Accident Analysis and Prevention 33 (2001) 187196

Fatal traffic accidents among trailer truck drivers and accidentcauses as viewed by other truck drivers

Helina Hakkanen *, Heikki SummalaDepartment of Psychology, Traffic Research Unit, Uni6ersity of Helsinki, PO Box 13, 00014 Helsinki, Finland

Received 28 October 1998; received in revised form 24 January 2000; accepted 8 February 2000

Abstract

Causality factors, the responsibility of the driver and driver fatigue-related factors were studied in fatal two-vehicle accidentswhere a trailer truck driver was involved during the period of 19911997 (n337). In addition, 251 long-haul truck drivers weresurveyed in order to study their views regarding contributing factors in accidents involving trucks and the development of possiblecountermeasure against driver fatigue. Trailer truck drivers were principally responsible for 16% of all the accidents. Youngerdriver age and driving during evening hours were significant predictors of being principally responsible. In addition, theprobability of being principally responsible for the accident increased by a factor of over three if the driver had a chronic illness.Prolonged driving preceding the accident, accident history or traffic offence history did not have a significant effect. Only 2% ofthe drivers were estimated to have fallen asleep while driving just prior to the accident, and altogether 4% of the drivers had beentired prior to the accident. Of the drivers 13% had however, been driving over 10 h preceding the accident (which has beencriminally punishably in Finland since 1995 under the EC regulation) but no individual factors had a significant effect inpredicting prolonged driving. The surveyed views regarding causes of truck accidents correspond well with the accident analysis.Accidents were viewed as being most often caused by other road users and driver fatigue was viewed to be no more than the fifth(out of eight) common cause of accidents. The probability of viewing fatigue as a more common cause increased significantly ifthe driver had experienced fatigue-related problems while driving. However, nearly half of the surveyed truck drivers expresseda negative view towards developing a technological countermeasure against driver fatigue. The negative view was not related topersonal experiences of fatigue-related problems while driving. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: In-depth accident analysis; Driver fatigue; Chronic illness; Fatal road accidents; Truck drivers; Technological countermeasure

www.elsevier.com:locate:aap

1. Introduction

Throughout the 1990s in Finland trailer truck drivershave been involved in approximately 8% of all the fatalroad accidents and in 2% of the accidents leading to apersonal injury. Due to the weight of the truck and thedifference in weight between the truck and the passen-ger car, that is often the other participant, the probabil-ity of fatality in an accident involving a truck issignificantly increased compared to accidents not in-volving trucks (Eicher et al., 1982; Evans, 1991).

An issue that has received considerable attention inthe truck industry is the association between the proba-

bility of being involved in an accident and driver fa-tigue. Generally truck drivers themselves do believedriver fatigue to be a problem for the industry andseveral previous studies have reported that some truckdrivers do have fatigue-related problems while driving(Feyer and Williamson, 1995; Arnold et al., 1997;Sluiter et al., 1999; Hakkanen and Summala, 2000a).However, when analysing accident data, the informa-tion concerning drivers possible fatigue is not alwaysavailable or might, in some respects, be unreliable.Nevertheless, it has been recently estimated that truckdrivers have been involved in approximately 3% offatigue-related accidents in the US, with driver fatigue-related accidents contributing one to 3% of all roadaccidents (Lyznicki et al., 1998).

Although a large proportion of truck drivers havebeen found to smoke regularly, be overweight and

* Corresponding author. Tel.:358-9-19123731; fax: 358-9-19123489.

E-mail addresses: hehakkanen@hotmail.com (H. Hakkanen),heidi.summala@helsinki.fi (H. Summala).

0001-4575:01:$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.PII: S00 0 1 -4575 (00 )00030 -0

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196188

unaware of their high blood pressure (Korelitz et al.,1993), surprisingly few previous studies have studiedthe association between truck drivers medical conditionand the probability of being involved in an accident.However, Dionne et al. (1995) showed that diabetictruck drivers (but not drivers with coronary heart dis-ease, hypertension or problems in binocular vision) hadin fact more accidents than drivers in good health.Another study showed a relationship between pro-longed driving, poor health status of the driver andmore frequent driver fatigue-related problems (Hak-kanen and Summala, 2000b).

Previous analysis concerning fatal accident rates oflarge trucks have more often shown that the probabilityof being involved in an accident is associated with somedriver characteristics or external factors. As with pri-vate drivers, the truck drivers probability of beinginvolved in an accident has been shown to be associatedwith driver age and driving experience. Campbell (1991)found that truck drivers under 27 years were over-in-volved in accidents, independently of time of day.Kaneko and Jovanis (1992), on the other hand, showedthat accident risk of truck drivers is highest amongthose with less than 5 years of driver experience. In ourprevious study, concerning fatal accident rates of trailertruckers during 19841989, we showed that driverswhose fatigue contributed to the accident were in factyounger that other drivers (Summala and Mikkola,1994).

Based on previous findings that indicate increasedaccident risk in truck drivers after prolonged driving(Hamelin, 1981; McDonald, 1984; Kaneko and Jovanis,1992; Summala and Mikkola, 1994), two approaches, interms of decreasing the accident risk, have been espe-cially stressed. Firstly, in countries belonging to theEuropean Union the truck driving hours are subject tocontrol under the EC Regulation No. 3820:85, accord-ing to which the maximum driving time is set to 10 hand the resting time to at least 11 h per 24 h period. InUnited States, federal regulation also limits permitteddriving hours to 10 h while the minimum resting time is8 h. Several previous studies have, however shown thatnon-compliance with the driving hour regulation israther common among truck drivers (Hertz, 1991;Braver et al., 1992). According to Braver et al. (1992),who found nearly three-fourths of the surveyed trailertruck drivers to violate the regulation, the primaryreasons for violation were economic. Nevertheless, it isalso possible that drivers generally believe themselvesable to cope with possible driver fatigue and rarelyconsider it leading to an accident.

Another approach in trying to decrease the accidentrisk associated with prolonged driving is that of devel-oping technological, in-car countermeasures to detectdriver fatigue. However, there still are certain limita-tions regarding this approach, first being that it is

unclear what measure (such as eye closure, deteriora-tion in steering skill, etc.) or what combination ofmeasures would be sufficiently valid and accessible indetecting driver fatigue. Altogether, it has been esti-mated that the research and development of a reliabletechnological countermeasure and the market penetra-tion will take up to 20 years (Brown, 1995, 1997). Inaddition, it has been highlighted that technologicalcountermeasures device may not prevent those driverswho are highly motivated to complete a journey, fromcontinuing to drive after the device alarms and that insome cases such a device could be used to support thecontinuation of journeys, despite increased subjectivefatigue (Summala and Mikkola, 1994; Brown, 1997;Summala et al., 1999). Furthermore, we are not awareof any study where truck drivers views on possibletechnological countermeasures have been addressed.

The purpose of the present study was, by analysingaccident data, to investigate contributing factors tofatal accidents involving trailer trucks and to identifyindividual and external factors that would predict driv-ers who had been principally responsible for the acci-dent. Specific emphasis was given to the medicalcondition of the drivers, driver fatigue and prolongeddriving. The frequency of falling asleep while driving asa contributor to the accidents and the effect of pro-longed driving (over 10 h) were estimated. In addition,a group of long-haul truck drivers were surveyed inorder to study their views regarding the contributingfactors to accidents involving trucks, including driverfatigue, and how they view the development of in-cartechnological countermeasure against driver fatigue.The issues of the present study are considered to givemore information in order to improve the possibleeducational and technological countermeasures relatedto traffic safety in truck drivers.

2. Method

2.1. Fatal accident data

This study used a database consisting of case studyreports of all fatal two-vehicle accidents where trailertruck drivers had been involved during the period of19911997. In Finland all the fatal accidents have been,since the early 1970s, studied in depth by the expertisemembers of local multidisciplinary teams consisting ofa police officer, road engineer, vehicle engineer, physi-cian and occasionally a psychologist. The study in-cludes on-the-spot investigation made immediately afterthe accident, interviews with survivors, eyewitnessesand relatives of the deceased victims along with thejoint meetings of the members (for details on themethod and the database, and for reviews of earlierstudies made from it, see Traffic Safety Committee ofInsurance Companies, 1987; Hantula, 1989, 1992).

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196 189

Detailed information concerning the persons whohave been involved in the accident is further gatheredfrom national traffic offence record and health carecentres, for example concerning previous offences andmedical history of the driver. Altogether the case studydatabase consists of over 300 different variables and themultidisciplinary teams also tend to identify a factorwhich triggered the accident process (causal factor, forexample failure to perceive the other participant). Thecausal factor indicates the very last preceding factorthat was essential in the occurrence of the accident andthat transferred the normal traffic situation into anaccident as well as made the accident irrevocable. Quiteoften the actual accident is staged in order to estimatethe factors contributing to the accident.

Driver fatigue is primarily investigated and assessedby the multidisciplinary teams in terms of interviewingthe survived driver and possible other participants. Thismay however, lead to some bias in the results as not alldrivers are willing to admit that they felt sleepy ordozed off prior to the accident. In case of a perisheddriver, the estimation of driver fatigue is based onanalysing the driving time, prior sleeping time, time ofday, etc. As both of these investigation methods havesome shortcomings the results regarding the frequencyof fatigue-related accidents are to be considered as onlyestimates.

The data for 19911997 consisted of 337 fatal two-vehicle accidents where a trailer truck driver was in-volved during his work time. All of the drivers weremale. According to the investigation performed by themultidisciplinary teams the trailer truck driver wasconsidered to be the main originator of the accident(being principally responsible) in 57 (17%) accidentswhereas in 280 (83%) accidents the trailer truck driverwas considered to be the second participant. Also 16single-vehicle accidents had occurred, involving onlythe trailer truck. However, single-vehicle accidents wereexcluded from the analysis regarding drivers responsi-bility, for two reasons. Firstly, previous studies haveindicated that there are significant differences in causal-ity factors between single and multi-vehicle accidents(Persaud and Mucsi, 1995; Ivan et al., 1999). In thepresent study, a fairly small number of single-vehicleaccidents did not make feasible a separate analysis forthem. Secondly, two-vehicle accidents make possible abetter controlled analyses of the exposure factors. Driv-ers who were second participants were thought toprovide an estimate for exposure by being involvedincidentally (Thorpe, 1964; Haight, 1973; Lyles et al.,1991; Summala and Mikkola, 1994). In head-oncrashes, that are most numerous fatalities involving aheavy vehicle in Finland where main road system con-sists of two-lane roads, the elementary unit of exposureis a meeting of two vehicles (e.g. Chapman, 1973;Summala, 1996). In each case the a priori chances of

drifting into the lane of opposite traffic should be thesame for both vehicles and responsibility can be consid-ered as a measure that indicates failure on performance,whether it be due to errors in attention, arousal level,etc.

In none of the accidents had the truck driver beenintoxicated or under the influence of drugs. Generally,the low number of alcohol related cases in the presentdata indicate, in accordance to earlier studies, thatdrunk driving is seldom a problem for truckers, interms of fatal accidents (Brown, 1993; Summala andMikkola, 1994).

2.2. Sur6ey

We collected data by sending an anonymous ques-tionnaire to 2000 randomly sampled members of anon-political organisation promoting truckers interests,which had a current membership of 16 508 personsinvolved in the truck industry. A total of 669 members(34%) with different work descriptions returned thequestionnaire. For the present study we focused ouranalysis on long-haul drivers (who often drive a trailertruck), leaving responses of other drivers and peoplewith other work descriptions outside of the analysis. Atotal of 251 respondents were long-haul drivers, all ofthem men.

The questionnaire contained questions of individualcharacteristics (such as age, driving experience, etc.)and the description of the preceding 3 months work. Inaddition, frequency of fatigue-related problems whiledriving (such as frequency of dozing off and near misssituations due to dozing off) were surveyed. Further-more, drivers were asked to rank eight different factors(such as weather condition, technological fault anddriver fatigue, see Fig. 1) in numbered order, based onwhat they viewed to be the most common contributorfor accidents where trucks are involved (giving numberone), the second common (giving number two), etc. Inaddition, a short and general description (without tech-nical details) was given of a technological countermea-sure to detect driver fatigue, and an open view wasasked to be written. These views were further classifiedas representing positive, negative or neutral views.

3. Results

3.1. Dri6ers at fault and not at-fault in fatal accidents

Table 1 presents the type of accidents that hadoccurred to trailer truck drivers who were principallyresponsible and to drivers who were second participantsin the accident. In both categories the most usualaccident type was a head-on crash with a vehicle com-ing from the opposite direction. This was expected due

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196190

Table 1Type of accident separately for drivers who were principally responsible for the accident and for second participants

Principally responsible Second participant Total (%)(%)(%)n280 n337n57

5.7Same driving course (overtaking, change of lane or rear-end collision) 7.717.53.610.5 4.7Same driving course, one vehicle turning

50.9Opposite driving courses, head-on collision 73.5 69.82.5Opposite driving courses, one vehicle turning 2.10.07.98.8 8.0Crossing driving course

5.3Crossing driving course, one vehicle turning 6.4 6.2Other 0.47.0 1.5

100.0 100.0Total 100.0

to the fact that in Finland the main road systemconsists mainly of two-lane highways.

Table 2 presents the primary causal factors for acci-dents where the driver was principally responsible.Most often the causal factor was related to driverserrors in operating the vehicle and cognitive functions.

In the following, univariate comparisons will be firstpresented, followed by multivariate (logistic regression)analysis. The univariate comparisons between drivercategories did not reveal any significant differences inindividual factors. Drivers who were principally respon-sible were somewhat younger but the difference did notquite reach a significant level (means 35.8 vs. 38.4 years,F1,3353.273, P0.071). There were no significant dif-ferences in driving experience with the truck in questionor the proportion of self-employed drivers (23 vs. 30%,respectively).

Although drivers who were responsible for the acci-dent more frequently had a discovered chronic illness(19 vs. 10%), the difference in the univariate compari-son did not quite reach a significant level (x23.146,df1, P0.076, data missing from ten drivers at faultand from 33 drivers not at fault). Of the drivers whowere responsible one had heart disease, hypertension aswell as musculoskeletal pains (the accident of this driverwas triggered by his sudden attack of illness). Otherchronic illnesses occurring were: (number of drivers inparenthesis) visual defect (1), insomnia (1), hearingdefect (1), diabetes (1), hypertension (2) and someother (2). Illnesses occurring among drivers who hadbeen second participants were: heart disease, visualdefect as well as diabetes (1), hypertension as well asdiabetes (1), hypertension as well as hearing defect (1),hearing defect (2), diabetes (2), hypertension (5), heartdisease (1), renal disease (1), musculoskeletal disease(1), difficult handicap (1) and some other (9). Com-pared to the healthy drivers, the drivers with an illnesswere, as expected, significantly older (44.7 vs. 37.2years, F1,29218.806, PB0.001), but there was no sig-nificant difference regarding their accident history, cov-ering the preceding 5 years.

Accident and especially traffic offence history of thedriver was considered to indicate, to a certain extent,driving style. Of all the drivers 39% had been involvedin an accident at least once and approximately 20% hadreceived at least five traffic offences but no significantdifferences occurred between driver categories (35 and40% with at least one accident and 19 and 20% with atleast five offences). Of the preceding offences 27%concerned causing a danger situation in traffic, 3%driving while intoxicated and 49% were indiscriminatelyclassified as other traffic offences.

Somewhat surprisingly there was no significant dif-ference between driver categories in relation to drivingtime preceding the accident (P0.578, MannWhitneyU-Test). Of the drivers who were responsible 17% andof those who were second participants 12% had beendriving for more than 10 h preceding the accident(which has been criminally punishable in Finland since1995, under the EC Regulation). Approximately 7% ofthe drivers in both driver categories reported furtherhaving slept less than 5 h while being off from workearlier.

Only two (4%) of the accidents where the driver wasresponsible had occurred during the early morninghours, between 00:00 and 06:00 h, compared to 16% ofthe accidents where the driver was the second partici-pant. Of the accidents where the driver was responsible32% occurred between 18:00 and 00:00 h, compared to

Table 2Different primary factors contributing for accidents among driverswho were principally responsible for the accident

n %

5.3Driver having fallen asleep while driving 31Attack of illness 1.8

Error in attention, anticipation or estimation 50.82915 26.3Error in operating the vehicle

7.0Technological faults 41Traffic environment 1.8

7.04Other reasons57Total 100.0

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196 191

Fig. 1. Mean values of ranks (out of eight categories) for views regarding the causes of accidents where trucks are involved (1, most commoncause; 2, second common cause, etc. 8, least common).

19% of the accidents where the truck driver was thesecond participant (0x24.869, df1, P0.027).This result may be partly explained by the increase intraffic density (compared to early morning hours) butindicates also, to some respect, that driving perfor-mance is subject to diurnal variation, making errorsmore likely at evening hours.

A logistic regression analysis was carried out in orderto analyse how the responsibility of the driver varieswith driver characteristics and time of day when all therelevant variables are controlled. For this purpose, theresponsibility of the driver was set as the dependentvariable, the independent variables were subdividedinto approximately equally sized groups, reference cate-gories were determined and missing values were re-placed with modes and where appropriate, with means.The results shown in Table 3 indicate that compared toover 50 year old drivers, the odds of being responsiblefor the accident was significantly increased, by a factorof 3.5 if the driver was younger than 30 years.

In addition, although the frequency of a chronicillness did not reach a significant level in the univariatecomparison the logistic regression analysis showed thatwhen other variables were controlled (and missing val-ues for the illness variable were replaced with noillness), a chronic illness of the driver increased theprobability of being responsible for the accident by afactor of over three. In other words, compared tohealthy drivers, drivers with an illness had over threetimes higher risk for being involved in an accident asthe responsible participant.

Time of day had also a significant predictive effect inthe logistic regression analysis. The probability of being

responsible increased by about two if the accidentoccurred during the evening hours. This may be partlydue to the effect of circadian rhythm, making errorsmore likely to occur at evening hours. However, moti-vational tendencies to finish the journey at the end ofthe day may also partly explain the result. The effectsof accident and traffic offence history as well as drivingtime preceding the accident on the whole and time sleptpreviously while off work remained non-significant. Inall, the model for predicting drivers responsibility withthe present independent variables did reach a significantlevel (x237.045, df23, P0.032), although theoverall R square remained rather low (R20.174).

3.2. Dri6er fatigue and prolonged dri6ing

According to the information gathered by the multi-disciplinary teams, driver fatigue contributed onlyslightly to the accidents. It was roughly estimated,based on the driver interviews and other relevant infor-mation (such as driving time preceding the accident,eyewitness testimonies, etc.) that three drivers who wereprincipally responsible for the accident (n57) hadactually fallen asleep while driving, just preceding theaccident. In addition, the multidisciplinary teams cau-tiously concluded that the driver had been tired preced-ing the accident in four of the accidents where he wasresponsible and in six accidents where he was thesecond participant.

A total of 13% of all the drivers had been driving formore than 10 h preceding the accident. A logisticregression analysis showed that neither driver age, driv-ing experience with the truck in question, self-employ-ment, occurrence of a chronic illness, accident history

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196192

Table 3Summary of logistic regression analysis for being principally responsible for the accident as the dependent variable

Independent variables Odds ratioB Statistical significance (P )

Dri6er age (]50 years as a reference category) 0.1251.237529 3.447 0.044

3039 0.691 1.996 0.2561.4840.395 0.5104049

0.062Dri6ing experience with the truck in question (o6er 100 000 km as reference category)1.30250 000100 000 km 0.272 0.026

0.5700.561 0.19950 000100 00 km1.031B10 000 km 0.357 0.088

0.685Self employment 0.3260.378

3.104 0.025Occurrence of an illness 1.133

Traffic offences during past 5 years (no offences as reference category) 0.3360.100One 1.106 0.843

Two 1.5630.447 0.3771.7220.543 0.328Three

1.200Four 0.301 0.162At least five 0.552 1.737 0.368

Traffic accidents during the past 5 years (no accidents as reference category) 0.6630.720 0.345One 0.3280.7530.284 0.632At least two

0.123Dri6ing time preceding the accident (B1 h as reference category)3.113 0.04613 1.1361.0540.052 0.93036

0.717610 2.048 0.2180.601\10 1.824 0.351

Preceding sleeping time while off work (\8 h as reference category) 0.5890.13778 0.872 0.747

67 1.1030.098 0.8440.404 0.1900.906B6

2.095Accident occurring between 08:00 and 00:00 h 0.0400.740

nor the responsibility in the present accident predicted,at a statistically significant level, driving over 10 h.Traffic offence history did not yield any additionalsignificant effect as such. However, when compared todrivers with no offences, drivers having at least fiveoffences did have an increased probability of havingbeen driving over 10 h. Prolonged driving preceding theaccident was also strongly related to the time of day,the probability increasing significantly during eveninghours. However, the overall logistic model did not quitereach a significant level (x226.280, df17, P0.069), and the R square of the model remained low(0.145).

3.3. Accident causes as 6iewed by other truck dri6ers

Fig. 1 presents the mean rank values for the possiblecauses of accidents where trucks are involved as viewedby the long-haul drivers responding to the question-naire. Most often accidents were thought to occur dueto other road users and errors in drivers perception,anticipation and:or estimation. Technological faults

and factors in traffic environment were considered to bethe least common causes and the most variance ap-peared in weather (SD2.8) and driver fatigue (SD2.0) as causes.

The views of causal estimates stem at least partlyfrom drivers own experiences. Drivers who viewedfatigue to be among the three most usual causes (29%in all) self-reported also having recently experiencedmore frequent dozing off while driving (P0.009) andnear miss situations due to dozing off (P0.018,MannWhitney U-Test, respectively). A total of 34%of these driver had dozed off at least twice and 15% hadexperienced a near miss situation (compared to 21 and6% of drivers viewing fatigue as a more rare cause). Inaddition, these drivers were less frequently self-em-ployed drivers (8 vs. 22%, x26.389, df1, P0.011).

The results of a multiple regression analysis, shownin Table 4, indicated that driver age, truck drivingexperience or shift type did not have a significant effectin predicting the rank for driver fatigue as a cause ofaccidents. However, self-employment did significantly

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196 193

predict the view so that driver fatigue was seen as amore rare cause. Furthermore, near miss situationspredicted the view so that the drivers with increasedfrequency of near miss situations viewed fatigue as amore common cause. In all, although the overall Rsquare remained low, the model did reach a significantlevel.

3.4. Views regarding the de6elopment of a technologicalcountermeasure against dri6er fatigue

A total of 47% of the drivers expressed a negativeview towards developing a technological countermea-sure to cope with driver fatigue. Most common argu-ments were that the driver should know himself whenhe is too tired to continue and that truck drivers arealready now under a too strict control. Somewhatsurprisingly, negative views towards the idea were ex-pressed more frequently by drivers who had been driv-ing a regular night shift or irregular day and night shiftduring the past 3 months than by drivers who had beendriving a regular daytime shift (66 vs. 34%, x24.357,df1, P0.037). However, drivers with a regularnight shift or irregular shift (60% of all) did not reportsignificantly more frequent fatigue-related problemswhile driving.

Furthermore, 34% of the surveyed drivers reportedhaving dozed off at least twice during the preceding 3months but nearly half of these drivers still expressednegative views. In addition, 9% of the drivers reportedthat they had previously been involved in an accidentdue to dozing off but still 38% of these drivers ex-pressed negative views towards a possible countermea-sure (compared to 56% of the accident-free drivers,x21.721, df1, P0.190).

The results of a logistic regression analysis indicatedthat neither driver age, truck driving experience, self-

employment, the rank for fatigue as a cause for acci-dents or frequency of dozing off had any significanteffect in predicting the negative view towards the possi-ble countermeasure. In addition, the effect of near misssituations related to dozing off did not quite reach asignificant level (P0.072). The only significant effectwas related to shift type and appeared already in theunivariate comparison (the probability of having a neg-ative view increasing if the driver had been driving anirregular night and day shift compared to driving aregular day shift). In all, the model did not reach asignificant level (x212.341, df9, P0.195, R20.048).

4. Discussion

The accidents where trailer trucks are involved areoften collisions between two vehicles driving the oppo-site driving courses. In accordance with our previousresults (Summala and Mikkola, 1994), only 17% of thetruck drivers were considered to be principally responsi-ble for the accident. As it is often only the truck driverwho survives from the accident to tell his own story ofwhat happened it could be argued that the reliability ofthe assessment regarding the responsibility is question-able. Such a bias was indeed shown by Naatanen andSummala (1972) in cyclist and pedestrian fatalities. Intwo-vehicle crashes however, the multidisciplinaryteams are generally able to reconstruct the accidentsequence on the basis of hard data (marks in thepavement, final position of vehicles, etc.) to the pointthat, for example the exact lateral location (lane) of thecrash can be determined. This implies responsibility interms of who lost the control and drifted away fromhis:her own lane.

In cases where the trailer truck driver was principallyresponsible for the accident, the crash was most oftentriggered due to an error in operating the vehicle or indrivers perception, anticipation or estimation. Unfor-tunately, however, even the present data, based onin-depth analysis of the accidents, cannot give us moredefinitive information regarding the causal factors. Ithas previously been expressed that some of the acci-dents involving inattention or wrong anticipation mightbe related to driver fatigue and that accident investiga-tors should further inquire why such a cognitive erroroccurred (Horne, 1992).

In the present study, the trailer truck driver wasestimated to have fallen asleep just prior to the accidentin only 2% of the accidents. This result correspondswell with our earlier study where the same database, forthe years 19841989 was used (Summala and Mikkola,1994). Also, consistently with the earlier results, driverfatigue altogether (taken into consideration those whohad fallen asleep and those who had felt tired preceding

Table 4Summary of multiple regression analysis for viewing driver fatigue asa cause of accidents involving trucks as the dependent variable

StatisticalbIndependentsignificancevariables

0.008 0.911Age0.096Lifetime truck driving experience 0.1860.175Self-employment (employee1, 0.005

self-employed2)Shift type (dayshift1, irregular or 0.5910.033

night shift2)0.215 0.001Near miss situations due to dozing

off0.097R2

0.079Adjusted R2

5.288Fdf 5.245Significance B0.001

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196194

the accident) was estimated to occur in 4% of the trailertruck drivers involved in an accident. However, sincethese results are based on interviews and analysis ofcircumstantial evidence they are to be considered asonly estimates. Determining what role, if any, fatigueplayed in a specific accident is often difficult for acci-dent investigators, although fatigue-related accidentseem to have certain characteristics (Lauber andKayten, 1989; Horne and Reyner, 1995; Pack et al.,1995). Previously however, it has been stated that fa-tigue-related crashes are often under-reported (Pack etal., 1995).

In general, identifying causes of accidents and possi-ble individual and external factors that increase theaccident risk of the driver improve the development ofeducational and technological countermeasures relatedto traffic safety. In accordance with a previous study oftruck drivers, the present results indicate that the riskof being involved in an accident as the responsibleparticipant is significantly increased among driversyounger than 30 years (Campbell, 1991). The presentresults indicate however, that younger driver age wasnot associated to more prone risk taking in terms ofprolonged driving. Also, contrary to earlier studies(Hamelin, 1987; Kaneko and Jovanis, 1992; Summalaand Mikkola, 1994), the effect of total driving hourspreceding the accident to the responsibility of the driverwas non-significant.

The present results also indicate that past drivingstyle (in terms of accident and traffic offence history)did not significantly contribute in predicting the respon-sibility of the drivers. These results are contrary to theprevious results obtained from private drivers (Robert-son and Baker, 1975; Summala et al., 1986; Rajalin,1994). However, it has also been previously shown thatheavy-vehicle drivers in general tend to drive morecautiously, in terms of violations, after being involvedin a fatal accident (Rajalin and Summala, 1997).

The present results raise an increased concern interms of drivers medical condition having a significanteffect on the responsibility of the driver. A total of 12%of the drivers had a chronic illness, which increased theodds of being responsible for the accident by a factor ofover three. However, the drivers sudden attack of anillness was a causal factor in only one accident. Mostoften the accidents where the truck driver with anillness was responsible were triggered by an error incognitive functions. However, there was no consistencybetween the type of an illnesses and causal factor. Onepossible explanation is that drivers with an illness hadgenerally also more lowered subjective well being andwere increasingly unable to react to the demands of thetraffic environment. However, the relatively smallamount of data in the present result does limit thepossibility of making any further conclusions but theresults do indeed support the need for further studies.

It is, however, to be remembered that in general themedical condition among truck drivers was surprisinglygood, although a previous study in United Statesshowed that truck drivers would benefit from a healthpromotion program (Korelitz et al., 1993). In addition,according to the multidisciplinary teams none of thedrivers suffered from exceptionally increased daytimesleepiness, often a common symptom in sleep disorders.Altogether, the present results support the view thatself-selection effectively discards those drivers withpoorer medical conditions or serious sleepiness-relatedproblems from this demanding profession (Hakkanen etal., 1999; Hakkanen & Summala, 2000a,b).

Although over 10% of all the drivers had been driv-ing over 10 h (now criminally punishable under the ECRegulation) preceding the accident it did not have asignificant effect on being responsible for the accident.However, considering driving hour restrictions as un-necessary based on the present results would be ratherinappropriate when the relatively small sample size ofthis study and the several previous results indicatingincreased risk of driver fatigue-related problems afterprolonged driving are taken into account. Unfortu-nately however, predicting driving hour violators seemsto be rather difficult since no individual factors, occur-rence of an illness or past accident history had asignificant effect in the present study. However, theodds of having been driving over ten hours precedingthe accident increased significantly during eveninghours. At these hours the subjective risk of gettingcaught is low which might, together with increasedmotivational tendencies to finish the day and demandsof the job, encourage prolonged driving.

The views of accident causes among the surveyedlong-haul drivers correspond well with the results of theaccident analysis and stem at least partly from driversown experiences. A total of 83% of the accidents wheretrailer truck drivers had been involved, during 19911997, were originated by other road users. Accidentswere also viewed, by the surveyed long-haul drivers, asoccurring most often due to other road users. Driverfatigue was in average ranked as the fifth commoncause of accidents involving trucks. Nevertheless, recentpersonal experiences of dozing off and related near misssituations made drivers rank driver fatigue higheramong the most frequent causes of accidents. In addi-tion, self-employed drivers particularly ranked driverfatigue to be a less common cause of accidents. Takinginto consideration their reluctance for restricting driv-ing hours, often expressed in public discussion in theEU, the result is hardly surprising.

Quite surprisingly however, nearly half of the driversexpressed a negative view towards developing a possibletechnological countermeasure to cope with driver fa-tigue. In addition, increased experiences of driver fa-tigue-related problems do not make drivers opinion of

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196 195

technological countermeasures more positive. This is, insome respects, in line with the theoretical position thatdrivers develop a strong subjective control, and thatthey feel they can manage the task despite some lapsesevery now and then (Naatanen and Summala, 1976;Summala, 1988). Furthermore, the negative view wasfound to be more common among drivers who drive anight shift than among drivers with a day shift al-though the former should need it more (at night thediurnal variation is at its lowest level and drivingperformance has been shown to be subject to thisvariation, Lenne et al., 1997). It can be hypothesisedthat the result also reflects drivers high subjectivecoping skills over fatigue, especially among those spe-cifically involved, and that drivers who drive a regularday shift view technical development positive in generalwhile not affecting them personally. It is like truckdrivers opinion of road safety campaigns for truckers:good for others but I dont need them, I already knowthe risks and can manage them (Summala andPihlman, 1993).

A noticeable result is indeed that the rank for fatigueas a more common cause of truck accidents has nosignificant effect on the view towards the countermea-sure. In general, the countermeasure was considered toexcessively restrict drivers ability to freely choosewhether to stop for a rest or to continue the journey,and in general, rather pessimistic views regarding thereliability of such a measure were expressed. Theseresults altogether raise the question of truck driversmotivation for using any possible countermeasure inthe future. Altogether the present results strongly indi-cate that the development of a possible countermeasureand its market penetration may be facing seriousresistance.

Acknowledgements

This study was supported by grants from the TrafficSafety Committee of Insurance Companies and theFinnish Cultural Foundation.

References

Arnold, P.K., Hartley, L.R., Corry, A., Hochstadt, D., Penna, F.,Feyer, A.M., 1997. Hours of work, and perceptions of fatigueamong truck drivers. Accident Analysis and Prevention 29 (4),471477.

Braver, E.R., Preusser, C.W., Preusser, D.F., Baum, H.M., Beilock,R., Ulmer, R., 1992. Long hours and fatigue: a survey of tractortrailer drivers. Journal of Public Health Policy 13, 341366.

Brown, I.D., 1993. Reducing accident risk for the heavy goods vehicledriver. Journal of Traffic Medicine 21 (1), 7983.

Brown, I.D., 1995. Methodological issues in driving fatigue research.In: Hartley, L. (Ed.), Fatigue and Driving. Taylor and Francis,London, pp. 155166.

Brown, I.D., 1997. Prospects for technological countermeasuresagainst driver fatigue. Accident Analysis and Prevention 29, 525531.

Campbell, K.L., 1991. Fatal accident involvement rates by driver agefor large trucks. Accident Analysis and Prevention 23, 287295.

Chapman, R., 1973. The concept of exposure. Accident Analysis andPrevention 5, 95110.

Dionne, G., Desjardins, D., Laberge-Nadeau, C., Maag, U., 1995.Medical condition, risk exposure, and truck drivers accidents: ananalysis with count data regression models. Accident Analysis andPrevention 27, 295305.

Eicher, J.P., Robertson, H.D., Toth, G.R., 1982. A report to Con-gress on large-truck accident causation. DOT:HS-806-300.Springfield, VA: National Technical Information Service.

Evans, L., 1991. Traffic Safety and the Driver. Van Nostrand Rein-hold, New York.

Feyer, A.M., Williamson, A.M., 1995. Managing driver fatigue in thelong-haul road transport industry: interim report of nationalresearch programme. In: Hartley, L. (Ed.), Fatigue and Driving.Taylor and Francis, London, pp. 2532.

Haight, F., 1973. Induced exposure. Accident Analysis and Preven-tion 5, 111126.

Hakkanen, H., Summala H., 2000a. Sleepiness at work in commercialtruck drivers, Sleep 23 (1), 4957.

Hakkanen, H., Summala H., 2000b. Driver sleepiness-related prob-lems, health status and prolonged driving among professionalheavy vehicle drivers. Transportation Human Factors 2 (2).

Hakkanen, H., Summala, H., Partinen, M., Tiihonen, M., Silvo, J.,1999. Blink duration as an indicator of driver sleepiness inprofessional bus drivers. Sleep 22 (6), 798802.

Hamelin, P., 1981. Les conditions temporelles de travail des conduc-teurs routiers et la securite routie`re. Travail Humain 44 (1), 521.

Hamelin, P., 1987. Lorry drivers time habits in work and theirinvolvement in traffic accidents. Ergonomics 30 (9), 13231333.

Hantula, L., 1989. Case studies of road accidents in Finland. Journalof Traffic Medicine 17, 2329.

Hantula, L., 1992. Development of a case study method of roadaccidents in Finland. Journal of Traffic Medicine 20, 2735.

Hertz, R.P., 1991. Hours of service violations among tractortrailerdrivers. Accident Analysis and Prevention 23, 2936.

Horne, J., 1992. Stay awake, stay alive. New Scientist 4 (1), 2024.Horne, J., Reyner, L.A., 1995. Sleep-related vehicle accidents. British

Medical Journal 310, 565567.Ivan, J.N., Raghubhushan, K.P., Ossenbruggen, P.J., 1999. Differ-

ences in causality factors for single and multi-vehicle crashes ontwo-lane roads. Accident Analysis and Prevention 31, 695704.

Kaneko, T., Jovanis, P., 1992. Multiday driving patterns and motorcarrier accident risk: a disaggregate analysis. Accident Analysisand Prevention 24 (5), 437456.

Korelitz, J.J, Fernandez, A.A., Uyeda, V.J., Spivey, G.H., et al.,1993. Health habits and risk factors among truck drivers visitinga health booth during a trucker trade show. American Journal ofHealth Promotion 8 (2), 117123.

Lauber, J.K., Kayten, P.J., 1989. Fatigue, alcohol and drug involve-ment in transportation accidents. Alcohol Drugs and Driving 5(3), 173184.

Lenne, G.M., Triggs, T.J., Redman, J.R., 1997. Time of day varia-tions in driving performance. Accident Analysis and Prevention29 (4), 431437.

Lyles, R.W., Stamatiadis, P., Lighthizer, D.R., 1991. Quasi-inducedexposure revisited. Accident Analysis and Prevention 23, 275285.

Lyznicki, J.M., Doege, T.C., Davis, R.M., Williams, M.A., 1998.Sleepiness, driving, and motor vehicle crashes. Journal of Ameri-can Medical Association 279 (23), 19081913.

McDonald, N., 1984. Fatigue, safety and the truck driver. Taylor andFrancis, London.

H. Hakkanen, H. Summala : Accident Analysis and Pre6ention 33 (2001) 187196196

Naatanen, R., Summala, H., 1972. Kevyt liikenne ja liikenneturval-lisuus (Unshielded traffic and traffic safety). Sosiaalilaaketieteelli-nen Aikakauslehti 10, 276281.

Naatanen, R., Summala, H., 1976. Road-user Behaviour and TrafficAccidents. North-Holland, Amsterdam.

Pack, A.I., Pack, A.M., Rodgman, E., Cucchiara, A., Dinges, D.F.,Schwab, C.W., 1995. Characteristics of accidents attributed to thedriver having fallen asleep. Accident Analysis and Prevention 27,769775.

Persaud, B., Mucsi, K., 1995. Microscopic accident potential modelsfor two-lane rural roads. Transportation Research Record 1485,TRB, National Research Council, Washington, DC, pp. 134139.

Rajalin, S., 1994. The connection between risky driving and involve-ment in fatal accidents. Accident Analysis and Prevention 26,555562.

Rajalin, S., Summala, H., 1997. What surviving drivers learn from afatal road accident. Accident Analysis and Prevention 29, 277283.

Robertson, L.S., Baker, S.P., 1975. Prior violation record of 1447drivers involved in fatal crashes. Accident Analysis and Preven-tion 7, 121128.

Sluiter, J.K., van der Beek, A.J., Frings-Dresen, M.H.W., 1999. Theinfluence of work characteristics on the need for recovery andexperienced health: a study on coach drivers. Ergonomics 42 (4),573583.

Summala, H., 1988. Risk control is not risk adjustment: the zero-risktheory of driver behaviour and its implications. Ergonomics 31(4), 491506.

Summala, H., 1996. Accident risk and driver behaviour. SafetyScience 22, 103111.

Summala, H., Pihlman, M., 1993. Activating a safety message fromtruck drivers memory: an experiment in a work zone. SafetyScience 16, 675687.

Summala, H., Mikkola, T., 1994. Fatal accidents among car andtruck drivers: effects of fatigue, age, and alcohol consumption.Human Factors 36 (2), 315326.

Summala, H., Kanninen, K., Kanninen, J., Rantanen, P., Virtanen,A., 1986. Passing when visibility obscured: can we predict acci-dent involvement from one single traffic maneuver? Proceedingsof the Human Factors Society 30, 249250.

Summala, H., Hakkanen, H., Mikkola, T., Sinkkonen, J., 1999. Taskeffects on fatigue symptoms in overnight driving. Ergonomics 42(6), 798806.

Thorpe, J.D., 1964. Calculating relative involvement rates of accidentwithout determining exposure. Australian Road Research 2 (1),2536.

Traffic Safety Committee of Insurance Companies. 1987. Road acci-dent Investigation Teams in Finland, 4th ed., Helsinki, Finland:Author.

.