance is also influenced by the design
and condition of the roadway, by the characteristics of the vehicle
being driven, and by the number and location of other vehicles sharing
the roadway. Those influences must be accounted for in the development
of continuous-monitoring systems.
The study also sought to identify any
behavioral methods used by drivers to ward off fatigue. Napping was a
frequent driver-initiated response to drowsiness and fatigue. The
alertness-enhancing effects of napping have been demonstrated in other
operator performance settings (e.g., commercial aviation) and should
be the subject of future research, including additional analysis of
the DFAS database. Other methods will be reported in the
countermeasures survey.
The highly significant time-of-day
effects on fatigue demonstrate that scheduling may be an important
countermeasure to CMV driver fatigue. From the driver fatigue and
alertness standpoint, the optimal schedule is one that appropriately
manages night driving. There are no known highway transportation
hours-of-service regulations in the world that address time-of-day
effect, even though shiftwork literature for many years has pointed
out a strong relationship between time of day and accidents and
incidents.
It cannot be concluded, however, that
shifting truck traffic to daylight hours would result in lower
accident rates. This measure would increase daytime traffic
congestion, possibly with a corresponding increase in accidents, and
would further increase the risk of accidents with passenger vehicles
which are more vulnerable in accidents with trucks because of their
difference in mass. Research is needed to establish the relative risks
of accidents between day and night driving for a variety of road and
vehicle types, and levels of traffic density, to establish the net
impacts on highway safety of day/night scheduling practices.
Another key to enhanced scheduling at
the fleet level may be the finding of large individual differences in
susceptibility to drowsiness while driving, as noted in this study. It
appears that some drivers may be much better than others at
maintaining alertness in the long-haul CMV environment, especially at
night -- a potential basis for driver selection and assignments of
runs should future research prove that these individual differences
are consistent over time.
Implications for Educational
Approaches
Two major project findings relevant to
driver education were the generally inadequate amounts of sleep
obtained by the driver subjects and the strong tendency for drowsiness
to be most associated with nighttime circadian effects. Drivers need
to be educated about how to obtain more sleep, especially if they will
drive at night. Further, study findings showed that drivers were
generally poor judges of their own levels of fatigue/alertness. This
finding indicates a need to train drivers to better assess their
current levels of fatigue while driving, perhaps by learning to become
more conscious of changes in their physical state and subtle changes
in their driving performance.
ASSESSMENT
OF RESULTS FOR FATIGUE MANAGEMENT
There is no quick fix and no single
solution to the fatigue problem. Sleep is the principal countermeasure
to fatigue. All drivers need to ensure that they obtain adequate
sleep. Drivers must also be afforded the opportunity to obtain
adequate sleep.
Changes in the hours-of-service
regulations alone will not solve the fatigue problem. Much can be done
to address driver fatigue through a combination of innovative
hours-of-service regulation and enforcement, education, driver work
scheduling, innovative fatigue management programs, driver screening,
fitness for duty and alertness monitoring systems, and additional
research.
Partnerships among government,
industry, drivers, safety groups, the scientific community, and
shippers are needed for effective solutions to the commercial motor
vehicle driver fatigue problem.
FUTURE
DIRECTIONS
The DFAS demonstrated that it is
possible to conduct a field study with substantial numbers of
commercial drivers (80), hauling revenue freight on many trips (360),
employing instrumentation to record numerous aspects of vehicle
control, surrogate test performance, driver physiology, face and road
video, and sleep studies for each principal sleep period. This was
accomplished without any motor vehicle crashes or other harm to the
study subjects or other participants. The data collected have been
used to document a number of fundamental characteristics of driver
fatigue and alertness and the archived DFAS database will continue to
support analyses of additional questions over the coming years.
Driver drowsiness/fatigue has become
the dominant human factors research issue relating to CMV
transportation. For example, the FHWA/OMC currently sponsors more than
a dozen research and education/outreach projects relating to CMV
driver drowsiness/fatigue. Recently completed, current, or planned
fatigue-related research includes studies on work, rest and recovery;
sleep apnea; multi-trailer vehicle driver stress and fatigue; highway
rest areas; use of on-board recorders for hours-of-service and other
regulatory compliance; driver fitness-for-duty testing; other
technological fatigue countermeasures; fatigue education for driver
and other CMV-related personnel; fleet-based driver wellness; the role
of shippers and receivers in HOS violations; scheduling practices and
fatigue; safety analysis of HOS "restart" options; sleeper
berth usage and fatigue; local/short-haul driver fatigue;
loading/unloading and fatigue; and improved crash causation analysis.
Many of these studies will build upon the research techniques
developed, practical lessons learned, and scientific knowledge gained
from the DFAS.
A separate report, to be published
under the sponsorship of Transport Canada and the Canadian Trucking
Research Institute of the Canadian Trucking Association, will present
the results and analysis of an additional field study that was
performed in coordination with this one. The data base in the Canadian
study covers an additional 55 trips performed under various driving
and days-off schedules lasting up to 10 days. That study will provide
additional information to build on the results of this study,
including results on the influence of different durations of multi-day
off-duty periods.
Return
to Top of Document
|