Accident Analysis and Prevention 75 (2015) 69–76
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Wrong-way driving crashes on French divided roads Emmanuel Kemel * CEREMA, DTerOuest/DIMER, MAN Avenue Viviani, 44000 Nantes, France
A R T I C L E I N F O
A B S T R A C T
Article history: Received 5 August 2014 Received in revised form 31 October 2014 Accepted 1 November 2014 Available online xxx
Context: The objective of divided roads is to increase users safety by posting unidirectional traffic flows. It happens however that drivers proceed in the wrong direction, endangering themselves as well as other users. The crashes caused by wrong-way drivers are generally spotlighted by the media and call for public intervention. Objectives: This paper proposes a characterization of wrong-way driving crashes occurring on French divided road on the 2008–2012 period. The objective is to identify the actors that delineate between wrong-way driving crashes and other crashes. Method: Building on the national injury road crash database, 266 crashes involving a wrong-way driver were identified. Their characteristics (related to timing, location, vehicle and driver) are compared to those of the 22,120 other crashes that occurred on the same roads over the same period. The comparison relies on descriptive statistics, completed by a logistic regression. Results: Wrong-way driving crashes are rare but severe. They are more likely to occur during night hours and on non-freeway roads than other crashes. Wrong-way drivers are older, more likely to be intoxicated, to be locals, to drive older vehicles, mainly passenger cars without passengers, than other drivers. Perspectives: The differences observed across networks can help prioritizing public intervention. Most of the identified WW-driving factors deal with cognitive impairment. Therefore, the specific countermeasures such as alternative road signs should be designed for and tested on cognitively impaired drivers. Nevertheless, WW-driving factors are also risk factors for other types of crashes (e.g. elderly driving, drunk driving and age of the vehicle). This suggest that, instead of (or in addition to) developing WW-driving specific countermeasures, managing these risk factors would help reducing a larger number of crashes. ã 2014 Elsevier Ltd. All rights reserved.
Keywords: Wrong-way driving Logistic regression Elderly driver Drunk driving
1. Introduction 1.1. Motivation Front-head collisions are the most severe and probably most feared road crashes. Divided roads, that impose an unidirectional traffic flow, have been developed with the primary objective to avoid this type of crash. Nevertheless drivers sometimes voluntarily or mistakenly proceed against the posted direction and drive the wrong-way (hereafter WW). The crashes caused by such WW drivers are also particularly disturbing as they are clear examples of failure from the driver and/or the infrastructure. A sizable grey literature, from local institutions and network managers, reports evaluations of wrong-way driving countermeasures. These reports are generally introduced by descriptive
* Tel.: +33 240 128519. E-mail address: [email protected] (E. Kemel). http://dx.doi.org/10.1016/j.aap.2014.11.002 0001-4575/ ã 2014 Elsevier Ltd. All rights reserved.
statistics on WW-driving crashes. Even if WW-driving crashes are rare, the literature shows the interest of divided-road managers for this issue. This may be because, by their severity, these crashes shock the public opinion and are particularly spotlighted by the media (Cooner and Ranft, 2008). They are also particularly unacceptable because divided-road are especially designed to avoid vehicles traveling against the traffic flow. Despite the human factors involved, a WW-driving crash can thus be perceived as a failure from the infrastructure and call for public intervention. In France the annual number of road fatalities has decreased from 8170 to 3653 between 2002 and 2012. Additional gains become harder and harder to obtain, and new niches of traffic safety are targeted by the authorities: WW-driving crashes are one of them. Avoiding WW-driving crashes can be achieved by preventing wrong-way entries or by reducing crash risk when they occur. The former countermeasures essentially consist of modifications of infrastructure’s geometry or signage. The later consist of patrols or electronic devices that detect WW drivers and radio or dynamic-
70
E. Kemel / Accident Analysis and Prevention 75 (2015) 69–76
sign messages that inform the violator and/or regular users of the risk (Scaramuzza and Cavegn, 2007; Vicedo, 2006; Cooner and Ranft, 2008; SETRA 2008). These countermeasures share the common characteristics that they are often expensive and their impact can be uncertain (Shepard, 1976; Carnis and Kemel, 2014), hence the need to characterize WW-driving crashes in order to choose suitable countermeasures. 1.2. Literature review Aside from the grey literature, the research literature is particularly scarce. To our knowledge, the paper by Lathrop et al. (2010) is the only one that provides a detailed statistical analysis of fatal WW-driving crashes. Their results are consistent with the patterns reported by the grey literature. In terms of prevalence, WW-driving crashes are rare but severe. In early 2000, they caused the death of 350 people per year in the US (Moler, 2002). The same figure was reported more recently by Zhou et al. (2014) for 2013. For instance, 6.7% of fatalities on New Mexico’s interstate highways are due to WW-driving crashes (Lathrop et al., 2010). According to the review of several technical reports proposed by Scaramuzza and Cavegn (2007), they represent 0.03% of crashes but between 2.6% and 6% of fatalities in Switzerland. In France, WW drivers are involved in 1% of injury crashes but 4% of fatalities (SETRA, 2008). The literature shows a consensus about the timing of these crashes and several characteristics of their authors (Scaramuzza and Cavegn, 2007; Vicedo, 2006; Cooner and Ranft, 2008; Lathrop et al., 2010; Zhou et al., 2014). WW-driving crashes generally occur during night hours. Their author is often an elderly or drunk driver. According to Vicedo (2006), WW drivers are also characterized by particular psychological profiles such as individuals with suicidal tendencies, or under the influence of drugs. The descriptive statistics reported by the grey literature are informative but do not always allow to identify predictors. First, these studies rarely report statistical tests, which limits the possibility to draw conclusions about patterns observed on small sample sizes. Second, these studies generally describe WW-driving crashes but do not refer to comparison groups. For instance, Cooner and Ranft (2008) and Zhou et al. (2014) report that most WW drivers are male. However Lathrop et al. (2010) show that this also applies to other crashes and that men are not over-represented in WW-driving crashes. Lathrop et al. (2010) are also the authors that investigate the largest set of possible WW-driving predictors, including presence of licit or illicit drug, vehicle type or seat-belt use. However, they focused their study on fatal crashes occurring on New Mexico interstate highways. Doing so they base their statistical analysis on 49 crashes only. Because WW-driving crashes are rare, working on a larger database may be suitable. This can be achieved by including non-fatal crashes in the analysis, and working at a national level. This later aspect would also allow to compare road types, which was rarely done in the literature. 1.3. Objective of the paper This paper presents a statistical analysis of WW-driving crashes that occurred in France between 2008 and 2012. Among the 22,386 crashes recorded during this period on divided roads, 266 involved a WW-driving vehicle. Besides timing and location of the crashes, the data offer information about users and their vehicles. The objective of the paper is to identify factors associated with WW-driving crashes, by comparing their characteristics with those of other crashes. Extending previous characterizations of WW-driving crashes from a larger database is the first contribution of the paper. The
database used in this study also allows to investigate new potential factors such as presence of passengers or other driver characteristics: novice or experienced, local or transit drivers. This investigation constitutes the second contribution. Like in Lathrop et al. (2010), all the contributing factors are assessed separately by statistical tests. Because several factors may be correlated, a logistic model is also estimated. This is the third contribution of the paper to the WW-driving literature. The next section presents the database that was used for the analysis. The results of the statistical analysis are reported in Section 3. The section starts with descriptive statistics and the related inference tests. The results of the logistic regression are then reported. Section 4 synthesizes the main results, discusses their limitation and addresses their implication for policy making and further research. 2. Method 2.1. Data The statistical data source is the French database of road injury crashes reported by police authorities (namely BAAC for Bulletin d’Analyse des Crashes Corporels). A road injury crash is defined as a crash involving at least one motorized vehicle traveling a public road and causing at least one injured victim (dead or not). For each of these crashes, an extensive paper form is filled by police officers on the crash scene. The forms are then recorded on computer files and made anonymous. This database is used for establishing the national overview of traffic safety every year (ONISR, 2012) and is also the main basis for French traffic safety studies (Lenguerrand et al., 2006; Amoros et al., 2007; de Lapparent, 2008; Moskal et al., 2012; Carnis and Blais, 2013). The data contains four parts, related respectively to the timing of the crash (day and hour), its location (road type), the vehicles involved (type and age), and the driver and passengers of each vehicle. This last set of variables is of prime interest for our analysis. For each person involved, a variable describe the consequences of the crash: unscathed, light injury without hospitalization, heavy injury requiring hospitalization and death within 30 days. Other variables deal with gender, age and mandatory safety-equipment use (seat-belt or helmet). For drivers, additional variables are available, such as the year the driving license was received, allowing to distinguish experienced from novice drivers1 and to identify drivers without a valid driving license. The license variable also contains a level labeled “school”, that refers to situations where the driver was taking a driving lesson when the crash occurred. Because the county of residence is also reported, a distinction can be made between local drivers (from the county) and transit drivers (from other counties). Regarding human factors, blood alcohol concentration (BAC) level is reported when over the legal limit of 0.5 g/l. For fatal crashes, drivers are also tested for illegal drugs and the data reports a positive test when the presence of at least one illegal drug is detected, without further distinction. Like most of the previous WW-driving studies, the present analysis focuses on non-urban divided roads. As reminded by an anonymous referee, wrong-way driving can also occur on urban roads. In France, urban and non-urban road safety issues are thus considered as different topics and, as far as wrong-way driving is concerned, the focus is generally set on non-urban divided roads. This is because urban and non urban roads have different characteristics. Speed limits are lower on urban roads. In terms
1
In France, drivers are considered novice for two years after receiving their license.
E. Kemel / Accident Analysis and Prevention 75 (2015) 69–76
71
Table 1 Severity of WW-driving and other crashes. Variable
Level
% Among WW-driving crashes (n = 266)
% Among other crashes (n = 22,120)
p-value
OR
Fatal crash
No Yes One Two Three and more
73.3 26.7 2.2 64.3 33.5
94.2 5.8 37.9 44.8 17.3
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Wrong-way driving crashes on French divided roads Emmanuel Kemel * CEREMA, DTerOuest/DIMER, MAN Avenue Viviani, 44000 Nantes, France
A R T I C L E I N F O
A B S T R A C T
Article history: Received 5 August 2014 Received in revised form 31 October 2014 Accepted 1 November 2014 Available online xxx
Context: The objective of divided roads is to increase users safety by posting unidirectional traffic flows. It happens however that drivers proceed in the wrong direction, endangering themselves as well as other users. The crashes caused by wrong-way drivers are generally spotlighted by the media and call for public intervention. Objectives: This paper proposes a characterization of wrong-way driving crashes occurring on French divided road on the 2008–2012 period. The objective is to identify the actors that delineate between wrong-way driving crashes and other crashes. Method: Building on the national injury road crash database, 266 crashes involving a wrong-way driver were identified. Their characteristics (related to timing, location, vehicle and driver) are compared to those of the 22,120 other crashes that occurred on the same roads over the same period. The comparison relies on descriptive statistics, completed by a logistic regression. Results: Wrong-way driving crashes are rare but severe. They are more likely to occur during night hours and on non-freeway roads than other crashes. Wrong-way drivers are older, more likely to be intoxicated, to be locals, to drive older vehicles, mainly passenger cars without passengers, than other drivers. Perspectives: The differences observed across networks can help prioritizing public intervention. Most of the identified WW-driving factors deal with cognitive impairment. Therefore, the specific countermeasures such as alternative road signs should be designed for and tested on cognitively impaired drivers. Nevertheless, WW-driving factors are also risk factors for other types of crashes (e.g. elderly driving, drunk driving and age of the vehicle). This suggest that, instead of (or in addition to) developing WW-driving specific countermeasures, managing these risk factors would help reducing a larger number of crashes. ã 2014 Elsevier Ltd. All rights reserved.
Keywords: Wrong-way driving Logistic regression Elderly driver Drunk driving
1. Introduction 1.1. Motivation Front-head collisions are the most severe and probably most feared road crashes. Divided roads, that impose an unidirectional traffic flow, have been developed with the primary objective to avoid this type of crash. Nevertheless drivers sometimes voluntarily or mistakenly proceed against the posted direction and drive the wrong-way (hereafter WW). The crashes caused by such WW drivers are also particularly disturbing as they are clear examples of failure from the driver and/or the infrastructure. A sizable grey literature, from local institutions and network managers, reports evaluations of wrong-way driving countermeasures. These reports are generally introduced by descriptive
* Tel.: +33 240 128519. E-mail address: [email protected] (E. Kemel). http://dx.doi.org/10.1016/j.aap.2014.11.002 0001-4575/ ã 2014 Elsevier Ltd. All rights reserved.
statistics on WW-driving crashes. Even if WW-driving crashes are rare, the literature shows the interest of divided-road managers for this issue. This may be because, by their severity, these crashes shock the public opinion and are particularly spotlighted by the media (Cooner and Ranft, 2008). They are also particularly unacceptable because divided-road are especially designed to avoid vehicles traveling against the traffic flow. Despite the human factors involved, a WW-driving crash can thus be perceived as a failure from the infrastructure and call for public intervention. In France the annual number of road fatalities has decreased from 8170 to 3653 between 2002 and 2012. Additional gains become harder and harder to obtain, and new niches of traffic safety are targeted by the authorities: WW-driving crashes are one of them. Avoiding WW-driving crashes can be achieved by preventing wrong-way entries or by reducing crash risk when they occur. The former countermeasures essentially consist of modifications of infrastructure’s geometry or signage. The later consist of patrols or electronic devices that detect WW drivers and radio or dynamic-
70
E. Kemel / Accident Analysis and Prevention 75 (2015) 69–76
sign messages that inform the violator and/or regular users of the risk (Scaramuzza and Cavegn, 2007; Vicedo, 2006; Cooner and Ranft, 2008; SETRA 2008). These countermeasures share the common characteristics that they are often expensive and their impact can be uncertain (Shepard, 1976; Carnis and Kemel, 2014), hence the need to characterize WW-driving crashes in order to choose suitable countermeasures. 1.2. Literature review Aside from the grey literature, the research literature is particularly scarce. To our knowledge, the paper by Lathrop et al. (2010) is the only one that provides a detailed statistical analysis of fatal WW-driving crashes. Their results are consistent with the patterns reported by the grey literature. In terms of prevalence, WW-driving crashes are rare but severe. In early 2000, they caused the death of 350 people per year in the US (Moler, 2002). The same figure was reported more recently by Zhou et al. (2014) for 2013. For instance, 6.7% of fatalities on New Mexico’s interstate highways are due to WW-driving crashes (Lathrop et al., 2010). According to the review of several technical reports proposed by Scaramuzza and Cavegn (2007), they represent 0.03% of crashes but between 2.6% and 6% of fatalities in Switzerland. In France, WW drivers are involved in 1% of injury crashes but 4% of fatalities (SETRA, 2008). The literature shows a consensus about the timing of these crashes and several characteristics of their authors (Scaramuzza and Cavegn, 2007; Vicedo, 2006; Cooner and Ranft, 2008; Lathrop et al., 2010; Zhou et al., 2014). WW-driving crashes generally occur during night hours. Their author is often an elderly or drunk driver. According to Vicedo (2006), WW drivers are also characterized by particular psychological profiles such as individuals with suicidal tendencies, or under the influence of drugs. The descriptive statistics reported by the grey literature are informative but do not always allow to identify predictors. First, these studies rarely report statistical tests, which limits the possibility to draw conclusions about patterns observed on small sample sizes. Second, these studies generally describe WW-driving crashes but do not refer to comparison groups. For instance, Cooner and Ranft (2008) and Zhou et al. (2014) report that most WW drivers are male. However Lathrop et al. (2010) show that this also applies to other crashes and that men are not over-represented in WW-driving crashes. Lathrop et al. (2010) are also the authors that investigate the largest set of possible WW-driving predictors, including presence of licit or illicit drug, vehicle type or seat-belt use. However, they focused their study on fatal crashes occurring on New Mexico interstate highways. Doing so they base their statistical analysis on 49 crashes only. Because WW-driving crashes are rare, working on a larger database may be suitable. This can be achieved by including non-fatal crashes in the analysis, and working at a national level. This later aspect would also allow to compare road types, which was rarely done in the literature. 1.3. Objective of the paper This paper presents a statistical analysis of WW-driving crashes that occurred in France between 2008 and 2012. Among the 22,386 crashes recorded during this period on divided roads, 266 involved a WW-driving vehicle. Besides timing and location of the crashes, the data offer information about users and their vehicles. The objective of the paper is to identify factors associated with WW-driving crashes, by comparing their characteristics with those of other crashes. Extending previous characterizations of WW-driving crashes from a larger database is the first contribution of the paper. The
database used in this study also allows to investigate new potential factors such as presence of passengers or other driver characteristics: novice or experienced, local or transit drivers. This investigation constitutes the second contribution. Like in Lathrop et al. (2010), all the contributing factors are assessed separately by statistical tests. Because several factors may be correlated, a logistic model is also estimated. This is the third contribution of the paper to the WW-driving literature. The next section presents the database that was used for the analysis. The results of the statistical analysis are reported in Section 3. The section starts with descriptive statistics and the related inference tests. The results of the logistic regression are then reported. Section 4 synthesizes the main results, discusses their limitation and addresses their implication for policy making and further research. 2. Method 2.1. Data The statistical data source is the French database of road injury crashes reported by police authorities (namely BAAC for Bulletin d’Analyse des Crashes Corporels). A road injury crash is defined as a crash involving at least one motorized vehicle traveling a public road and causing at least one injured victim (dead or not). For each of these crashes, an extensive paper form is filled by police officers on the crash scene. The forms are then recorded on computer files and made anonymous. This database is used for establishing the national overview of traffic safety every year (ONISR, 2012) and is also the main basis for French traffic safety studies (Lenguerrand et al., 2006; Amoros et al., 2007; de Lapparent, 2008; Moskal et al., 2012; Carnis and Blais, 2013). The data contains four parts, related respectively to the timing of the crash (day and hour), its location (road type), the vehicles involved (type and age), and the driver and passengers of each vehicle. This last set of variables is of prime interest for our analysis. For each person involved, a variable describe the consequences of the crash: unscathed, light injury without hospitalization, heavy injury requiring hospitalization and death within 30 days. Other variables deal with gender, age and mandatory safety-equipment use (seat-belt or helmet). For drivers, additional variables are available, such as the year the driving license was received, allowing to distinguish experienced from novice drivers1 and to identify drivers without a valid driving license. The license variable also contains a level labeled “school”, that refers to situations where the driver was taking a driving lesson when the crash occurred. Because the county of residence is also reported, a distinction can be made between local drivers (from the county) and transit drivers (from other counties). Regarding human factors, blood alcohol concentration (BAC) level is reported when over the legal limit of 0.5 g/l. For fatal crashes, drivers are also tested for illegal drugs and the data reports a positive test when the presence of at least one illegal drug is detected, without further distinction. Like most of the previous WW-driving studies, the present analysis focuses on non-urban divided roads. As reminded by an anonymous referee, wrong-way driving can also occur on urban roads. In France, urban and non-urban road safety issues are thus considered as different topics and, as far as wrong-way driving is concerned, the focus is generally set on non-urban divided roads. This is because urban and non urban roads have different characteristics. Speed limits are lower on urban roads. In terms
1
In France, drivers are considered novice for two years after receiving their license.
E. Kemel / Accident Analysis and Prevention 75 (2015) 69–76
71
Table 1 Severity of WW-driving and other crashes. Variable
Level
% Among WW-driving crashes (n = 266)
% Among other crashes (n = 22,120)
p-value
OR
Fatal crash
No Yes One Two Three and more
73.3 26.7 2.2 64.3 33.5
94.2 5.8 37.9 44.8 17.3
