Traffic Injury Prevention

ISSN: 1538-9588 (Print) 1538-957X (Online) Journal homepage: https://www.tandfonline.com/loi/gcpi20

Factors associated with driving under the

influence of alcohol

Emir Smailović, Krsto Lipovac, Dalibor Pešić & Boris Antić

To cite this article: Emir Smailović, Krsto Lipovac, Dalibor Pešić & Boris Antić (2019) Factors
associated with driving under the influence of alcohol, Traffic Injury Prevention, 20:4, 343-347, DOI:
10.1080/15389588.2019.1605168

To link to this article: https://doi.org/10.1080/15389588.2019.1605168

View supplementary material

Published online: 21 May 2019.

Submit your article to this journal

Article views: 98

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=gcpi20
TRAFFIC INJURY PREVENTION
2019, VOL. 20, NO. 4, 343–347
https://doi.org/10.1080/15389588.2019.1605168

Factors associated with driving under the influence of alcohol

Emir Smailovic , Krsto Lipovac , Dalibor Pesic, and Boris Antic
University of Belgrade, Faculty of Transport and Traffic Engineering, Belgrade, Serbia

ABSTRACT ARTICLE HISTORY

Objective: This study examined the risk factors of driving under the influence of alcohol (DUI) Received 3 December 2018
among drivers of specific vehicle categories (DSC). On the basis of this research, the variables Accepted 1 April 2019
related to DUI and involvement in traffic crashes were defined. The analysis was conducted for car
KEYWORDS
drivers, bicyclists, motorcyclists, bus drivers, and truck drivers.
Driving under the influence;
Method: The research sample included drivers involved in traffic crashes on the territory of Serbia in risk factors; drivers of
2016 (60,666). Two types of analyses were conducted in this study. Logistic regression established specific vehicle category;
the correlation between DUI and DSC and the The Technique for Order of Preference by Similarity logistic regression; TOPSIS
to Ideal Solution (Multi-criteria decision making) method was applied to consider the scoring and
explore the potential for the prevalence of DUI on the basis of 2 data sets (DUI and non DUI).
Results: The study results showed that driver error and male drivers were the 2 most significant
risk factors for DUI, with the highest scores and potential for prevalence. The nonuse of restraint
systems, driver experience, and driver age are the factors with a significant prediction of involve-
ment in an accident and an insignificant prediction of DUI.
Conclusions: Following the development of the logistic prediction models for DUI drivers, testing
of the model was conducted for 3 control driver groups: Car, motorcycle, and bicycle. The predic-
tion model with a probability greater than 50% showed that 77% of car drivers were under the
influence of alcohol. Similarly, the prediction percentage for motorcyclists and bicyclists amounted
to 71 and 67%, respectively. The recommendation of the study is that drivers whose DUI probabil-
ity is above 50% should be potentially suspected of DUI. The results of this study can help to
understand the problem of DUI among specific driver categories and detect DUI drivers, with the
aim of creating successful traffic safety policy.

Introduction of alcohol in a driver’s breath, the time needed for a reac-

The occurrence of traffic accidents is connected to speed or
driving under the influence of alcohol, whereas the risk of
injuries is connected to the use of seat belts (World Health
Organization 2004). Alcohol consumption is undoubtedly
connected to the occurrence of traffic accidents (Mann et al.
2010), and it has negative effects on drivers’ capabilities in
terms of attention, tracking, reaction time, information
processing, perception, psychomotor skills, visual function,
and increased risk-taking (Ogden and Moskowitz 2004;
Penning et al. 2010).
Numerous studies have shown that the presence of alco-
hol in blood exponentially increases the risk of collision
(Asbridge et al. 2004; Compton and Bering 2015; Mann
et al. 2001, 2010). A driver’s performance under the influ-
ence of alcohol is lower compared to the performance of
sober drivers. In their research, Li et al. (2016) examined
this by testing 52 Chinese drivers in a simulator, where it
was shown that the drivers’ performance, such as reaction
time and lateral position on the track, was highly correlated
with consumption of alcohol. With an increase of one unit
tion is decreased by 0.3%. For details about the background
and factors associated with driving under the influence
(DUI) see Appendix A (online supplement).
In this study, the risk factors related to involvement in
traffic crashes are separately analyzed depending on the
driver category. Namely, the actions against DUI prevalence
should involve all drivers. The actions directed at one driver
category (e.g., car drivers) will not have an effect on other
driver categories. The research has shown that drivers of dif-
ferent categories have a different risk of being involved in
traffic crashes and DUI. Thus, because of their number, car
drivers have the highest share of involvement in traffic
crashes. On the other hand, bicyclists have a lower share of
involvement in crashes but suffer more severe consequences.
Zero alcohol tolerance for commercial drivers, which has
been implemented in a few countries of Europe, is due to
the consequences of DUI truck vehicles. This is why there is
a need to study risk factors of being involved in traffic
crashes and DUI for different driver categories, as well as
for the whole population of drivers.

CONTACT Emir Smailovic e.smailovic@sf.bg.ac.rs Faculty of Transport and Traffic Engineering, Department for Traffic Safety and Road Vehicles, Vojvode
Stepe 305, Belgrade 11000, Serbia.
Associate Editor Kathy Stewart oversaw the review of this article.
Supplemental data for this article can be accessed on the publisher’s website.
ß 2019 Taylor & Francis Group, LLC
344  ET AL.
E. SMAILOVIC
All drivers of specific vehicle categories (DSC) are inclined
to DUI. The initial hypothesis is that commercial drivers are
involved in alcohol-related traffic crashes less frequently, so dif-
ferent risk factors are expected than for car drivers, bicyclists,
and motorcyclists. For the efficient decrease in DUI prevalence,
the importance and potential of factors for all driver groups
have to be determined. The contribution of the study is defin-
ing the risk factors for each driver category individually and
determining the priority of the prevalence of DUI for the
whole population of drivers. Engaging significant resources
(human and financial) in order to effect a certain factor will
not lead to satisfactory results if the factor’s potential is limited.
The aim of this study is to examine the potential of the preva-
lence factor in order to direct the activities in an optimal man-
ner. The prevalence potential was observed on the basis of
drivers’ involvement in DUI crashes and non-DUI crashes.
The analysis of risk factors for DSC was carried out in order
to establish the potential for prevalence and drivers of specific
vehicle category, applying 5 different logistic models. The score
and potential for the prevalence of risk factors was obtained
using the Technique for Order of Preference by Similarity to
Ideal Solution (TOPSIS) for DUI crashes and non-DUI crashes.
Methods
This study is based on the factors associated with DUI
among drivers involved in road traffic crashes in Serbia in
2016. For each driver involved in a traffic crash in Serbia,
the data are recorded in the National Traffic Accident
Database (see Appendix B, online supplement). The database
contains data on drivers involved in traffic crashes and
details of the crash. The national database includes the data
obtained from the official police investigation of the traffic
crash conducted immediately following the crash.
The study is based on all drivers involved in traffic
crashes in Serbia in 2016. Fatally injury drivers, injured driv-
ers, and drivers without injuries were the subjects of this
study. Frivers without injuries were included in this study
because a large number of these drivers were under the
influence of alcohol. Namely, out of 60,666 drivers involved
in crashes, about 10% were DUI, and about 8% of DUI driv-
ers had no injuries (see Appendix B). In other words, 4 out
of 5 of DUI drivers had no injuries.
Two types of analyses were conducted in this study.
Binary logistic regression established the correlation between
DSC and DUI. Logistic regression was applied to analyze
the association between DUI and risk factors using 5 logistic
models: car drivers, bicyclists, motorcyclists, bus drivers, and
truck drivers. Each of these models represents DSC. Using
the TOPSIS method, we obtained the score of the risk fac-
tors for DUI and the potential for prevalence. For the details
about the logistic regression and TOPSIS method, see
Appendix C (see online supplement).
Variables
The need for the comprehensive analysis of DUI drivers of
different vehicle categories in traffic crashes means that the
largest number of variables should be included in order to
examine the effect of individual variables. Reviewing the lit-
erature, we found a few variables related to DUI and
involvement in traffic crashes. Variables such as gender, age,
and driver experience have been proven more than once,
but their potential for prevention is less well known. In add-
ition, data availability and reliability represent limitations for
the realization of such a study. Rejecting unclear and
unknown variables collected in the research on traffic
crashes, we systematized the selection of the variables exam-
ined in this study into the following 8 factors:

Age (number of years)

Gender (male or female)

Use of restraint systems (seat belt or helmet)

Driver experience (number of years between obtaining a
driver’s license and involvement in a traffic crash; not
including bicyclists)

Trip/journey purpose (mandatory or nonmandatory)

Drivers of specific vehicle categories

Responsibility for the crash: Driver; driver of the other
vehicle or other

Area of the crash (urban or rural).
Involvement in a traffic crash represents a combination
of a number of risk factors, such as exposure, speeding, lack
of attention, mistakes, alcohol consumption, and others.
Alcohol, as one of the influential factors, was isolated in this
research and its impact was studied. Exposure was not the
subject of the research; it was studied indirectly through the
correlation between exposure and involvement in traffic
crashes. In addition to exposure, factors such as speed, alco-
hol consumption and lack of attention are correlated with
involvement in traffic crashes. The study considered the
drivers involved in traffic crashes, which involves all of the
mentioned impacts, and alcohol was isolated and dealt with
separately in this study.
For car, bus, and truck drivers, the variable use of
restraint systems implies the use of seat belts, whereas for
bicyclists and motorcyclists the variable suggests the use of
helmets. It should be underlined that there is no legal obli-
gation for bicyclists to use a helmet in Serbia, whereas
motorcyclists are obliged to use one.
The variable driver experience implies the number of
years between obtaining a driver’s license and the driver’s
involvement in a traffic crash (i.e., 2016).
Examining previous studies, it was noticed that the pur-
pose of the journey was correlated with involvement in traf-
fic crashes (Soltani et al. 2016). The variable journey purpose
was divided into only 2 categories, according to previous
research by Soltani et al. (2016). The mandatory category
implies going to and from work, going to and from school,
and driving as part of one’s job, whereas a nonmandatory
purpose indicates driving during free time/for fun, driving
for the sake of driving, and driving during holidays.
This study analyzes commercial drivers (bus drivers and
truck drivers) because DUI among this group differs signifi-
cantly from that among other drivers. Commercial drivers
 TRAFFIC INJURY PREVENTION 345

Table 1. Sample description of DUI and non-DUI crashes.

Car driversa Bicyclistsb Motorcyclistsb Bus driversa Truck driversb
Positive (>0.00 mg/L) (%) 10.5 14.8 13.6 1.5 4.7
Age (mean) (years) 41.7/41.7 47.2/45.6 37.5/35.7 41.8/43.3 43.8/39.8
Driver experience (mean)c (years) 17.6/18.9 — 12.5/9.6 15.1/15.1 20.3/16.9
Gender Female (%) 6.3/17.4 7.1/25.9 1.4/1.6 8.3/0.4 1.3/3.0
Male (%) 93.7/82.6 92.9/74.1 98.6/98.4 91.7/99.6 98.7/97.0
Nonuse of restraint system,a,b (%) 5.4/2.0 96.4/84.1 21.4/9.9 16.7/5.2 2.6/1.6
Area Rural (%) 18.0/17.3 7.1/6.9 17.1/15.4 16.7/12.5 35.3/28.1
Urban (%) 82.0/82.7 92.9/93.1 82.9/84.6 83.3/87.5 64.7/71.9
Nonmandatory journey (%) 63.4/48.1 64.3/64.6 74.3/65.0 8.3/0.2 9.6/2.6
Driver responsibility (mistakes) (%) 86.8/53.3 71.4/46.6 77.1/45.9 66.7/37.6 81.4/57.7
a
Seat belt.
b
Helmet.
c
Not including bicyclists.

have a lower motivation to drive under the influence of Logistic regression

alcohol because they are afraid of losing their jobs (Park
et al. 2017). This is why the results of this study are import-
ant for linking DUI and commercial drivers. The initial
hypothesis is that commercial drivers are involved in traffic
crashes under the influence of alcohol to a lesser degree.
The variable responsibility for the crash implies a driver’s
mistakes that caused the traffic crash. The driver’s responsi-
bility for the crash represents the driver’s mistake that, in
accordance with the police investigation, represents the
cause of the crash and, as such, is registered by a police offi-
cer. These mistakes include driving without the right of way;
overtaking; speeding; misjudgment of the situation; and the
use of a cell phone. Responsibilities for the crash related to
another driver, pedestrian mistakes, traffic environment, or
vehicle breakdown represent the responsibilities of
another driver.
Results
Description of the sample
In all groups, motorcyclists were the youngest (35.7 years;
SD ¼ 12.3), with the least driving experience (9.6 years; SD
¼ 10.5). Bicyclists were the oldest group of drivers (45.6
years; SD ¼ 19.2), and car drivers had the most driving
experience (18.9 years; SD ¼ 12.4). Bicyclists were most
often involved in crashes under the influence of alcohol
(14.8%), followed by motorcyclists (13.6%), car drivers
(10.5%), truck drivers (4.7%), and bus drivers (1.5%). The
difference in driving experience for DUI crashes and non-
DUI crashes was noticeable regarding motorcyclists and
truck drivers, where older drivers were more often DUI
(Table 1).
Truck drivers used seat belts to the highest degree
(98.4%), followed by car drivers (98%) and bus drivers
(94.8%). In addition, 90.1% of motorcyclists and 15.9% of
bicyclists used a helmet at the time of the crash (Table 1).
Female drivers were involved in traffic crashes less fre-
quently and were less frequently under the influence of alco-
hol. A slightly smaller percentage of DUI drivers were
involved in traffic crashes during mandatory journeys, and
DUI drivers were more frequently responsible for the occur-
rence of traffic crashes than sober drivers (Table 1).
Five models of logistic regression are presented in Appendix
F (see online supplement). Model 1 includes car drivers,
where all variables, with the exception of age, are statistically
significant (Wald’s chi-square; P < .05, P < .01, P < .001).
Male drivers have greater odds for DUI (odds ratio
[OR] ¼ 3.623). Model 1 shows that the probability of DUI
for car drivers decreases by 0.5% with a 1-year increase in
the driver age. The odds for DUI are 1.8 times greater on
nonmandatory journeys. A driver has around a 6 times
greater odds of making a mistake if he or she is DUI (OR
¼ 6.486).
Model 2 of the logistic regression shows that statistically
significant variables of DUI for bicyclists are gender (DUI
males have 5 times greater odds [OR ¼ 5.371]) and mistake
(DUI bicyclists have 2.5 times greater odds of making a mis-
take [OR ¼ 2.514]).
The model for motorcyclists shows that the factors hel-
met use and responsibility for a crash are statistically signifi-
cant. DUI motorcyclists have smaller odds to use a helmet
and greater odds of making a mistake (Appendix F).
Statistically significant variables for DUI bus drivers are
gender and journey purpose. They have greater odds of DUI
if they are male and on a nonmandatory journey
(Appendix F).
For truck drivers, statistically significant variables for
DUI are the area, journey purpose, and driver’s mistakes.
Truck drivers have smaller odds of DUI if they are in an
urban area; 4.5 greater odds of DUI if they are on a non-
mandatory journey; and 3.3 greater odds of making mistake
if they are under the influence of alcohol (Appendix F).
TOPSIS method
The score of the risk factors is calculated on the basis of
ORs previously obtained through 5 logistic models
(Appendix G, see online supplement). The potential for
prevalence was observed by comparing DUI crash and non-
DUI crash scales of risk factors.
The factor scores of the results using the TOPSIS method
offer a clear picture of the significance of the factors related
to DUI. According to the results, we may conclude that
drivers’ mistakes and male drivers are the 2 most significant
factors related to DUI, with a score higher than 0.5. The
346  ET AL.
E. SMAILOVIC
third most significant risk factor for DUI is the nonmanda-
tory purpose of the journey.
The other factors are of lesser significance for DUI:
Driver experience, driver age, the area of driving, and non-
use of restraint systems.
For non-alcohol-related crashes, the priority of risks is
different. Involvement in traffic crashes by non-DUI drivers
is related to the nonuse of seat belts, and other factors have
significantly lower scores (Appendix G).
Discussion
This study identified DUI risk factors for 5 driver categories.
The risk factors help us better understand the problem of
DUI, lead to efficient prevalence measures, and facilitate the
identification of DUI drivers in traffic. Random police testing
has a limited scope if the factors characterizing DUI are not
comprehensively observed. Testing drivers in the period of
greater alcohol consumption (at night or on the weekend) is
not sufficient for the efficient reduction of DUI prevalence. It
is necessary to recognize driver groups inclined to DUI.
The key challenge of the model development in this study
is the prediction of DUI on the basis of available data.
Analysis of the model results raises the issue of application
of the model. The criterion for includion in the model was
that all suspected drivers were under the influence of alcohol
would require the testing of a relatively small number of
drivers and leave most DUI drivers unnoticed. On the other
hand, setting the criterion to a low probability of identifying
DUI drivers would enable testing of a large number of driv-
ers. In this situation, a large number of tests would not
identify DUI drivers due to the required low accuracy. On
the basis of the analysis of the developed logistic regression
models and the viability of the requested application, we
suggest that testing should be conducted for drivers who
have a DUI probability estimated as over 50%. This would
lower the risk of not including all drivers and most of the
tests would be successful. In order to explain this, we used
testing of the predictive performance of the model. A con-
trol sample was isolated from the initial data set. The con-
trol sample was isolated for car drivers, motorcyclists, and
bicyclists. Applying the model to the control sample, we
obtained the result that our models showed a DUI predic-
tion accuracy of 77% for car drivers, 71% for DUI motorcy-
clists, and 67% for DUI bicyclists. The prediction accuracy
represents the result of the model for drivers under the
influence of alcohol (with probability above 50%) in relation
to actual DUI drivers (in the control sample). Thus, the
results were validated, which justifies development of the
model. For details about the purpose of study, see Appendix
D (online supplement).
The TOPSIS DUI crash model showed the risk of DUI
drivers’ involvment in crashes, and the TOPSIS model for
other crashes showed the risk of non-DUI drivers’ involve-
ment in crashes. In this manner, DUI factors were verified
and the potential of DUI prevalence in relation to the fac-
tors was determined. The factors that are significantly
related to involvement in traffic crashes and have a low
correlation with DUI cannot be interpreted as DUI predic-
tors but as predictors of involvement in non-DUI crashes.
The results of the study show that DUI predictors differ
from the predictors of involvement in a crash. Considering
these factors in this light led to isolation of the partial con-
tribution of DUI.
It is important to emphasize the partial contribution of each
predictor for further development of the presented model of
DUI driver detection, as well as for the development of new
DUI prediction models. The future development of DUI pre-
diction models can include all or some of the predictors
studied in this article. Thus, it is significant to underline the
partial contribution of all factors examined in this article, in
order to make the development of future studies more efficient.
This study has shown the highest correlation between
drivers’ mistakes and DUI. For the whole population of
drivers, the score of the factor driving mistakes is 0.856
(maximum ¼ 1.000), which represents the most significant
DUI predictor. For non-DUI drivers, driving mistakes have
a smaller significance as a predictor (p  0.001). These
results indicate that DUI drivers make mistakes while driv-
ing more frequently than sober drivers.
Male gender is the second most significant predictor of
DUI for the whole population of drivers, which is in accord-
ance with the results of other studies (MacLeod et al. 2015;
Mann et al. 2010). The stated result does not refer to motor-
cyclists, whose chances of being under the influence of alco-
hol are approximately the same for male and female drivers.
Thus, male motorcyclists do not have a higher risk of DUI
than female drivers.
The score of the factor nonmandatory journey is signifi-
cantly lower than the previous 2 and is a risk factor with a
slight influence on DUI. In comparison to non-DUI crashes,
a nonmandatory journey is significantly related to DUI
crashes, which gives significance to it. In other words, the
specific importance of the nonmandatory journey for DUI is
present, but its influence on the whole population of drivers
is limited. These results indicate that drivers consume alcohol
more frequently outside of working hours, so its correlation
with DUI is significant. On the other hand, it is often difficult
to differentiate between a mandatory and a nonmandatory
journey, so its effect on the whole population is limited.
For commercial drivers, this factor is particularly import-
ant as a predictor, which can be a result of nonmandatory
driving or the change in behavior when professional drivers
are not at work. It is easier to recognize this type of driver
in commercial vehicle drivers, so the results of this study
are useful for identifying DUI drivers. A commercial vehicle
driver has an 80% higher odds of being under the influence
of alcohol during a nonmandatory journey in comparison to
when he or she is driving for business/work.
The correlation of other factors and DUI has been deter-
mined, but their partial contribution to DUI is debatable.
Studying non-alcohol-related crashes has shown that the
influence of other factors on the occurrence of traffic
crashes is higher than their partial contribution to DUI
crashes. These results can be explained by a greater influ-
ence of other factors on the occurrence of a crash than their

contribution to DUI. Apart from the fact that there is about
a 30% likelihood for a DUI driver to use a seat belt, which
is in accordance with the research in Norway (Bogstrand
et al. 2015), the particular significance for the prediction of
DUI is low. For example, the nonuse of a seat belt is the
most significant risk factor for being involved in traffic
crashes for drivers who are not under the influence of alco-
hol, though its correlation with DUI is significantly lower.
Thus, the nonuse of restraint systems cannot be considered
to be a predictor of DUI; this factor is related more to
involvement in crashes than to DUI.
The correlation of other factors and involvement in traf-
fic crashes is significantly lower than the nonuse of restraint
systems. Driver experience and age are the factors with a
significant prediction of involvement in a crash and the low-
est prediction of DUI. These results indicate that older and
younger drivers or drivers with less or more experience are
inclined to DUI. An increase in age among motorcyclists
and bus drivers leads to a slight increase in the risk of
involvement in crashes, whereas in other driver groups an
increase in age slightly decreases the risk of involvement in
crashes (about 0.5% per year).
The higher risk for motorcyclists and bus drivers with an
increase of age can be a result of decreased attentiveness
and psychomotor abilities, undertaking risky maneuvers
more frequently, etc.
Exposure to traffic is higher in urban areas where there is
hgih traffic flow, which leads to an expected higher risk of
being involved in a crash. This study shows that DUI driv-
ing in urban areas is not significantly more common than in
rural areas, which is in accordance with previous research
(Hobday and Meuleners 2018).
ORCID
Emir Smailovic http://orcid.org/0000-0001-9342-722X
Krsto Lipovac http://orcid.org/0000-0002-6207-9629
TRAFFIC INJURY PREVENTION 347
References
Asbridge M, Mann RE, Flam Zalcman R, Stoduto G. The criminaliza-
tion of impaired driving in Canada: assessing the deterrent impact
of Canada’s first per se law. J Stud Alcohol. 2004;65:450–459.
Bogstrand TS, Larsson M, Holtan A, Staff T, Vindenes V, Gjerde H.
Associations between driving under the influence of alcohol or
drugs. Speeding and seatbelt use among fatally injured car drivers in
Norway. Accid Anal Prev. 2015;78:14–19.
Compton RP, Berning A. Drug and Alcohol Crash Risk. Washington:
National Highway Traffic Safety; 2015. DOT HS 812 117.
Hobday M, Meuleners L. Alcohol and non-aclohol-related motor
vehicle crashes in Perth, Autralia: do alcohol outlets make a differ-
ence? Accid Anal Prev. 2018;113:117–124.
Li M, He Y, Zhou Z, et al. MicroRNA-223 ameliorates alcoholic liver
injury by inhibiting the IL-6-p47phox-oxidative stress pathway in
neutrophils. Gut. 2016.
MacLeod eK, Karriker-Jaffe JK, Ragland RR, Satariano AW, Kelley-
Baker T, Lacey HJ. Acceptance of drinking and driving and alcohol-
involved driving crashes in California. Accid Anal Prev. 2015;81:
134–142.
Mann RE, Stoduto G, Macdonald S, Shaikh A, Bondy S, Jonah B. The
effects of introducing or lowering legal per se blood alcohol limits
for driving: an international review. Accid Anal Prev. 2001;33:
61–75.
Mann ER, Stoduto G, Vingilis E, et al. Alcohol and driving factors in
collision risk. Accid Anal Prev. 2010;42:1538–1544.
Ogden EJ, Moskowitz H. Effects of alcohol and other drugs on driver
performance. Traffic Inj Prev. 2004;5:185–198.
Park HC, Kim DK, Kho SY, Park YP. Cross-classified multilevel mod-
els for severity of commercial motor vehicle crashes considering het-
erogeneity among companies and regions. Accid Anal Prev. 2017;
106:305–314.
Penning R, Veldstra JL, Daamen AP, Olivier B, Verster JC. Drugs of
abuse, driving and traffic safety. Curr Drug Abuse Rev. 2010;3:
23–32.
Soltani M, Seyedabrishami S, Mamdoohi A, Alyari Kordehdeh V.
Driver risk assessment model considering trip characteristics using
insurance data system. Procedia Eng. 2016;161:1160–1165.
World Health Organization. World Report on Traffic Injury Prevention.
2004. Available at: https://www.who.int/violence_injury_prevention/
publications/road_traffic/world_report/en/. Accessed February 5,
2017.