- Email: [email protected]
The scope and nature of the drowsy driving problem in New York state
A&d. Anal. and Prev., Vol. 28, No. 4, pp. 511-51’7, 1996 Copyright 8 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved oool-4575196 $15.00 + 0.00
Pergamon
THE SCOPE AND NATURE OF THE DROWSY DRIVING PROBLEM IN NEW YORK STATE* ANNET. MCCARTT, ‘8 STEPHENA. RIBNER,~ALLAN I. PACKSand MARK C. HAMMERI ‘Institute for Traffic Safety Management and Research, University at Albany, State University of New York, 80 Wolf Road, Suite 607, Albany, NY 12205-2604, USA, ‘Fact Finders, Incorporated, 2010 Western Avenue, Albany, NY 12203-5015, USA; 3Center for Sleep and Respiratory Neurobiology, University of Pennsylvania, 3600 Spruce Street, Philadelphia, PA 19104-4283, USA (Accepted 27 September 1995)
Abstract-Atelephone survey was conducted of a random sample of New York State licensed drivers to determine the prevalence and circumstances of drowsy driving. Based on the survey responses, 54.6% of the drivers had driven while drowsy within the past year; 22.6% had ever fallen asleep at the wheel without having a crash, 2.8% had ever crashed when they fell asleep, and 1.9% had crashed when driving while drowsy. Of the reported crashes due to driving while drowsy or falling asleep at the wheel, 82.5% involved the driver alone in the vehicle, 60.0% occurred between 11:OOp.m. and 7:OOa.m., 47.5% were drive-off-road crashes, and 40.0% occurred on a highway or expressway. Multiple regression analysis suggested that the following driver variables are predictive of an increased frequency of driving drowsy: demographic characteristics (younger drivers, more education, and men); sleep patterns (fewer hours of sleep at night and greater frequency of trouble staying awake during the day); work patterns (greater frequency of driving for job and working rotating shifts); and driving patterns (greater number of miles driven annually and fewer number of hours a person can drive before becoming drowsy). Copyright 0 1996 Elsevier Science Ltd Keywords-Drowsy
driving, Fatigued driving, Fall-asleep crashes, Drowsy driver crashes, Driver survey
Wales Road Safety Bureau 1993). Further, apart from the compilation of crash data, scant research has been conducted on the number and characteristics of crashes that involve driver drowsiness or falling asleep or on the prevalence of drowsy driving that does not result in a crash. Data for 1989-1993 from the U.S. National Highway Traffic Safety Administration’s General Estimates System (GES) indicate that nationwide there were an estimated 56,000 crashes annually in which driver drowsiness/fatigue was indicated on the police accident report; this represents about 1% of all police-reported crashes. Data from the agency’s Fatal Accident Reporting System (FARS) indicate that either drowsiness or fatigue was cited as a factor in an annual average of 1,357 fatal crashes during 1989-1993, representing 3.6% of all fatal crashes during that time period (Knipling and Wang 1994). The problem of drowsy driving is believed to be especially severe among certain groups of drivers. For example, the National Transportation Safety Board (NTSB 1990) estimates that nearly one-third of fatalto-the-driver heavy truck crashes involve fatigue, and
INTRODUCTION A review of current research on the nature and scope of drowsy driver crashes indicates that there are large gaps in our knowledge about this problem. To date, the understanding of the prevalence and circumstances of drowsy driver crashes in the United States has been based largely on crash data gathered by government agencies. However, it is believed by many researchers that the prevalence of driver drowsiness in crashes is under-reported, since there is generally little evidence upon which to base a finding that drowsiness was a factor in a crash, and since potential liability may motivate a driver to fail to report his or her drowsiness or fatigue. Surveys of motorists conducted in England, Finland, and New South Wales suggest that drowsy driving may be more prevalent than the crash data would indicate (Horne and Reyner 1995; Martikainen et al. 1992; New South *Presented at the 39th Annual Meeting the Advancement of Automotive Medicine: Chicago, Illinois. tAuthor for correspondence.
of the Association for 16-18 October 1995.
511
A.T. MCCARTT et al.
512
that fatigue and fatigue/drug interactions are implicated in a greater number of fatal heavy truck crashes than alcohol and other drugs alone. Recently, the NTSB (1995) conducted a study of nighttime single vehicle heavy truck crashes in which the driver survived; the study found that 58% of these crashes were fatigue-related. Several studies have associated shiftwork, especially rotating shifts, with an increased risk for crashes due to sleepiness or fatigue (Colquhoun 1976; Folkard and Monk 1989; Richardson et al. 1989/1990). Based on their greater need for sleep and their lifestyles, young drivers would also appear to be at high risk for drowsy driving, and crash studies and surveys of youth support this hypothesis (Carskadon 1989/1990; Knipling and Wang 1994; Pack et al. 1995). The limited state of knowledge about the problem of drowsy driving may be contrasted with the level of knowledge of alcohol-related crashes. Although alcohol-related crashes are also known to be under-reported, there are physical indicators of intoxication and reliable tests to determine the blood alcohol concentration of a driver. Based on the availability of these tests and the cumulative results of research on the characteristics of alcohol-related crashes, a good understanding of the scope and characteristics of alcohol-related crashes has been achieved, despite the under-reporting of such crashes. Since there are no reliable measures to indicate that a driver fell asleep or was fatigued, it is important to use alternative research methods in addition to analyses of crash data to study the problem. To document more fully the prevalence and nature of drowsy driving in New York State, a statewide telephone survey was designed and conducted by Fact Finders, Inc., and the Institute for Traffic Safety Management and Research, under the auspices of the New York State Task Force on the Impact of Fatigue on Driving. In addition to information on driver attitudes and reported behaviors, the survey gathered information on a number of factors hypothesized to be associated with drowsy driving. It was hoped that this information would assist in the development of a better understanding of the characteristics of drivers who may be at higher risk for drowsy driving and the types of driving situations that may contribute to drowsy driving. METHODS In order to accomplish the reliable and representative measurement of attitudes, perceptions, and behaviors related to drowsy driving, a scientific random probability sample was designed. The sample was comprised of 1000 randomly selected licensed
drivers (or those with a learner’s permit) residing in New York’s 62 counties. The number of interviews designated for each county was determined by that county’s population of licensed drivers relative to the total population of licensed drivers in the state. Telephone numbers were generated by randomdigit dialing, a sampling technique which assures an equal, unbiased probability of inclusion in the sample for all households with a telephone, including households with unlisted and newly established telephone numbers. In theory, the statistical sampling error associated with the overall findings ranges from *1.8% to 3.1%. Seventeen percent of the drivers contacted declined to participate in the survey. The survey questionnaire was developed by the Institute for Traffic Safety Management and Research and members of the Research Team of the Task Force on the Impact of Fatigue on Driving, and subsequently reviewed and pilot-tested by Fact Finders. The initial draft incorporated portions of a survey instrument developed at the University of Pennsylvania Center for Sleep and Respiratory Neurobiology. Draft versions of the instrument were pilot-tested to insure that the final questionnaire would be efficient, comprehensive, valid, and free from contextual biases. Approximately 30 pilot tests were conducted. This paper describes the results of the telephone survey, with particular emphasis on motorists’ experiences with drowsy driving, including crashes due to falling asleep or drowsiness, and on driver correlates associated with the reported frequency of drowsy driving episodes. Following a summary of the characteristics of the survey respondents, the tabulated results are presented for selected survey items related to the awareness of drowsy driving, drowsy driving experiences in the past year, preventive actions, and lifetime drowsy driving incidents. Next, an examination of driver correlates with the frequency of drowsy driving experiences is presented. The correlates include driver demographic characteristics, and driving, work, and sleep and wake patterns. The examination consists of bivariate analyses of the association between the hypothesized correlates and the frequency of reported drowsy driving within the past year, and multiple regression models with reported frequency of drowsy driving as the dependent variable and driver demographic characteristics and work, driving, and sleep patterns as the independent variables. RESULTS Summary of sample demographics The gender, age group, and region of residence of the survey sample and the population of New York
The scope and nature of the drowsy driving problem in New York State
State licensed drivers are provided in Table 1. The sample was highly representative of the population of licensed drivers. Awareness of drowsy driving issue
With “drowsy” defined as “so tired you could easily fall asleep,” 54.7% of the survey respondents reported that being drowsy greatly affects their ability to drive safely. Survey respondents indicated that being drowsy has a greater effect on their ability to drive safely than either adverse weather or having two drinks of wine, beer, or liquor (Table 2). More than one-quarter of the respondents (26.1%) knew someone who had a crash due to falling asleep at the wheel or drowsiness. Driving while drowsy experiences
The survey results indicated that 54.6% of drivers experienced driving while drowsy in the last year, and 2.5% drove drowsy “very often” (Table 3). All respondents were presented with a series of different situations and asked whether they had driven while drowsy in the last year in each of these situations. The results are presented in Table 4. Of the total survey sample, about one in four drivers reported that they drove drowsy during the past year in each Table 1. Demographics of survey sample and population of New York State licensed drivers
Gender Men Women Age Group 16-24 years old 25-34 years old 35-44 years old 45-54 years old 55-64 years old 65 years or over Region Upstate Long Island New York City
Survey sample (N=1000)
Total NYS drivers (N= 10.3M)
so.19 49.9%
52.8% 47.2%
9.9% 19.6% 26.1% 16.7% 11.9% 15.9%
12.8% 22.9% 22.3% 16.7% 11.5% 13.8%
53.8% 19.0% 27.2%
53.6% 18.8% 21.6%
513
Table 3. Reported frequency of driving while drowsy in last year Frequency of drowsy driving
(N=999)
Have driven drowsy in last year Very often Sometimes Rarely Have not driven drowsy in last year
Table 4. Situations in which drivers experienced drowsiness in Iast year Drowsy driving situations
(N= 1000)
On a long trip At night Due to lack of sleep During the day Due to an illness or medical condition Due to taking medications Due to drinking alcohol
29.5% 27.7% 26.6% 22.9% 4.6% 3.5% 1.8%
Note: All respondents were asked if they had driven while drowsy in the last year in each situation; some respondents had driven drowsy in more than one situation.
of these situations: on a long trip (29.5%), at night (27.7%), due to lack of sleep (26.6%), or during the day (22.9%). Fewer than one in 20 drivers drove drowsy due to an illness or medical condition (4.6%), taking medications (3.5%) or drinking alcohol (1.8%). Drivers reported that they could drive 5.4 hours, on average, before starting to feel drowsy. One-third of the respondents (32.9%) reported that since they began driving they have continued to drive when they knew they needed rest.
The most common preventive actions taken by drivers to fight drowsiness are to stop driving (72.3%); to change the environment in the car by opening the windows, using the radio, etc. (51.1%); or to consume food or a beverage (32.5%). Thirty-one percent said that they seek to prevent situations in which they are likely to become drowsy by, for example, getting a good night’s sleep, or taking a nap before driving. When asked if they stopped in the last year at a roadside rest area when they felt drowsy, 45.2% of respondents said that they had done so. Two-thirds
Table 2. Reported effect on ability to drive of being drowsy, drinking alcohol, and rain or snow
Being drowsy 2 drinks of wine, beer, or liquor Heavy rain or snow
45.4%
Preventive actions
Note: Data for New York State licensed drivers are from the New York State Department of Motor Vehicles, Division of Research and Evaluation,“Li~nses on File, 1993”.
Driving conditions
54.6% 2.5% 14.1% 38.0%
Great deal 54.7% 29.3% 29.1%
Effect on ability to drive safely Some-what Not much at all 26.2% 33.9% 46.7%
19.0% 36.8% 24.1%
*Of the total sample, 26.6% responded, “I don’t drink alcohol.”
0’) (930) (710)* (965)
A.T. MCCART-Iet al.
514
of these drivers (66.7%) said that stopping helped to combat drowsiness “a great deal.” Of all drivers surveyed, 29.9% needed or wanted to stop at a roadside rest area within the past year when a rest area was not available. 61.8% of all drivers said that they would be very likely to stop at a rest area if they became drowsy. While 42.7% of male drivers would be very likely to stop if they were alone at night, only 16.8% of female drivers would be very likely to do so. Drowsy driving crashes and other incidents In terms of their lifetime driving experiences, 2.8% of the respondents reported that they have fallen asleep at the wheel and had a crash, 1.9% have been drowsy while driving and had a crash, and 22.6% have fallen asleep at the wheel without crashing (Table 5). In all, 24.7% of the respondents reported falling asleep at the wheel at some point in their driving career, including incidents resulting in a crash and incidents in which no crash occurred. The respondents who had had a crash due to
falling asleep or drowsiness or who fell asleep without crashing were asked a series of questions about the circumstances of the most recent incident. The questions addressed presumed characteristics of fall-asleep/drowsy crashes (compared to crashes in general), based on studies of police-reported fallasleep/drowsy crashes, crash investigation studies, and research on the causes and consequences of drowsiness and fatigue (Institute for Traffic Safety Management and Research 1993; Knipling and Wierwille 1994; NYS Task Force on the Impact of Fatigue on Driving 1994; Pack et al. 1995). Selected results are summarized in Table 6; these results should be interpreted with caution, given the relatively small number of drivers involved in fall-asleep/drowsy incidents. The reported fall-asleep/drowsy crashes had the characteristics that have come to be associated with fall-asleep crashes. Of the fall-asleep/drowsy crashes reported by survey respondents, 82.5% involved the driver alone in the vehicle, 60.0% occurred between 11:00 p.m. and 7:00 a.m., 47.5% were drive-off-road crashes, and 40.0% occurred on a highway or expressway. Compared to the fall-asleep/drowsy crashes, the Table 5. Reported
lifetime drowsy
and fall-asleep
Incidents Have Have Have None
fallen had a had a of the
incidents (N = 1000)
asleep at the wheel and not crashed crash when fell asleep at the wheel crash when driving while drowsy above
22.6% 2.8% 1.9% 74.7%
Note: First 3 categories are nor mutually exclusive; 24.7% have fallen asleep, including crash and no-crash incidents.
fall-asleep/no crash incidents were less likely to occur between 11:OOp.m. and 7:00 a.m. or to involve driving off the road, and more likely to occur on a highway or expressway. On average, the drivers in the fallasleep/no crash group had been driving fewer hours and had been awake fewer hours prior to the crash than the drivers in the fall-asleep/drowsy crash group, and were less likely to have been alone in the vehicle. In addition, the drivers in the fall-asleep/no crash group were less likely to have been working a night shift or a lot of overtime the week before the incident and less likely to have consumed alcohol or medication prior to the incident. Correlates offrequency of drowsy driving An examination of the driver correlates of reported drowsy driving was undertaken. The outcome variable was the reported frequency of drowsy driving within the past year: very often, sometimes, rarely, or never. Four sets of hypothesized driver correlates included demographics, and driving, work, and sleep/wake patterns. Table 7 summarizes the bivariate associations between the hypothesized correlates and the outcome variable. The significance of the association was based on the chi-square statistic for the categorical variables, with the dependent variable defined as sometimes/ very often, rarely, or never driving drowsy, and on the correlation coefficient r for the numerical variables. Drivers 25-34 years old, and men more than women, drove drowsy more frequently within the past year. Higher levels of education were associated with increased frequency of drowsy driving. Greater frequency of drowsy driving was associated with greater reported numbers of hours worked per week, working more than one job, working rotating shifts, and driving as part of work responsibilities. Drivers who work shifts including day, evening, and nighttime hours were more likely than other workers to drive drowsy. The survey instrument provided comprehensive measurements of a driver’s sleep patterns as well as assessments of the quality of sleep. Most of these sleep factors were associated with the reported frequency of drowsy driving in the last year. For example, respondents who rated the quality of their sleep as fair or poor were more likely to have sometimes/very often driven drowsy than respondents who rated their sleep quality as good or excellent. The reported frequency of drowsy driving was inversely related to the reported number of hours of sleep each night, and directly related to the difference between the reported number of hours needed each night and the number of hours obtained. In addition,
The scope and nature of the drowsy driving problem in New York State
515
Table 6. Circumstances of reported fall-asleep/drowsy crashes and fall-asleep/no crash incidents Fail-asl~p/drowsy crash* (N = 40)
Circumstances of incident
Fall-asleep/no crasht (N= 152)
60.0%$ 40.0% 47.5% 82.5% 42.5% 35.0% 10.0% 13.5 4.0
11 p.m.-7 a.m. Highway or expressway Drive-off-road Driver alone Worked night sh~t/overt~e Drank alcohol Took medication Mean hours awake pre-incident Mean hours driving pre-incident
36.2%$ 72.4% 26.0% 66.9% 36.4% 1.9% 3.2% 12.0 2.8
*Theoretical statistical sampling error &9.3%-15.5% tLimited to incidents within past 10 years; theoretical statistical sampling error & 4.8967.996 $An additional 12.5% of drivers who crashed and an additional 14.5% of drivers who did not crash indicated that the incident occurred during unspecified evening or nighttime hours.
drivers who frequently have trouble staying awake during the day were more likely than those who do not have this problem to have sometimes/very often driven drowsy. The measurements included in the
study as possible indicators of sleep disorders (snoring, gasping, or cessation of breathing during sleep) were not significantly associated with the reported frequency of drowsy driving.
Table 7. Summary of bivariate associations of driver variables with reported frequency of drowsy driving in last year
Driver variables
N
Frequency of drowsy driving Sometimes or often Rarely Never
Age group
16-24 25-34 35-44 45--54 55-64 65f
1;: 258
Men Women
501 498
166 118 158
Gender Education
%
13.3 42.9 26.2 41.0 17.8 41.1 18.7 38.0 11.9 41.5 7.0 24.1 x2=21.7, pt0.001 40.1 21.0 12.2 35.9
43.9 32.8 41.1 43.4 46.6 68.4 38.9 51.8
x2=29.3, p
Secondary or less Some college Bachelor’s degree Post-bachelor’s Work 2+ jobs
333 324 178 161
Yes No Rotating shifts Yes No Shifts worked Day
126 571
Day + eve/night Day + eve + night Eve and/or night Drive for job Daily Sometimes Rarely Never QuaRty of sleep Excellent Good Fair Poor
% 96 X2=58.3, pt0.001
118 569 349 201 112 34 200 136 112 241 260 481 200 50
14.4 29.1 16.4 42.0 16.9 44.9 21.7 41.6 x2=9.5, p=o.o09 27.0 46.0 18.6 40.6 x2 = 10.2, p = 0.006 30.5 39.0 17.9 42.2 X2= 30.0, p < 0.001 14.6 39.3 23.4 48.3 32.1 36.6 17.6 41.2 x2 = 20.0, p = 0.003 23.0 39.0 44.1 26.5 46.4 23.2 12.9 41.1 xz = 22.6, p = 0.001 36.5 12.7 36.8 15.0 23.5 43.0 36.0 26.0
56.5 41.7 38.2 36.6 27.0 40.8 30.5 39.9 z: 31.3 41.2 38.0 29.4 30.4 46.1 50.8 48.2 33.5 38.0
Driver variables
N
Frequency of drowsy driving Sometimes or often Rarely Never
% Suliicieacy of sleep
Not enough Enough Too much Diiuity
253 482 244
falling asleep
Every night Sometimes Rarely Never
2:: 426 244
Wake during nigh
Every night Sometimes Rarely Never Trouble stay awake day
Daily Sometimes Rarely Never Napping frequency
261 360 273 94 26 184 400 383
Daily Sometimes Rarely Never
65 224 359 348
Daily Sometimes Rarely Hours work/week Hours sleep/night Sleep debt (Sleep Need-Get) Lifetime crashes Miles drive/year Trips/year 3 + hrs Hours drive/day Hrs drive before become drowsy
785 161 43 694 994 980
Driving frequency
96
x2=36.8, p
26.9 41.9 13.3 37.1 12.7 37.3 x2 = 22.9, p =O.OOi 23.1 42.3 23.6 33.6 12.2 43.0 15.2 34.4 x2=22.5, p=O.OOl 15.7 34.5 20.8 35.6 11.4 48.0 19.1 29.8 I(‘= 82.3, p
993 746 975 776
r=0.21, p
785
r= -0.11, p=O.OOI
31.2 49.6 50.0 34.6 42.8 44.8 50.4 49.8 43.6 40.7 51.1 46.2 34.8 27.9 60.8 35.4 46.9 39.6 51.7 39.7 33.5 32.6
A.T. MCCAR~
516
As would be expected, the reported frequency of drowsy driving was positively associated with the reported number of miles driven annually, number of hours driven each day, frequency of driving, and number of trips of three or more hours taken each year. The reported frequency of drowsy driving was also associated with the reported number of lifetime crashes and was inversely related to the number of hours that can be driven before drowsiness occurs. Multiple regression analysis Stepwise multiple regression analysis was used to examine the relative importance of driver characteristics and behaviors in predicting the frequency of driving drowsy in the last year. The dependent variable in each regression model was the reported frequency of driving drowsy in the last year (very often, sometimes, rarely, never). In the first phase of the analysis, a regression model was built for each of the four sets of predictor variables. Then a full model was built, including all variables from each of the four sets of factors. The results of the regression analyses for the four sets of content variables are shown in Table 8. Care was taken in selecting the variable sets to minimize high intercorrelations, or multicollinearity, among the multiple independent predictor variables. In the model of demographic variables, age group, gender, and education level emerged as significant predictors of drowsy driving. Among the work-related variables, working rotating shifts, number of hours worked per week, and frequency of driving for work were significant predictors of the frequency of drowsy driving. From the relatively large set of sleep-related factors
Table 8. Stepwise
Dependent
multiple
variable:
regression set
frequency
analyses
of driving
with each predictor
drowsy
in last year Beta
Model I: Demographic predictors (R’ = O&5)** Age group Gender Education level Model II: Work-related predictors (R’ = 0.036)** Work rotating shifts Number hours work per week Frequency of driving for job Model III: Sleeprelated predictors (R"=0.078)** Frequency of trouble staying awake during day Number of hours sleep per night Model IV: Driving behavior predictors (R'=0.077)** Number of miles drive annually Number of hours drive before become drowsy Number lifetime crashes Number of hours drive per day
*p< 0.05 **p-Co.01
-0.17** -0.14** 0.11* 0.11** 0.10** 0.09* 0.21** -0.16** 0.20** -0.15** 0.11** 0.09**
et al.
Table 9. Stepwise
Dependent
multiple
variable:
regression predictors
frequency
analysis
of driving
drowsy
with
all
driver
in last year Beta
Driver predictors: (R*= 0.177)** Frequency of trouble staying awake during day Number of miles drive annually Age group Number of hours sleep per night Education level Number of hours drive before become drowsy Frequency of driving for job Gender Work rotating shifts
0.19** 0.15** -0.13** -0.11** 0.12** -0.12** 0.08* -0.10* 0.08*
*p < 0.05 **p
DISCUSSION Self-reported driver data, such as those reported here, provide an important means to expand our understanding of the scope and nature of drowsy driving episodes. The results of a telephone survey of a representative sample of New York State licensed drivers suggest that not only do drivers consider drowsy driving to be a serious highway safety concern, but drivers also frequently engage in this risky beha-
The scope and nature
of the drowsy driving
vior. More than half of the drivers reported that they drove drowsy within the last year. The survey results also indicate that about onequarter of the drivers fell asleep at the wheel at some point in their driving career; 2.8% have had a crash when they fell asleep at the wheel, and 1.9% have had a crash due to drowsiness. The characteristics of the fall-asleep/drowsy crashes described by survey respondents are similar to those derived from studies of data for crashes where the driver was reported to have fallen asleep or been drowsy. Although the sample of incidents is relatively small and respondents were asked to recall incidents which sometimes occurred a number of years ago, a comparison of the incidents where the driver crashed with the incidents in which no crash occurred indicate that there may be factors related to the driver’s behavior or the environment that may mitigate the severity of the incident when a driver falls asleep. For example, compared to the fall-asleep/drowsy crashes, the fallasleep/no crash incidents were less likely to involve a driver alone in the vehicle or to occur between 11:00 p.m. and 7:00 a.m., and more likely to occur on a highway or expressway. Knowledge of the types of drivers who may be at high risk for drowsy driving may be helpful in designing effective countermeasures, including public information and education programs. Therefore, in addition to drivers’ attitudes, perceptions, and reported behaviors related to drowsy driving, the survey gathered information on a number of driver variables hypothesized to be related to drowsy driving. These variables included driver demographic characteristics, and reported driving, work, and sleep/wake patterns. Multiple regression analysis was used to examine the relative importance of these variables in explaining the reported frequency of drowsy driving. Although the analyses must be regarded as exploratory, the findings suggest that there are certain driver characteristics that appear to be associated with increased frequency of driving drowsy; these characteristics relate to sleep patterns (greater frequency of trouble staying awake during the day and fewer hours of sleep per night); demographic characteristics (younger drivers, higher level of education, and men); work patterns (greater frequency of driving for job and working rotating shifts); and driving patterns (greater number of miles driven annually and fewer number of hours a person can drive before becoming drowsy). While this set of variables was statistically significant in predicting the frequency of driving drowsy, much remains to be understood about the phenome-
problem
in New York State
517
non of drowsy driving. The relatively modest predictive power of the models suggests that additional research focusing on the relationship between driver and crash variables and fall-asleep incidents, including crashes, is warranted. Acknowledgements-We are grateful to Rosalie F. Lux for assistance in preparation of the manuscript. This study was funded by the New York State Governor’s Traffic Safety Committee.
REFERENCES Carskadon, M. A. Adolescent sleepiness: increased risk in a high-risk population. Alcohol, Drugs and Driving, 5(4)/ 6(1):317-328; October 1989-March 1990. Colquhoun, W. P. Accidents, injuries and shift work. In: Shift Work and Health. DHHS Pub. No. (NIOSH) 76-203. Washington, DC: Government Printing Office; 1976. Folkard, S.; Monk, T. Shiftwork and performance. Hum. Factors 21:483-492; 1989. Horne, J. A.; Reyner, L. A. Sleep related vehicle accidents. Br. Med. J. 310: 565-567; 1995. Institute for Traffic Safety Management and Research. Proceedings: highway safety forum on fatigue, sleep disorders and traffic safety, Albany, New York, December 1, 1993. Knipling, R. R.; Wang, J. S. Crashes and fatalities related to driver drowsiness/fatigue. Research Note, November 1994. Washington, DC: National Highway Traffic Safety Administration. Knipling, R. R.; Wierwille, W. W. Vehicle-based drowsy driver detection: current status and future prospects. Paper presented at the IVHS America 4th Annual Meeting, Atlanta, Georgia, 17-20 April, 1994. Martikainen, K.; Hasan, J.; Urponen, H.; Vuori, I.; Partinen, M. Daytime sleepiness: a risk factor in community life. Acta Neurol. Stand. 86:337-341; 1992. National Transportation Safety Board. Factors that affect fatigue in heavy truck accidents, vols l-2. Safety Study NTSB/SS-95/01. Washington, DC; 1995. National Transportation Safety Board. Fatigue, alcohol, other drugs, and medical factors in fatal-to-the-driver heavy truck crashes. Safety Study NTSB/SS-90/01. Washington, DC, 1990. New South Wales Road Safety Bureau. Problem definition and countermeasure summary: fatigue; 1993. New York State Task Force on the Impact of Fatigue on Driving. Report and recommendations: research team on the nature and scope of drowsy driver crashes in New York State. Albany, New York: Institute for Traffic Safety Management and Research; 1994. Pack, A., Pack, A. M.; Rodgman, E.; Cucchiara, A.; Dinges, D. F.; Schwab, C. W. Characteristics of crashes attributed to the driver having fallen asleep. Accid. Anal. Prev. 27: 769-775; 1995. Richardson, G. S.; Miner, J. D.; Czeisler, C. A. Impaired driving performance in shiftworkers: the role of the circadian system in a multifactorial model. Alcohol, Drugs and Driving 5(4)/6( 1):265-273; October 1989-March 1990.
Pergamon
THE SCOPE AND NATURE OF THE DROWSY DRIVING PROBLEM IN NEW YORK STATE* ANNET. MCCARTT, ‘8 STEPHENA. RIBNER,~ALLAN I. PACKSand MARK C. HAMMERI ‘Institute for Traffic Safety Management and Research, University at Albany, State University of New York, 80 Wolf Road, Suite 607, Albany, NY 12205-2604, USA, ‘Fact Finders, Incorporated, 2010 Western Avenue, Albany, NY 12203-5015, USA; 3Center for Sleep and Respiratory Neurobiology, University of Pennsylvania, 3600 Spruce Street, Philadelphia, PA 19104-4283, USA (Accepted 27 September 1995)
Abstract-Atelephone survey was conducted of a random sample of New York State licensed drivers to determine the prevalence and circumstances of drowsy driving. Based on the survey responses, 54.6% of the drivers had driven while drowsy within the past year; 22.6% had ever fallen asleep at the wheel without having a crash, 2.8% had ever crashed when they fell asleep, and 1.9% had crashed when driving while drowsy. Of the reported crashes due to driving while drowsy or falling asleep at the wheel, 82.5% involved the driver alone in the vehicle, 60.0% occurred between 11:OOp.m. and 7:OOa.m., 47.5% were drive-off-road crashes, and 40.0% occurred on a highway or expressway. Multiple regression analysis suggested that the following driver variables are predictive of an increased frequency of driving drowsy: demographic characteristics (younger drivers, more education, and men); sleep patterns (fewer hours of sleep at night and greater frequency of trouble staying awake during the day); work patterns (greater frequency of driving for job and working rotating shifts); and driving patterns (greater number of miles driven annually and fewer number of hours a person can drive before becoming drowsy). Copyright 0 1996 Elsevier Science Ltd Keywords-Drowsy
driving, Fatigued driving, Fall-asleep crashes, Drowsy driver crashes, Driver survey
Wales Road Safety Bureau 1993). Further, apart from the compilation of crash data, scant research has been conducted on the number and characteristics of crashes that involve driver drowsiness or falling asleep or on the prevalence of drowsy driving that does not result in a crash. Data for 1989-1993 from the U.S. National Highway Traffic Safety Administration’s General Estimates System (GES) indicate that nationwide there were an estimated 56,000 crashes annually in which driver drowsiness/fatigue was indicated on the police accident report; this represents about 1% of all police-reported crashes. Data from the agency’s Fatal Accident Reporting System (FARS) indicate that either drowsiness or fatigue was cited as a factor in an annual average of 1,357 fatal crashes during 1989-1993, representing 3.6% of all fatal crashes during that time period (Knipling and Wang 1994). The problem of drowsy driving is believed to be especially severe among certain groups of drivers. For example, the National Transportation Safety Board (NTSB 1990) estimates that nearly one-third of fatalto-the-driver heavy truck crashes involve fatigue, and
INTRODUCTION A review of current research on the nature and scope of drowsy driver crashes indicates that there are large gaps in our knowledge about this problem. To date, the understanding of the prevalence and circumstances of drowsy driver crashes in the United States has been based largely on crash data gathered by government agencies. However, it is believed by many researchers that the prevalence of driver drowsiness in crashes is under-reported, since there is generally little evidence upon which to base a finding that drowsiness was a factor in a crash, and since potential liability may motivate a driver to fail to report his or her drowsiness or fatigue. Surveys of motorists conducted in England, Finland, and New South Wales suggest that drowsy driving may be more prevalent than the crash data would indicate (Horne and Reyner 1995; Martikainen et al. 1992; New South *Presented at the 39th Annual Meeting the Advancement of Automotive Medicine: Chicago, Illinois. tAuthor for correspondence.
of the Association for 16-18 October 1995.
511
A.T. MCCARTT et al.
512
that fatigue and fatigue/drug interactions are implicated in a greater number of fatal heavy truck crashes than alcohol and other drugs alone. Recently, the NTSB (1995) conducted a study of nighttime single vehicle heavy truck crashes in which the driver survived; the study found that 58% of these crashes were fatigue-related. Several studies have associated shiftwork, especially rotating shifts, with an increased risk for crashes due to sleepiness or fatigue (Colquhoun 1976; Folkard and Monk 1989; Richardson et al. 1989/1990). Based on their greater need for sleep and their lifestyles, young drivers would also appear to be at high risk for drowsy driving, and crash studies and surveys of youth support this hypothesis (Carskadon 1989/1990; Knipling and Wang 1994; Pack et al. 1995). The limited state of knowledge about the problem of drowsy driving may be contrasted with the level of knowledge of alcohol-related crashes. Although alcohol-related crashes are also known to be under-reported, there are physical indicators of intoxication and reliable tests to determine the blood alcohol concentration of a driver. Based on the availability of these tests and the cumulative results of research on the characteristics of alcohol-related crashes, a good understanding of the scope and characteristics of alcohol-related crashes has been achieved, despite the under-reporting of such crashes. Since there are no reliable measures to indicate that a driver fell asleep or was fatigued, it is important to use alternative research methods in addition to analyses of crash data to study the problem. To document more fully the prevalence and nature of drowsy driving in New York State, a statewide telephone survey was designed and conducted by Fact Finders, Inc., and the Institute for Traffic Safety Management and Research, under the auspices of the New York State Task Force on the Impact of Fatigue on Driving. In addition to information on driver attitudes and reported behaviors, the survey gathered information on a number of factors hypothesized to be associated with drowsy driving. It was hoped that this information would assist in the development of a better understanding of the characteristics of drivers who may be at higher risk for drowsy driving and the types of driving situations that may contribute to drowsy driving. METHODS In order to accomplish the reliable and representative measurement of attitudes, perceptions, and behaviors related to drowsy driving, a scientific random probability sample was designed. The sample was comprised of 1000 randomly selected licensed
drivers (or those with a learner’s permit) residing in New York’s 62 counties. The number of interviews designated for each county was determined by that county’s population of licensed drivers relative to the total population of licensed drivers in the state. Telephone numbers were generated by randomdigit dialing, a sampling technique which assures an equal, unbiased probability of inclusion in the sample for all households with a telephone, including households with unlisted and newly established telephone numbers. In theory, the statistical sampling error associated with the overall findings ranges from *1.8% to 3.1%. Seventeen percent of the drivers contacted declined to participate in the survey. The survey questionnaire was developed by the Institute for Traffic Safety Management and Research and members of the Research Team of the Task Force on the Impact of Fatigue on Driving, and subsequently reviewed and pilot-tested by Fact Finders. The initial draft incorporated portions of a survey instrument developed at the University of Pennsylvania Center for Sleep and Respiratory Neurobiology. Draft versions of the instrument were pilot-tested to insure that the final questionnaire would be efficient, comprehensive, valid, and free from contextual biases. Approximately 30 pilot tests were conducted. This paper describes the results of the telephone survey, with particular emphasis on motorists’ experiences with drowsy driving, including crashes due to falling asleep or drowsiness, and on driver correlates associated with the reported frequency of drowsy driving episodes. Following a summary of the characteristics of the survey respondents, the tabulated results are presented for selected survey items related to the awareness of drowsy driving, drowsy driving experiences in the past year, preventive actions, and lifetime drowsy driving incidents. Next, an examination of driver correlates with the frequency of drowsy driving experiences is presented. The correlates include driver demographic characteristics, and driving, work, and sleep and wake patterns. The examination consists of bivariate analyses of the association between the hypothesized correlates and the frequency of reported drowsy driving within the past year, and multiple regression models with reported frequency of drowsy driving as the dependent variable and driver demographic characteristics and work, driving, and sleep patterns as the independent variables. RESULTS Summary of sample demographics The gender, age group, and region of residence of the survey sample and the population of New York
The scope and nature of the drowsy driving problem in New York State
State licensed drivers are provided in Table 1. The sample was highly representative of the population of licensed drivers. Awareness of drowsy driving issue
With “drowsy” defined as “so tired you could easily fall asleep,” 54.7% of the survey respondents reported that being drowsy greatly affects their ability to drive safely. Survey respondents indicated that being drowsy has a greater effect on their ability to drive safely than either adverse weather or having two drinks of wine, beer, or liquor (Table 2). More than one-quarter of the respondents (26.1%) knew someone who had a crash due to falling asleep at the wheel or drowsiness. Driving while drowsy experiences
The survey results indicated that 54.6% of drivers experienced driving while drowsy in the last year, and 2.5% drove drowsy “very often” (Table 3). All respondents were presented with a series of different situations and asked whether they had driven while drowsy in the last year in each of these situations. The results are presented in Table 4. Of the total survey sample, about one in four drivers reported that they drove drowsy during the past year in each Table 1. Demographics of survey sample and population of New York State licensed drivers
Gender Men Women Age Group 16-24 years old 25-34 years old 35-44 years old 45-54 years old 55-64 years old 65 years or over Region Upstate Long Island New York City
Survey sample (N=1000)
Total NYS drivers (N= 10.3M)
so.19 49.9%
52.8% 47.2%
9.9% 19.6% 26.1% 16.7% 11.9% 15.9%
12.8% 22.9% 22.3% 16.7% 11.5% 13.8%
53.8% 19.0% 27.2%
53.6% 18.8% 21.6%
513
Table 3. Reported frequency of driving while drowsy in last year Frequency of drowsy driving
(N=999)
Have driven drowsy in last year Very often Sometimes Rarely Have not driven drowsy in last year
Table 4. Situations in which drivers experienced drowsiness in Iast year Drowsy driving situations
(N= 1000)
On a long trip At night Due to lack of sleep During the day Due to an illness or medical condition Due to taking medications Due to drinking alcohol
29.5% 27.7% 26.6% 22.9% 4.6% 3.5% 1.8%
Note: All respondents were asked if they had driven while drowsy in the last year in each situation; some respondents had driven drowsy in more than one situation.
of these situations: on a long trip (29.5%), at night (27.7%), due to lack of sleep (26.6%), or during the day (22.9%). Fewer than one in 20 drivers drove drowsy due to an illness or medical condition (4.6%), taking medications (3.5%) or drinking alcohol (1.8%). Drivers reported that they could drive 5.4 hours, on average, before starting to feel drowsy. One-third of the respondents (32.9%) reported that since they began driving they have continued to drive when they knew they needed rest.
The most common preventive actions taken by drivers to fight drowsiness are to stop driving (72.3%); to change the environment in the car by opening the windows, using the radio, etc. (51.1%); or to consume food or a beverage (32.5%). Thirty-one percent said that they seek to prevent situations in which they are likely to become drowsy by, for example, getting a good night’s sleep, or taking a nap before driving. When asked if they stopped in the last year at a roadside rest area when they felt drowsy, 45.2% of respondents said that they had done so. Two-thirds
Table 2. Reported effect on ability to drive of being drowsy, drinking alcohol, and rain or snow
Being drowsy 2 drinks of wine, beer, or liquor Heavy rain or snow
45.4%
Preventive actions
Note: Data for New York State licensed drivers are from the New York State Department of Motor Vehicles, Division of Research and Evaluation,“Li~nses on File, 1993”.
Driving conditions
54.6% 2.5% 14.1% 38.0%
Great deal 54.7% 29.3% 29.1%
Effect on ability to drive safely Some-what Not much at all 26.2% 33.9% 46.7%
19.0% 36.8% 24.1%
*Of the total sample, 26.6% responded, “I don’t drink alcohol.”
0’) (930) (710)* (965)
A.T. MCCART-Iet al.
514
of these drivers (66.7%) said that stopping helped to combat drowsiness “a great deal.” Of all drivers surveyed, 29.9% needed or wanted to stop at a roadside rest area within the past year when a rest area was not available. 61.8% of all drivers said that they would be very likely to stop at a rest area if they became drowsy. While 42.7% of male drivers would be very likely to stop if they were alone at night, only 16.8% of female drivers would be very likely to do so. Drowsy driving crashes and other incidents In terms of their lifetime driving experiences, 2.8% of the respondents reported that they have fallen asleep at the wheel and had a crash, 1.9% have been drowsy while driving and had a crash, and 22.6% have fallen asleep at the wheel without crashing (Table 5). In all, 24.7% of the respondents reported falling asleep at the wheel at some point in their driving career, including incidents resulting in a crash and incidents in which no crash occurred. The respondents who had had a crash due to
falling asleep or drowsiness or who fell asleep without crashing were asked a series of questions about the circumstances of the most recent incident. The questions addressed presumed characteristics of fall-asleep/drowsy crashes (compared to crashes in general), based on studies of police-reported fallasleep/drowsy crashes, crash investigation studies, and research on the causes and consequences of drowsiness and fatigue (Institute for Traffic Safety Management and Research 1993; Knipling and Wierwille 1994; NYS Task Force on the Impact of Fatigue on Driving 1994; Pack et al. 1995). Selected results are summarized in Table 6; these results should be interpreted with caution, given the relatively small number of drivers involved in fall-asleep/drowsy incidents. The reported fall-asleep/drowsy crashes had the characteristics that have come to be associated with fall-asleep crashes. Of the fall-asleep/drowsy crashes reported by survey respondents, 82.5% involved the driver alone in the vehicle, 60.0% occurred between 11:00 p.m. and 7:00 a.m., 47.5% were drive-off-road crashes, and 40.0% occurred on a highway or expressway. Compared to the fall-asleep/drowsy crashes, the Table 5. Reported
lifetime drowsy
and fall-asleep
Incidents Have Have Have None
fallen had a had a of the
incidents (N = 1000)
asleep at the wheel and not crashed crash when fell asleep at the wheel crash when driving while drowsy above
22.6% 2.8% 1.9% 74.7%
Note: First 3 categories are nor mutually exclusive; 24.7% have fallen asleep, including crash and no-crash incidents.
fall-asleep/no crash incidents were less likely to occur between 11:OOp.m. and 7:00 a.m. or to involve driving off the road, and more likely to occur on a highway or expressway. On average, the drivers in the fallasleep/no crash group had been driving fewer hours and had been awake fewer hours prior to the crash than the drivers in the fall-asleep/drowsy crash group, and were less likely to have been alone in the vehicle. In addition, the drivers in the fall-asleep/no crash group were less likely to have been working a night shift or a lot of overtime the week before the incident and less likely to have consumed alcohol or medication prior to the incident. Correlates offrequency of drowsy driving An examination of the driver correlates of reported drowsy driving was undertaken. The outcome variable was the reported frequency of drowsy driving within the past year: very often, sometimes, rarely, or never. Four sets of hypothesized driver correlates included demographics, and driving, work, and sleep/wake patterns. Table 7 summarizes the bivariate associations between the hypothesized correlates and the outcome variable. The significance of the association was based on the chi-square statistic for the categorical variables, with the dependent variable defined as sometimes/ very often, rarely, or never driving drowsy, and on the correlation coefficient r for the numerical variables. Drivers 25-34 years old, and men more than women, drove drowsy more frequently within the past year. Higher levels of education were associated with increased frequency of drowsy driving. Greater frequency of drowsy driving was associated with greater reported numbers of hours worked per week, working more than one job, working rotating shifts, and driving as part of work responsibilities. Drivers who work shifts including day, evening, and nighttime hours were more likely than other workers to drive drowsy. The survey instrument provided comprehensive measurements of a driver’s sleep patterns as well as assessments of the quality of sleep. Most of these sleep factors were associated with the reported frequency of drowsy driving in the last year. For example, respondents who rated the quality of their sleep as fair or poor were more likely to have sometimes/very often driven drowsy than respondents who rated their sleep quality as good or excellent. The reported frequency of drowsy driving was inversely related to the reported number of hours of sleep each night, and directly related to the difference between the reported number of hours needed each night and the number of hours obtained. In addition,
The scope and nature of the drowsy driving problem in New York State
515
Table 6. Circumstances of reported fall-asleep/drowsy crashes and fall-asleep/no crash incidents Fail-asl~p/drowsy crash* (N = 40)
Circumstances of incident
Fall-asleep/no crasht (N= 152)
60.0%$ 40.0% 47.5% 82.5% 42.5% 35.0% 10.0% 13.5 4.0
11 p.m.-7 a.m. Highway or expressway Drive-off-road Driver alone Worked night sh~t/overt~e Drank alcohol Took medication Mean hours awake pre-incident Mean hours driving pre-incident
36.2%$ 72.4% 26.0% 66.9% 36.4% 1.9% 3.2% 12.0 2.8
*Theoretical statistical sampling error &9.3%-15.5% tLimited to incidents within past 10 years; theoretical statistical sampling error & 4.8967.996 $An additional 12.5% of drivers who crashed and an additional 14.5% of drivers who did not crash indicated that the incident occurred during unspecified evening or nighttime hours.
drivers who frequently have trouble staying awake during the day were more likely than those who do not have this problem to have sometimes/very often driven drowsy. The measurements included in the
study as possible indicators of sleep disorders (snoring, gasping, or cessation of breathing during sleep) were not significantly associated with the reported frequency of drowsy driving.
Table 7. Summary of bivariate associations of driver variables with reported frequency of drowsy driving in last year
Driver variables
N
Frequency of drowsy driving Sometimes or often Rarely Never
Age group
16-24 25-34 35-44 45--54 55-64 65f
1;: 258
Men Women
501 498
166 118 158
Gender Education
%
13.3 42.9 26.2 41.0 17.8 41.1 18.7 38.0 11.9 41.5 7.0 24.1 x2=21.7, pt0.001 40.1 21.0 12.2 35.9
43.9 32.8 41.1 43.4 46.6 68.4 38.9 51.8
x2=29.3, p
Secondary or less Some college Bachelor’s degree Post-bachelor’s Work 2+ jobs
333 324 178 161
Yes No Rotating shifts Yes No Shifts worked Day
126 571
Day + eve/night Day + eve + night Eve and/or night Drive for job Daily Sometimes Rarely Never QuaRty of sleep Excellent Good Fair Poor
% 96 X2=58.3, pt0.001
118 569 349 201 112 34 200 136 112 241 260 481 200 50
14.4 29.1 16.4 42.0 16.9 44.9 21.7 41.6 x2=9.5, p=o.o09 27.0 46.0 18.6 40.6 x2 = 10.2, p = 0.006 30.5 39.0 17.9 42.2 X2= 30.0, p < 0.001 14.6 39.3 23.4 48.3 32.1 36.6 17.6 41.2 x2 = 20.0, p = 0.003 23.0 39.0 44.1 26.5 46.4 23.2 12.9 41.1 xz = 22.6, p = 0.001 36.5 12.7 36.8 15.0 23.5 43.0 36.0 26.0
56.5 41.7 38.2 36.6 27.0 40.8 30.5 39.9 z: 31.3 41.2 38.0 29.4 30.4 46.1 50.8 48.2 33.5 38.0
Driver variables
N
Frequency of drowsy driving Sometimes or often Rarely Never
% Suliicieacy of sleep
Not enough Enough Too much Diiuity
253 482 244
falling asleep
Every night Sometimes Rarely Never
2:: 426 244
Wake during nigh
Every night Sometimes Rarely Never Trouble stay awake day
Daily Sometimes Rarely Never Napping frequency
261 360 273 94 26 184 400 383
Daily Sometimes Rarely Never
65 224 359 348
Daily Sometimes Rarely Hours work/week Hours sleep/night Sleep debt (Sleep Need-Get) Lifetime crashes Miles drive/year Trips/year 3 + hrs Hours drive/day Hrs drive before become drowsy
785 161 43 694 994 980
Driving frequency
96
x2=36.8, p
26.9 41.9 13.3 37.1 12.7 37.3 x2 = 22.9, p =O.OOi 23.1 42.3 23.6 33.6 12.2 43.0 15.2 34.4 x2=22.5, p=O.OOl 15.7 34.5 20.8 35.6 11.4 48.0 19.1 29.8 I(‘= 82.3, p
993 746 975 776
r=0.21, p
785
r= -0.11, p=O.OOI
31.2 49.6 50.0 34.6 42.8 44.8 50.4 49.8 43.6 40.7 51.1 46.2 34.8 27.9 60.8 35.4 46.9 39.6 51.7 39.7 33.5 32.6
A.T. MCCAR~
516
As would be expected, the reported frequency of drowsy driving was positively associated with the reported number of miles driven annually, number of hours driven each day, frequency of driving, and number of trips of three or more hours taken each year. The reported frequency of drowsy driving was also associated with the reported number of lifetime crashes and was inversely related to the number of hours that can be driven before drowsiness occurs. Multiple regression analysis Stepwise multiple regression analysis was used to examine the relative importance of driver characteristics and behaviors in predicting the frequency of driving drowsy in the last year. The dependent variable in each regression model was the reported frequency of driving drowsy in the last year (very often, sometimes, rarely, never). In the first phase of the analysis, a regression model was built for each of the four sets of predictor variables. Then a full model was built, including all variables from each of the four sets of factors. The results of the regression analyses for the four sets of content variables are shown in Table 8. Care was taken in selecting the variable sets to minimize high intercorrelations, or multicollinearity, among the multiple independent predictor variables. In the model of demographic variables, age group, gender, and education level emerged as significant predictors of drowsy driving. Among the work-related variables, working rotating shifts, number of hours worked per week, and frequency of driving for work were significant predictors of the frequency of drowsy driving. From the relatively large set of sleep-related factors
Table 8. Stepwise
Dependent
multiple
variable:
regression set
frequency
analyses
of driving
with each predictor
drowsy
in last year Beta
Model I: Demographic predictors (R’ = O&5)** Age group Gender Education level Model II: Work-related predictors (R’ = 0.036)** Work rotating shifts Number hours work per week Frequency of driving for job Model III: Sleeprelated predictors (R"=0.078)** Frequency of trouble staying awake during day Number of hours sleep per night Model IV: Driving behavior predictors (R'=0.077)** Number of miles drive annually Number of hours drive before become drowsy Number lifetime crashes Number of hours drive per day
*p< 0.05 **p-Co.01
-0.17** -0.14** 0.11* 0.11** 0.10** 0.09* 0.21** -0.16** 0.20** -0.15** 0.11** 0.09**
et al.
Table 9. Stepwise
Dependent
multiple
variable:
regression predictors
frequency
analysis
of driving
drowsy
with
all
driver
in last year Beta
Driver predictors: (R*= 0.177)** Frequency of trouble staying awake during day Number of miles drive annually Age group Number of hours sleep per night Education level Number of hours drive before become drowsy Frequency of driving for job Gender Work rotating shifts
0.19** 0.15** -0.13** -0.11** 0.12** -0.12** 0.08* -0.10* 0.08*
*p < 0.05 **p
DISCUSSION Self-reported driver data, such as those reported here, provide an important means to expand our understanding of the scope and nature of drowsy driving episodes. The results of a telephone survey of a representative sample of New York State licensed drivers suggest that not only do drivers consider drowsy driving to be a serious highway safety concern, but drivers also frequently engage in this risky beha-
The scope and nature
of the drowsy driving
vior. More than half of the drivers reported that they drove drowsy within the last year. The survey results also indicate that about onequarter of the drivers fell asleep at the wheel at some point in their driving career; 2.8% have had a crash when they fell asleep at the wheel, and 1.9% have had a crash due to drowsiness. The characteristics of the fall-asleep/drowsy crashes described by survey respondents are similar to those derived from studies of data for crashes where the driver was reported to have fallen asleep or been drowsy. Although the sample of incidents is relatively small and respondents were asked to recall incidents which sometimes occurred a number of years ago, a comparison of the incidents where the driver crashed with the incidents in which no crash occurred indicate that there may be factors related to the driver’s behavior or the environment that may mitigate the severity of the incident when a driver falls asleep. For example, compared to the fall-asleep/drowsy crashes, the fallasleep/no crash incidents were less likely to involve a driver alone in the vehicle or to occur between 11:00 p.m. and 7:00 a.m., and more likely to occur on a highway or expressway. Knowledge of the types of drivers who may be at high risk for drowsy driving may be helpful in designing effective countermeasures, including public information and education programs. Therefore, in addition to drivers’ attitudes, perceptions, and reported behaviors related to drowsy driving, the survey gathered information on a number of driver variables hypothesized to be related to drowsy driving. These variables included driver demographic characteristics, and reported driving, work, and sleep/wake patterns. Multiple regression analysis was used to examine the relative importance of these variables in explaining the reported frequency of drowsy driving. Although the analyses must be regarded as exploratory, the findings suggest that there are certain driver characteristics that appear to be associated with increased frequency of driving drowsy; these characteristics relate to sleep patterns (greater frequency of trouble staying awake during the day and fewer hours of sleep per night); demographic characteristics (younger drivers, higher level of education, and men); work patterns (greater frequency of driving for job and working rotating shifts); and driving patterns (greater number of miles driven annually and fewer number of hours a person can drive before becoming drowsy). While this set of variables was statistically significant in predicting the frequency of driving drowsy, much remains to be understood about the phenome-
problem
in New York State
517
non of drowsy driving. The relatively modest predictive power of the models suggests that additional research focusing on the relationship between driver and crash variables and fall-asleep incidents, including crashes, is warranted. Acknowledgements-We are grateful to Rosalie F. Lux for assistance in preparation of the manuscript. This study was funded by the New York State Governor’s Traffic Safety Committee.
REFERENCES Carskadon, M. A. Adolescent sleepiness: increased risk in a high-risk population. Alcohol, Drugs and Driving, 5(4)/ 6(1):317-328; October 1989-March 1990. Colquhoun, W. P. Accidents, injuries and shift work. In: Shift Work and Health. DHHS Pub. No. (NIOSH) 76-203. Washington, DC: Government Printing Office; 1976. Folkard, S.; Monk, T. Shiftwork and performance. Hum. Factors 21:483-492; 1989. Horne, J. A.; Reyner, L. A. Sleep related vehicle accidents. Br. Med. J. 310: 565-567; 1995. Institute for Traffic Safety Management and Research. Proceedings: highway safety forum on fatigue, sleep disorders and traffic safety, Albany, New York, December 1, 1993. Knipling, R. R.; Wang, J. S. Crashes and fatalities related to driver drowsiness/fatigue. Research Note, November 1994. Washington, DC: National Highway Traffic Safety Administration. Knipling, R. R.; Wierwille, W. W. Vehicle-based drowsy driver detection: current status and future prospects. Paper presented at the IVHS America 4th Annual Meeting, Atlanta, Georgia, 17-20 April, 1994. Martikainen, K.; Hasan, J.; Urponen, H.; Vuori, I.; Partinen, M. Daytime sleepiness: a risk factor in community life. Acta Neurol. Stand. 86:337-341; 1992. National Transportation Safety Board. Factors that affect fatigue in heavy truck accidents, vols l-2. Safety Study NTSB/SS-95/01. Washington, DC; 1995. National Transportation Safety Board. Fatigue, alcohol, other drugs, and medical factors in fatal-to-the-driver heavy truck crashes. Safety Study NTSB/SS-90/01. Washington, DC, 1990. New South Wales Road Safety Bureau. Problem definition and countermeasure summary: fatigue; 1993. New York State Task Force on the Impact of Fatigue on Driving. Report and recommendations: research team on the nature and scope of drowsy driver crashes in New York State. Albany, New York: Institute for Traffic Safety Management and Research; 1994. Pack, A., Pack, A. M.; Rodgman, E.; Cucchiara, A.; Dinges, D. F.; Schwab, C. W. Characteristics of crashes attributed to the driver having fallen asleep. Accid. Anal. Prev. 27: 769-775; 1995. Richardson, G. S.; Miner, J. D.; Czeisler, C. A. Impaired driving performance in shiftworkers: the role of the circadian system in a multifactorial model. Alcohol, Drugs and Driving 5(4)/6( 1):265-273; October 1989-March 1990.