This article was downloaded by: [New York University] On: 11 July 2015, At: 06:36 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick Place, London, SW1P 1WG
International Journal of Injury Control and Safety Promotion Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/nics20
Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach a
b
b
Harold B. Weiss , Sigal Kaplan & Carlo Giacomo Prato a
Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand
b
Department of Transport, Technical University of Denmark, Kgs. Lyngby, Denmark Published online: 06 Jul 2015.
Click for updates To cite this article: Harold B. Weiss, Sigal Kaplan & Carlo Giacomo Prato (2015): Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach, International Journal of Injury Control and Safety Promotion, DOI: 10.1080/17457300.2015.1056807 To link to this article: http://dx.doi.org/10.1080/17457300.2015.1056807
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
International Journal of Injury Control and Safety Promotion, 2015 http://dx.doi.org/10.1080/17457300.2015.1056807
Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach Harold B. Weissa*, Sigal Kaplanb and Carlo Giacomo Pratob a
Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand; bDepartment of Transport, Technical University of Denmark, Kgs. Lyngby, Denmark
Downloaded by [New York University] at 06:36 11 July 2015
(Received 19 March 2014; accepted 14 May 2015) The over-representation of young drivers in road crashes remains an important concern worldwide. Cluster analysis has been applied to young driver sub-groups, but its application by analysing crash occurrence is just emerging. We present a classification analysis that advances the field through a holistic overview of crash patterns useful for designing youthtargeted road safety programmes. We compiled a database of 8644 New Zealand crashes from 2002 to 2011 involving at least one 1524-year-old driver and a fatal or serious injury for at least one road user. We considered crash location, infrastructure characteristics, environmental conditions, demographic characteristics, driving behaviour, and pre-crash manoeuvres. The analysis yielded 15 and 8 latent classes of, respectively, single-vehicle and multi-vehicle crashes, and average posterior probabilities measured the odds of correct classification that revealed how the identified clusters contain mostly crashes of a particular class and all the crashes of that class. The results raised three major safety concerns for young drivers that should be addressed: (1) reckless driving and traffic law violations; (2) inattention, error, and hazard perception problems; and (3) interaction with road geometry and lighting conditions, especially on high-speed open roads and state highways. Keywords: road crashes; young driver problem; clustering analysis; latent class analysis
1. Introduction A major ongoing global public health concern is the overrepresentation of young drivers in road crashes and injuries (Twisk & Stacey, 2007; Williams, 2003). Graduated Driver Licensing (GDL) systems aim at increasing the driving experience, while reducing the crash risk. While the initial effectiveness of GDL programmes has been validated (Begg & Stephenson, 2003; Hedlund & Compton, 2005; Williams & Shults, 2010; Williams, Tefft, & Grabowski, 2012), road crashes still remain a leading cause of death among young people. In New Zealand, in the decade following the implementation of the GDL programme, a 50% reduction was observed in the fatal and serious injury rate of 1524-year-old car occupants (Begg & Stephenson, 2003), but road crashes remained the leading cause of death with 42.6 fatalities per 100,000 person years (Kypri, Chalmers, & Langley, 2002). Recent suggestions to improve GDL systems have moved in two main directions: (1) focusing on the ‘problem young driver’ by targeting high-risk young driver groups (Scott-Parker, Watson, King, & Hyde, 2013) with the aim of designing better youth-oriented road safety programmes; (2) addressing the ‘young driver problem’ by exploring crash occurrence patterns *Corresponding author. Email: [email protected] Ó 2015 Taylor & Francis
(Hasselberg & Laflamme, 2009) with the aim of targeting hazardous situations with focused solutions, for example, the hazard perception tests conducted in Australia, New Zealand, and the UK (Scialfa, Borkenhagen, Lyon, & Desch^enes, 2013). Researchers investigated the ‘problem young driver’ by identifying young drivers’ sub-groups from self-reported questionnaires exploring risk perceptions and attitudes towards traffic violations, risky driving behaviour, and crash involvement (Brandau, Daghofer, Hofmann, & Spitzer, 2011; Lucidi et al., 2010; Marengo, Settanni, & Vidotto, 2012; Scott-Parker et al., 2013). Other studies explored the factors contributing to the high crash risk of young drivers. In New Zealand, the research focus has been on gender (Lewis-Evans, 2010; Reeder et al., 1998), substance abuse (Begg et al., 2009; Kypri et al., 2006), cell phone use (Begg et al., 2009; Hallett, Lambert, & Regan, 2011), and risk-taking behaviour (Blows et al., 2005; Ivers et al., 2009) from the personal perspective, and on parental influence (Brookland, Begg, Langley, & Ameratunga, 2010; Brookland & Begg, 2011), peer pressure (Keall, Frith, & Patterson, 2004; Lam et al., 2003), and night-time driving (Keall et al., 2004) from the social perspective. These studies examined the crash risk by analysing self-report questionnaires
Downloaded by [New York University] at 06:36 11 July 2015
2
H.B. Weiss et al.
regarding driving habits, attitudes, and crash involvement of young drivers. In contrast, exploring the ‘young driver problem’ by analysing young drivers’ actual crash patterns is still in its nascent stage (Hasselberg & Laflamme, 2009), despite its usefulness for formulating crash scenarios on the basis of actual crash data (Lenard & Danton, 2010) and prioritizing safety issues and preventive measures (Kaplan & Prato, 2013; Prato, Gitelman, & Bekhor, 2012; Theofilatos & Efthymiou, 2012). We provide a holistic overview of multidimensional crash patterns that can help discern the underlying circumstances and crash characteristics of both the ‘young driver problem’ and the ‘problem young driver’. This holistic overview is important for prioritizing preventive measures, matching law enforcement efforts to potentially risky circumstances, and improving driver education, training, and GDL programmes. In order to achieve this overview, we analysed 8644 crashes that occurred in New Zealand in the 10-year period from 2002 to 2011, involved at least one 1524-year-old driver, and resulted in a fatality or serious injury for at least one road user. We compiled data provided by the Land Transport Safety Authority (LTSA) from police reports containing information regarding crash location, infrastructure characteristics, light and weather conditions, demographic characteristics, risky driving behaviours, and pre-crash manoeuvres of the young drivers. We performed latent class analysis (LCA) to cluster the crashes, and we interpreted the results with the aim of highlighting the most relevant issues within the ‘young driver problem’. The remainder of this paper is structured as follows. The next section summarizes the data and the applied methodology. Then, clustering results for single-vehicle and multi-vehicle crashes are presented and discussed separately. Finally, conclusions are drawn and further research directions are proposed. 2. Methods 2.1. Data The analysed LTSA data-set contains police reports of crashes involving at least one 1524-year-old driver occurring in New Zealand between 2002 and 2011 that resulted in serious injuries involving hospitalization (e.g., concussions, internal injuries, fractures, lacerations) or fatalities within 30 days of the crash occurrence. The database contains details regarding the persons injured in the crash, including age, gender, alcohol or drug intoxication, restraint use, licence type, licence validity, role in the crash, and injury severity. For motor-vehicle occupants, the database specifies vehicle type, registration date, precrash manoeuvre, and collision point. Additional information comprises crash circumstances including crash type, day and time of day, severity level, collision manner, vehicles and road users, infrastructure characteristics (e.g., curvature, road surface conditions, speed limits), illumination, and weather conditions.
We analysed 8644 records of relevant crashes whose characteristics are reported in Table 1. The characteristics were collected by the police officers who compiled the reports, and the quality of the reports was then verified to allow LTSA to release the reports for research purposes. Although data were presented for characteristics objectively difficult to assess (e.g., distraction, fatigue), it is assumed that the quality control guarantees that police officers had sufficient evidence, thus limiting the potential bias. Among the crashes, 86.0% resulted in serious injuries and 14.0% in at least one fatality. Over two-thirds of the young drivers were male, and their age distribution was skewed towards younger age under 20 years old. The vast majority of young drivers drove passenger cars with vehicle age generally over 10 years, and were assigned at least partial fault in over four out of five crashes. Intoxication was verified or suspected in almost one out of four cases, reckless driving and speeding were recorded in almost half of the cases, and the presence of passengers was reported in one-third of the cases. 2.2.
Model
We performed crash pattern recognition by applying LCA, also known as model-based clustering or finite-mixture clustering. We selected LCA over alternative clustering methodologies because of the existence of an underlying statistical model that allows computing of cluster probabilities for new records, the ability to represent similarity across clusters, and the availability of goodness-of-fit criteria that facilitate objective decisions regarding the number of clusters (Depaire, Wets, & Vanhoof, 2008; Magidson & Vermunt, 2002). These advantages have been found beneficial in uncovering crash patterns in Belgium and Denmark (Depaire et al., 2008; Kaplan & Prato, 2013). LCA fostered the classification of road crashes into C latent classes, with uncertain class membership and unknown cluster size prior to the analysis. Consider a vector of N road crashes characterized by a vector of M variables (yi D y1, … , yM), and let Yi (Yi D Yi1, … , YiM) be the vector of values of crash n for the M variables. The LCA model is formulated as follows (Lanza, Collins, Lemmon, & Schafer, 2007): pðYi j uÞ D
K X
PðCk ÞpðYi j Ck ; uk Þ
(1)
kD1
where k is an indicator of latent class Ck, K is the number of latent classes, P(Ck) is the prevalence of latent class Ck in the data-set, p(YijCk,uk) is the conditional multivariate probability that a crash in class Ck would be characterized by vector Yi, and uk is a vector of parameters to be estimated. We made simplifying assumptions in order to derive an estimable LCA model with reasonable parametric
International Journal of Injury Control and Safety Promotion
3
Table 1. Sample characteristics. Variable
Categories
Per cent
Categories
Serious injury
86.0%
Fatal injury
14.0%
15 16 17 18 19
3.8% 8.2% 12.4% 13.5% 13.3%
20 21 22 23 24
12.2% 11.0% 9.5% 8.6% 7.6%
Male
68.5%
Female
31.5%
Young driver ethnicity
European Maori
57.9% 19.8%
Asian Other Unknown
5.9% 1.9% 14.6%
Young driver licence
Full Restricted
37.6% 34.2%
Learner Not valid
15.5% 12.7%
Young driver vehicle
SUV Car
3.7% 86.9%
Van Heavy vehicle
7.7% 1.6%
Less than 2 years 25 years
1.1% 2.8%
510 years More than 10 years
18.2% 72.6%
Young driver fault
Yes
83.8%
No
16.2%
Young driver intoxication
Yes
23.2%
No
76.8%
Young driver seat belt use
Yes
55.0%
No Unknown
8.7% 36.3%
Young driver manoeuvre
Traveling Lane changing Lost control
27.0% 5.4% 39.4%
Turning Parking/reversing Avoiding pedestrian None
17.0% 3.5% 6.8% 0.8%
Reckless driving Speeding Distraction
22.9% 17.5% 22.6%
Fatigue Inexperience None
6.5% 13.2% 36.0%
Young driver passenger age
No passengers Only young
67.8% 26.2%
Only adult Young and adult
4.9% 1.1%
Young driver passenger gender
No passengers Only male
67.8% 16.0%
Only female Male and female
5.6% 10.6%
Young driver passenger ethnicity
No passengers Same ethnicity
67.8% 24.2%
Different ethnicity
8.0%
No object Road Side
53.3% 9.3% 14.2%
Vehicle not moving Construction Animal
3.9% 19.0% 0.3%
Crash injury severity Young driver age (years)
Downloaded by [New York University] at 06:36 11 July 2015
Young driver gender
Young driver vehicle age
Young driver factors
Object hit
Per cent
(Continued)
4
H.B. Weiss et al.
Table 1. (Continued ) Variable
Categories
Per cent
Categories
Vulnerable road user involved
Pedestrian Cyclist
8.1% 3.5%
Motorcyclist/moped None
6.4% 82.0%
Single-vehicle
56.4%
Multiple-vehicle
43.6%
1524 years old 2564 years old
23.2% 66.8%
65 years old or more
10.0%
Male
62.4%
Female
37.6%
Same ethnicity
45.5%
Different ethnicity
54.5%
Other driver licencea
Full Restricted
76.8% 9.9%
Learner Not valid
4.3% 9.0%
Other driver vehiclea Other driver intoxicationa
SUV Car Yes
7.6% 74.6% 7.6%
Van Heavy vehicle No
12.2% 5.7% 92.4%
Other driver seat belt usea
Yes
71.1%
Travelling Lane changing Lost control
51.5% 8.5% 4.8%
No Unknown Turning Parking/reversing Avoiding pedestrian None
4.3% 24.6% 29.3% 5.0% 0.5% 0.5%
Reckless driving Speeding Distraction
22.2% 7.8% 15.5%
Fatigue Inexperience None
0.6% 1.9% 52.0%
Yes
43.5%
No
56.5%
Road type
Motorway State highway Open road
3.6% 25.6% 23.3%
Major urban road Minor urban road
25.2% 22.2%
Speed limit
50 km/h or less 5070 km/h One One Straight Easy curve
45.5% 8.3% 4.0% 82.2% 57.1% 19.3%
7090 km/h 90 km/h or more Two Two or more Moderate curve Severe curve
5.8% 40.5% 96.0% 17.8% 19.6% 3.9%
Level
76.4%
Hilly
23.6%
Intersection Section (
International Journal of Injury Control and Safety Promotion Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/nics20
Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach a
b
b
Harold B. Weiss , Sigal Kaplan & Carlo Giacomo Prato a
Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand
b
Department of Transport, Technical University of Denmark, Kgs. Lyngby, Denmark Published online: 06 Jul 2015.
Click for updates To cite this article: Harold B. Weiss, Sigal Kaplan & Carlo Giacomo Prato (2015): Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach, International Journal of Injury Control and Safety Promotion, DOI: 10.1080/17457300.2015.1056807 To link to this article: http://dx.doi.org/10.1080/17457300.2015.1056807
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
International Journal of Injury Control and Safety Promotion, 2015 http://dx.doi.org/10.1080/17457300.2015.1056807
Fatal and serious road crashes involving young New Zealand drivers: a latent class clustering approach Harold B. Weissa*, Sigal Kaplanb and Carlo Giacomo Pratob a
Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand; bDepartment of Transport, Technical University of Denmark, Kgs. Lyngby, Denmark
Downloaded by [New York University] at 06:36 11 July 2015
(Received 19 March 2014; accepted 14 May 2015) The over-representation of young drivers in road crashes remains an important concern worldwide. Cluster analysis has been applied to young driver sub-groups, but its application by analysing crash occurrence is just emerging. We present a classification analysis that advances the field through a holistic overview of crash patterns useful for designing youthtargeted road safety programmes. We compiled a database of 8644 New Zealand crashes from 2002 to 2011 involving at least one 1524-year-old driver and a fatal or serious injury for at least one road user. We considered crash location, infrastructure characteristics, environmental conditions, demographic characteristics, driving behaviour, and pre-crash manoeuvres. The analysis yielded 15 and 8 latent classes of, respectively, single-vehicle and multi-vehicle crashes, and average posterior probabilities measured the odds of correct classification that revealed how the identified clusters contain mostly crashes of a particular class and all the crashes of that class. The results raised three major safety concerns for young drivers that should be addressed: (1) reckless driving and traffic law violations; (2) inattention, error, and hazard perception problems; and (3) interaction with road geometry and lighting conditions, especially on high-speed open roads and state highways. Keywords: road crashes; young driver problem; clustering analysis; latent class analysis
1. Introduction A major ongoing global public health concern is the overrepresentation of young drivers in road crashes and injuries (Twisk & Stacey, 2007; Williams, 2003). Graduated Driver Licensing (GDL) systems aim at increasing the driving experience, while reducing the crash risk. While the initial effectiveness of GDL programmes has been validated (Begg & Stephenson, 2003; Hedlund & Compton, 2005; Williams & Shults, 2010; Williams, Tefft, & Grabowski, 2012), road crashes still remain a leading cause of death among young people. In New Zealand, in the decade following the implementation of the GDL programme, a 50% reduction was observed in the fatal and serious injury rate of 1524-year-old car occupants (Begg & Stephenson, 2003), but road crashes remained the leading cause of death with 42.6 fatalities per 100,000 person years (Kypri, Chalmers, & Langley, 2002). Recent suggestions to improve GDL systems have moved in two main directions: (1) focusing on the ‘problem young driver’ by targeting high-risk young driver groups (Scott-Parker, Watson, King, & Hyde, 2013) with the aim of designing better youth-oriented road safety programmes; (2) addressing the ‘young driver problem’ by exploring crash occurrence patterns *Corresponding author. Email: [email protected] Ó 2015 Taylor & Francis
(Hasselberg & Laflamme, 2009) with the aim of targeting hazardous situations with focused solutions, for example, the hazard perception tests conducted in Australia, New Zealand, and the UK (Scialfa, Borkenhagen, Lyon, & Desch^enes, 2013). Researchers investigated the ‘problem young driver’ by identifying young drivers’ sub-groups from self-reported questionnaires exploring risk perceptions and attitudes towards traffic violations, risky driving behaviour, and crash involvement (Brandau, Daghofer, Hofmann, & Spitzer, 2011; Lucidi et al., 2010; Marengo, Settanni, & Vidotto, 2012; Scott-Parker et al., 2013). Other studies explored the factors contributing to the high crash risk of young drivers. In New Zealand, the research focus has been on gender (Lewis-Evans, 2010; Reeder et al., 1998), substance abuse (Begg et al., 2009; Kypri et al., 2006), cell phone use (Begg et al., 2009; Hallett, Lambert, & Regan, 2011), and risk-taking behaviour (Blows et al., 2005; Ivers et al., 2009) from the personal perspective, and on parental influence (Brookland, Begg, Langley, & Ameratunga, 2010; Brookland & Begg, 2011), peer pressure (Keall, Frith, & Patterson, 2004; Lam et al., 2003), and night-time driving (Keall et al., 2004) from the social perspective. These studies examined the crash risk by analysing self-report questionnaires
Downloaded by [New York University] at 06:36 11 July 2015
2
H.B. Weiss et al.
regarding driving habits, attitudes, and crash involvement of young drivers. In contrast, exploring the ‘young driver problem’ by analysing young drivers’ actual crash patterns is still in its nascent stage (Hasselberg & Laflamme, 2009), despite its usefulness for formulating crash scenarios on the basis of actual crash data (Lenard & Danton, 2010) and prioritizing safety issues and preventive measures (Kaplan & Prato, 2013; Prato, Gitelman, & Bekhor, 2012; Theofilatos & Efthymiou, 2012). We provide a holistic overview of multidimensional crash patterns that can help discern the underlying circumstances and crash characteristics of both the ‘young driver problem’ and the ‘problem young driver’. This holistic overview is important for prioritizing preventive measures, matching law enforcement efforts to potentially risky circumstances, and improving driver education, training, and GDL programmes. In order to achieve this overview, we analysed 8644 crashes that occurred in New Zealand in the 10-year period from 2002 to 2011, involved at least one 1524-year-old driver, and resulted in a fatality or serious injury for at least one road user. We compiled data provided by the Land Transport Safety Authority (LTSA) from police reports containing information regarding crash location, infrastructure characteristics, light and weather conditions, demographic characteristics, risky driving behaviours, and pre-crash manoeuvres of the young drivers. We performed latent class analysis (LCA) to cluster the crashes, and we interpreted the results with the aim of highlighting the most relevant issues within the ‘young driver problem’. The remainder of this paper is structured as follows. The next section summarizes the data and the applied methodology. Then, clustering results for single-vehicle and multi-vehicle crashes are presented and discussed separately. Finally, conclusions are drawn and further research directions are proposed. 2. Methods 2.1. Data The analysed LTSA data-set contains police reports of crashes involving at least one 1524-year-old driver occurring in New Zealand between 2002 and 2011 that resulted in serious injuries involving hospitalization (e.g., concussions, internal injuries, fractures, lacerations) or fatalities within 30 days of the crash occurrence. The database contains details regarding the persons injured in the crash, including age, gender, alcohol or drug intoxication, restraint use, licence type, licence validity, role in the crash, and injury severity. For motor-vehicle occupants, the database specifies vehicle type, registration date, precrash manoeuvre, and collision point. Additional information comprises crash circumstances including crash type, day and time of day, severity level, collision manner, vehicles and road users, infrastructure characteristics (e.g., curvature, road surface conditions, speed limits), illumination, and weather conditions.
We analysed 8644 records of relevant crashes whose characteristics are reported in Table 1. The characteristics were collected by the police officers who compiled the reports, and the quality of the reports was then verified to allow LTSA to release the reports for research purposes. Although data were presented for characteristics objectively difficult to assess (e.g., distraction, fatigue), it is assumed that the quality control guarantees that police officers had sufficient evidence, thus limiting the potential bias. Among the crashes, 86.0% resulted in serious injuries and 14.0% in at least one fatality. Over two-thirds of the young drivers were male, and their age distribution was skewed towards younger age under 20 years old. The vast majority of young drivers drove passenger cars with vehicle age generally over 10 years, and were assigned at least partial fault in over four out of five crashes. Intoxication was verified or suspected in almost one out of four cases, reckless driving and speeding were recorded in almost half of the cases, and the presence of passengers was reported in one-third of the cases. 2.2.
Model
We performed crash pattern recognition by applying LCA, also known as model-based clustering or finite-mixture clustering. We selected LCA over alternative clustering methodologies because of the existence of an underlying statistical model that allows computing of cluster probabilities for new records, the ability to represent similarity across clusters, and the availability of goodness-of-fit criteria that facilitate objective decisions regarding the number of clusters (Depaire, Wets, & Vanhoof, 2008; Magidson & Vermunt, 2002). These advantages have been found beneficial in uncovering crash patterns in Belgium and Denmark (Depaire et al., 2008; Kaplan & Prato, 2013). LCA fostered the classification of road crashes into C latent classes, with uncertain class membership and unknown cluster size prior to the analysis. Consider a vector of N road crashes characterized by a vector of M variables (yi D y1, … , yM), and let Yi (Yi D Yi1, … , YiM) be the vector of values of crash n for the M variables. The LCA model is formulated as follows (Lanza, Collins, Lemmon, & Schafer, 2007): pðYi j uÞ D
K X
PðCk ÞpðYi j Ck ; uk Þ
(1)
kD1
where k is an indicator of latent class Ck, K is the number of latent classes, P(Ck) is the prevalence of latent class Ck in the data-set, p(YijCk,uk) is the conditional multivariate probability that a crash in class Ck would be characterized by vector Yi, and uk is a vector of parameters to be estimated. We made simplifying assumptions in order to derive an estimable LCA model with reasonable parametric
International Journal of Injury Control and Safety Promotion
3
Table 1. Sample characteristics. Variable
Categories
Per cent
Categories
Serious injury
86.0%
Fatal injury
14.0%
15 16 17 18 19
3.8% 8.2% 12.4% 13.5% 13.3%
20 21 22 23 24
12.2% 11.0% 9.5% 8.6% 7.6%
Male
68.5%
Female
31.5%
Young driver ethnicity
European Maori
57.9% 19.8%
Asian Other Unknown
5.9% 1.9% 14.6%
Young driver licence
Full Restricted
37.6% 34.2%
Learner Not valid
15.5% 12.7%
Young driver vehicle
SUV Car
3.7% 86.9%
Van Heavy vehicle
7.7% 1.6%
Less than 2 years 25 years
1.1% 2.8%
510 years More than 10 years
18.2% 72.6%
Young driver fault
Yes
83.8%
No
16.2%
Young driver intoxication
Yes
23.2%
No
76.8%
Young driver seat belt use
Yes
55.0%
No Unknown
8.7% 36.3%
Young driver manoeuvre
Traveling Lane changing Lost control
27.0% 5.4% 39.4%
Turning Parking/reversing Avoiding pedestrian None
17.0% 3.5% 6.8% 0.8%
Reckless driving Speeding Distraction
22.9% 17.5% 22.6%
Fatigue Inexperience None
6.5% 13.2% 36.0%
Young driver passenger age
No passengers Only young
67.8% 26.2%
Only adult Young and adult
4.9% 1.1%
Young driver passenger gender
No passengers Only male
67.8% 16.0%
Only female Male and female
5.6% 10.6%
Young driver passenger ethnicity
No passengers Same ethnicity
67.8% 24.2%
Different ethnicity
8.0%
No object Road Side
53.3% 9.3% 14.2%
Vehicle not moving Construction Animal
3.9% 19.0% 0.3%
Crash injury severity Young driver age (years)
Downloaded by [New York University] at 06:36 11 July 2015
Young driver gender
Young driver vehicle age
Young driver factors
Object hit
Per cent
(Continued)
4
H.B. Weiss et al.
Table 1. (Continued ) Variable
Categories
Per cent
Categories
Vulnerable road user involved
Pedestrian Cyclist
8.1% 3.5%
Motorcyclist/moped None
6.4% 82.0%
Single-vehicle
56.4%
Multiple-vehicle
43.6%
1524 years old 2564 years old
23.2% 66.8%
65 years old or more
10.0%
Male
62.4%
Female
37.6%
Same ethnicity
45.5%
Different ethnicity
54.5%
Other driver licencea
Full Restricted
76.8% 9.9%
Learner Not valid
4.3% 9.0%
Other driver vehiclea Other driver intoxicationa
SUV Car Yes
7.6% 74.6% 7.6%
Van Heavy vehicle No
12.2% 5.7% 92.4%
Other driver seat belt usea
Yes
71.1%
Travelling Lane changing Lost control
51.5% 8.5% 4.8%
No Unknown Turning Parking/reversing Avoiding pedestrian None
4.3% 24.6% 29.3% 5.0% 0.5% 0.5%
Reckless driving Speeding Distraction
22.2% 7.8% 15.5%
Fatigue Inexperience None
0.6% 1.9% 52.0%
Yes
43.5%
No
56.5%
Road type
Motorway State highway Open road
3.6% 25.6% 23.3%
Major urban road Minor urban road
25.2% 22.2%
Speed limit
50 km/h or less 5070 km/h One One Straight Easy curve
45.5% 8.3% 4.0% 82.2% 57.1% 19.3%
7090 km/h 90 km/h or more Two Two or more Moderate curve Severe curve
5.8% 40.5% 96.0% 17.8% 19.6% 3.9%
Level
76.4%
Hilly
23.6%
Intersection Section (
