- Email: [email protected]
Graduated driver license nighttime compliance in U.S. teen drivers involved in fatal motor vehicle crashes
Accident Analysis and Prevention 56 (2013) 110–117
Contents lists available at SciVerse ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Graduated driver license nighttime compliance in U.S. teen drivers involved in fatal motor vehicle crashes Dustin Carpenter a,b,∗ , Joyce C. Pressley a,c a b c
Department of Epidemiology, Columbia University, Mailman School of Public Health, 722 West 168th Street, New York, NY 10032, USA Saint Louis University School of Medicine, 1402 South Grand Blvd., Saint Louis, MO 63104, USA Department of Health Policy and Management, Columbia University, 722 W 168th Street, New York, NY 10032, USA
a r t i c l e
i n f o
Article history: Received 13 December 2011 Accepted 15 December 2011 Keywords: Motor vehicle crash Graduated driver licensing Teen driver
a b s t r a c t Background: Examination of teen driver compliance with graduated driver licensing (GDL) laws could be instrumental in identifying factors associated with persistently high motor vehicle mortality rates. Methods: Fatality analysis reporting system (FARS) data from the years 2006 to 2009 were used in this nation-wide cross-sectional study of drivers covered by a state nighttime GDL law (n = 3492). A new definition of weekend, based on the school night in relation to the teenage social landscape, redefined Friday night as a weekend night and Sunday night as a weekday/school night and compared it to previous weekend definitions. Multiple logistic regression was used to examine independent effects of demographic, behavioral, environmental, contextual, and other factors on compliance with nighttime GDL laws. All analyses were performed in Stata version 11. Results: Given coverage under nighttime GDL laws, drivers aged 15–17 years were non-compliant in 14.9% of the fatal MVCs in which they were involved, and nearly one-fifth (18.8%) of all fatalities aged 15–17 years were associated with non-compliance. Mortality risk was 10% higher using a revised social (school night) versus traditional (Sat–Sun) weekend definitions. In multivariable analysis, drivers noncompliant with nighttime GDL laws were more likely to be drinking (OR = 4.97, 3.85–6.40), unbelted (OR = 1.58, 1.25–1.99), driving on the weekend (OR = 1.82, 1.47–2.24), and killed (OR = 1.31, 1.04–1.65). Conclusion: GDL non-compliance contributes to teen motor vehicle mortality. Legislative and enforcement efforts targeting non-school night driving, seatbelt nonuse and alcohol have potential to further reduce teen driving mortality. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Motor vehicle crashes (MVCs) are the number one killer of U.S. teens (Chen et al., 2000; Corden et al., 2009; Gonzales et al., 2005; Shope, 2007), taking the lives of approximately 3000 teens aged 15–19 and accounting for 33–40% of deaths for this age group each year (CDC, 2010b; Gonzales et al., 2005; Karaca-Mandic and Ridgeway, 2010). Two of the most important contributors to teen MVC morbidity and mortality include driving at night and presence of other teen passengers (Williams, 2003). Approximately 15% of all vehicle-miles traveled by 16–17-year-old drivers occur between 9:00 pm and 5:59 am, but this time period accounts for approximately 40% of fatal crashes in this age group (Lin and Fearn, 2003; Williams and Preusser, 1997; Williams, 2003). Nearly 50% of teen
∗ Corresponding author at: Floraville Road, Millstadt, IL 62260, USA. Tel.: +1 618 567 5560. E-mail addresses: [email protected], [email protected] (D. Carpenter), [email protected] (J.C. Pressley). 0001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2011.12.014
MVC fatalities occur with other teen passengers and no supervising adult present (Williams and Ferguson, 2002; Williams, 2003), with fatal crash risk increasing with each additional teen passenger (Chen et al., 2000). In order to mitigate the excess risk posed to and by teen drivers, every state has passed some form of graduated driver licensing (GDL) legislation. The core of GDL licensure consists of nighttime driving restrictions and underage passenger-carrying limitations. Nighttime driving restrictions decrease fatal nighttime crashes (Williams, 2007), and passenger limitations reduce mortality and the proportion of crashes with multiple teen passengers present (Masten and Hagge, 2004; Williams, 2007). The combination of laws may be more effective than either law in isolation (Chen et al., 2006; Williams, 2007), with stronger laws producing larger reductions than weak ones (Mccartt et al., 2010; Morrisey and Grabowski, 2006; Pressley et al., 2009; Traynor, 2009). Despite the dramatic lowering of teen MVC morbidity and mortality by GDL laws, the crash rate for teens remains high. Individual, behavioral and contextual factors such as alcohol involvement, restraint use, weather, and the degree of urbanization
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
111
affect driving risk in all age groups (CDC, 2010a). Recently, national studies have begun to explore how the relationship between some of these factors and GDL laws impacts teen MVC morbidity and mortality (Morrisey and Grabowski, 2006). Additionally, a limited number of studies using surveys have attempted to investigate teens’ level of reported compliance with nighttime and passenger restrictions (Goodwin and Foss, 2004; Goodwin et al., 2006; Williams et al., 2002). These studies suggest non-compliance is a significant issue and that teens are more likely to be non-compliant with passenger limitations than nighttime driving restrictions (Williams, 2007), but they do not directly reveal non-compliance’s impact and contribution to teen MVC mortality. Because passenger restrictions vary, ranging from states prohibiting the carrying of any teen passengers to allowing three or more teen passengers, the effect of non-compliance and its relationship to teen motor vehicle mortality is less clear than for nighttime driving restrictions. This study examines factors associated with nighttime GDL compliance in fatal crashes involving a teen driver. It also explores whether traditional definitions of the weekend fully capture teen weekend driving risk. An improved understanding of the relation between individual, behavioral, environmental, contextual, and state-level factors with GDL non-compliance could be instrumental in further reducing teen driving fatalities.
2.3.1. Compliance with state GDL Compliance with nighttime driving restrictions is individually coded for each state because the times when teens can legally drive varies across states (e.g., Idaho – sunrise to sunset (lifted at 16 years of age); Alaska – 1–5 am (lifted at 16 years and 6 months of age); Vermont – no nighttime restriction (thus excluded); see Insurance Institute for Highway Safety (2011) for further details). FARS data contains variables to determine compliance: age of driver, state of driver residence, and time and date of crash. States that passed laws during the study timeframe had teens who were listed as covered or not covered in relation to when the law went into effect. State-specific nighttime restrictions and variables for time of crash, and, in some cases, day of the week of the crash were used to assess whether a crash occurred during restricted hours. A crash was categorized as non-compliant with nighttime GDL laws if it occurred outside of legally permitted driving times. A teen driver was considered non-compliant with passenger GDL restrictions if more passengers of the specified ages were present than allowed by the state-specific law.
2. Materials and methods
2.3.2. Demographic variables Age and gender were obtained for each driver and passenger involved in each MVC using uniform FARS coding.
2.1. Data sources Analyses for traffic fatalities and relevant covariates for the years 2006–2009 utilized data obtained from the National Highway Traffic Safety Administration’s (NHTSA) fatality analysis reporting system (FARS). FARS is a motor vehicle crash (MVC) census database that contains every traffic crash on public roads that results in a fatality of a motorist or non-motorist within 30 days of the crash in all 50 states, Washington D.C., and Puerto Rico. FARS is a publically accessible, de-identified database (NHTSA, 2009). Intercensal denominator estimates were acquired from the Center for Disease Control and Prevention’s (CDC) web-based injury statistics query and reporting system (WISQARS) for total population by age and state. WISQARS population data are taken from the United States Census Bureau and the BridgedRace Post-censal Population Estimates of the United States, which is a collaboration of the U.S. Census Bureau and the National Center for Health Statistics (NCHS) updated annually (www.cdc.gov/nchs/about/major/dvs/popbridge/popbridge.htm). The years 2006–2007 are the most recently available data for WISQARS. In order to obtain population figures for the years 2008 and 2009, the percent change from 2006 to 2007 was used to estimate the population for 2008–2009 (CDC, 2010a,b).
2.2. Study population The study population was comprised of fatal MVCs involving at least one teen driver between the ages of 15 and 17 years who was covered by a state nighttime GDL law (n = 3492). Since FARS includes data on all MVCs that result in a fatality, the drivers who comprise our study population did not necessarily die in the crash (i.e. the death could have been a passenger in the driver’s vehicle, who was not included in our study population). FARS data provides age in increments of one year (i.e. integers), but some states lift their GDL provisions at non-integer time-points. At the time of this
study, eleven states had laws that precluded inclusion of at least one age group, resulting in exclusion of 544 teen driver crashes.1 2.3. Variable classications
2.3.3. Crash fatality In order to determine driver case fatality rates (CFR), given involvement in a crash, a variable was dichotomized into fatal and non-fatal injury. Fatal injuries comprised the numerators and were used with population census data to establish age-specific mortality rates per 100,000 population. 2.3.4. Teen residence/crash location The variable, “road function”, identifies crash location as rural or urban based on population density data in the area surrounding the crash. 2.3.5. Interstate highways FARS provides uniform coding for dichotomizing crashes that occur on an interstate or non-interstate highway, providing an indicator for roadways with varying levels of access. 2.3.6. Weather The FARS variable for adverse road conditions such as rain, sleet, snow, fog, mist, and severe crosswinds or other conditions that existed at the time of crash was dichotomized, indicating presence or absence of adverse weather conditions. 2.3.7. Day of the week FARS coding for day of the week and time of day was used to create a dichotomous variable for weekday/weekend. Three definitions of weekend were examined in this study to assess which more accurately reflects crash risk among teen drivers: (1) “traditional weekend” was coded as “weekend” for the days Saturday and Sunday, while the remainder of the week was coded as “weekday”; (2) “expanded weekend” (from the 2009 FARS codebook) defines the weekend as the time period from 6:00 pm on Fridays until 5:59 am
1 Of note, Missouri lifts its GDL at 17 years and 11 months, and, in order not to lose its data, all 17-year olds were considered covered under its provision.
112
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
on Monday mornings; and (3) “social weekend” defines weekends as the period from Friday at 5:00 pm through 4:59 pm Sunday. 2.3.8. Alcohol involvement The presence of alcohol was assessed using a derived composite variable, “dr drink”, that incorporates charges filed by police, police-reported alcohol involvement and alcohol tests that note presence of sufficient evidence to consider alcohol as a contributing factor in the crash. Presence of alcohol is coded as a dichotomous variable. 2.3.9. Seatbelt restraints The FARS variable, “restraint use”, indicates seat belt use as present or absent at the time of crash.
“weekend + no adult passenger present”, and “interstate highway + driving over the speed limit”. Interaction terms were added one at a time into the final multivariable model. Significance was assessed using the likelihood ratio test and the associated p-value of the interaction term with the cut-off = 0.05. 3.1. Traditional, expanded, and social weekend In order to assess the impact that each weekend definition has for explaining the risk of MVC death among teenagers, an adjustment for differences in length of time covered by the weekend variables was made to facilitate comparisons between the three definitions. All data was coded, transformed, and analyzed using Stata version 11.
2.3.10. Speeding A variable indicating whether or not a driver was speeding at the time of the crash utilized FARS’ default coding.
4. Results
2.3.11. Number of passengers The number of passengers in the vehicle, besides the driver, is derived from the FARS variable for number of vehicle occupants, with the number of occupants minus one to account for the driver. Youth passengers are defined as those aged 25 years or younger, and adult passengers are those aged 25 years and older. This definition was used because studies have shown that older teens and young adults increase access to alcohol for younger teens (Wagenaar et al., 1993) and increase crash risk (Chen et al., 2000). Insurers use a similar definition for driver age because of the aforementioned risk associated with younger drivers and passengers (www.iihs.org).
There were 20,967 teens aged 15–17 years involved in fatal MVCs between 2006 and 2009, of which 7415 were drivers. Of the drivers, coverage with state nighttime GDL laws could be determined for 6871 (92.7%) cases. Of these, 3492 (50.8%) were covered by a state nighttime GDL law. These teens comprise the study population. The population characteristics of teen drivers covered by a nighttime GDL law are shown by compliance status in Table 1. Coverage by a nighttime GDL provision differed significantly by age, with coverage decreasing with each increasing year of age for 15-, 16-, and 17-year olds (90.1% vs. 84.4% vs. 29.5%). Of the 3492 teens covered by a nighttime GDL law, 520 (14.9%) were in violation of nighttime GDL restrictions at the time of the fatal crash. Of the 1464 (41.9%) who died in crashes, 275 (18.8%) were non-compliant with their nighttime GDL restriction.
2.3.12. Seat belt Law A variable was dichotomized to indicate whether or not drivers were covered by a primary seat belt law as reported by the Insurance Institute of Highway Safety (www.iihs.org).
4.1. Study population
4.2. Risk factors 2.3.13. Fatality rate in person-years Fatality rates for the three definitions of weekend were calculated using person-years as the denominator. The age-specific population for each year was summed to determine the number of person-years each age group contributed over the study period. 3. Statistical analysis Given coverage by a nighttime GDL provision, the proportion of teen drivers and teen fatalities aged 15–17 involved in fatal MVCs while drivers were non-compliant with their state’s GDL law was calculated. These proportions are reported for the nighttime driving restriction by key covariates. Chi-squared tests, chi-squared test for trends and Student’s t-test were used to assess the statistical significance of categorical variables and group means, respectively, between compliant and non-compliant drivers. Significance is defined as p-values ≤0.05. We based our model building strategy on the recommendations in Hosmer and Lemeshow (2010)’s Applied Logistic Regression. Variables hypothesized to have an association with compliance with GDL laws were analyzed univariately with the outcome of interest, non-compliance with the nighttime GDL. Variables with p-values ≤0.25 were considered for the final multivariable logistic regression model. Correlations were then assessed for all predictor variables of interest. Two variables, “not wearing seat belt” and “death in the crash”, had an ‘r’ of 0.30, and were ultimately left in the final model because of their epidemiological significance. Several interaction terms were considered, including “weekend + drinking and driving”, “gender + drinking and driving”, “drinking and driving + not wearing a seat belt”, “gender + not wearing a seat belt”,
4.2.1. Age Drivers aged 17 years were more likely to drink and drive than 15- or 16-year-old drivers (12.6%, 11.5%, and 9.8%, 2 for trend by age = 6.0, p = 0.049). Younger teens were more likely to be unbelted (47.9%, 34.0% and 36.8%, 2 for trend by age = 29.6, p < 0.001), speeding (39.1%, 34.7% and 29.8%, 2 for trend by age = 15.7, p < 0.001), involved in single vehicle crashes (83.4%, 83.2% and 79.2%, 2 for trend by age = 11.3, p = 0.024), and die in their crashes (44.9%, 43.7% and 38.8%, 2 for trend by age = 9.0, p = 0.011) compared to older teen drivers. Younger drivers were also more likely to have a passenger older than 25 at the time of crash (16.0%, 6.8% and 4.8%, 2 for trend = 71.4, p < 0.001), but driver age was not related to weekend vs. weekday crash (2 for trend = 1.3, p = 0.53) or driving with more than two youth passengers (2 for trend = 1.8, p = 0.41). 4.2.2. Gender Male drivers were more likely than female drivers to be drinking and driving (13.0% vs. 7.1%, 2 = 27.7, p < 0.001), unbelted (41.1% vs. 28.8, 2 = 47.5, p < 0.001), and speeding (37.1% vs. 26.3%, 2 = 40.9, p < 0.001). However, male drivers were not significantly more likely to die in their crashes (2 = 2.5, p = 0.12), crash on the weekend (2 = 2.5, p = 0.12) or to have been driving with more than two youth passengers (2 = 0.76, p = 0.38) compared to female drivers. 4.2.3. Traditional, expanded, and social weekend Three weekend definitions and their effect on measuring teen mortality risk are examined in Table 2. In our study population, the social definition of the weekend had a 10.1% and 10.3% higher mortality rate compared to the traditional and expanded weekend
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
113
Table 1 Population characteristics for 15–17-year-old drivers complaint and non-compliant with nighttime GDL. Total covered by nighttime GDL (n = 3478) Non-compliant (n = 520) Compliant (n = 2958)
Total (n = 3478)
p-Value
Personal characteristics Age 15 16 17 Male (male) Fatality (died)
388 (13.1%) 1489 (50.3%) 1081 (36.6%) 1935 (65.4%) 1189 (40.3%)
87 (16.7%) 255 (49.1%) 178 (34.2%) 390 (75.0%) 275 (53.0%)
475 (13.7%) 1744 (50.1%) 1259 (36.2%) 2325 (66.9%) 1464 (42.2%)
<0.001 <0.001
Environmental characteristics Weather (bad weather) Rural (rural) Interstate highway
302 (10.2%) 1766 (60.2%) 410 (14.1%)
45 (8.7%) 258 (49.8%) 93 (18.0%)
347 (10.0%) 2024 (58.7%) 503 (14.7%)
0.282 <0.001 0.02
171 (32.9%) 253 (53.2%) 224 (43.1%) 1.26 (1.43) 1.11 (1.32) 145 (27.9%) 0.07 (0.36) 25 (4.8%) 79 (27.0%)
386 (11.1%) 1196 (37.0%) 1168 (33.6%) 1.06 (1.31) 0.93 (1.20) 824 (23.7%) 0.08 (0.36) 249 (7.2%) 401 (21.8%)
<0.001 <0.001 <0.001 <0.001 <0.001 0.015 0.138 0.024 0.02
Behavioral characteristics Drinking and driving Not wearing seatbelt Speeding Mean number of passengers (SD) Mean number pass ≤25 (SD) Increased youth pass (GE 2 pass ≤25) Mean number of adult pass >25 (SD) Adult passenger present (>25) Non-compliance with passenger GDL
215 (7.3%) 943 (34.2%) 944 (31.9%) 1.04 (1.25) 0.9 (1.18) 679 (23.0%) 0.09 (0.36) 224 (7.6%) 322 (20.9%)
0.081
Contextual characteristics Social weekend (5 pm Fri to 5 pm Sun) Traditional weekend (Sat–Sun)
1069 (36.1%) 919 (31.1%)
281 (54.0%) 274 (52.7%)
1350 (38.8%) 1193 (34.3%)
<0.001 <0.001
State-level laws Primary seatbelt law
1247 (42.9%)
219 (42.9%)
1466 (42.9%)
0.99
Table 2 Impact of different definitions of the weekend on fatality rates. Social weekend
Traditional weekend
Expanded weekend
Crashes involving 15–17-year-old drivers Number of 15–17-year-old fatalities over study period Person-years for 15–17-year olds for 2006–2009/100,000 Proportion of time each definition captures Fatality rate
1808 (39.8%) 523 0.286 12.1/100,000/person-year
1625 (35.8%) 523 0.286 10.9/100,000/person-year
2027 (44.7%) 523 0.357 10.9/100,000/person-year
Crashes involving all-age drivers Number of 15–17-year-old fatalities over study period Person-years for 15–17-year olds for 2006–2009/100,000 Proportion of time each definition captures Fatality rate
2905 (40.9%) 523 0.286 19.1/100,000/person-year
2651 (37.4%) 523 0.286 17.7/100,000/person-year
3322 (46.8%) 523 0.357 17.8/100,000/person-year
160 140 120
Friday Saturday Mon-Thurs Sunday
100 80 60 40 20 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
4.2.4. Alcohol, speeding, and seatbelts Drivers who were speeding (50.8% vs. 30.1%, 2 = 132.9, p < 0.001) or drinking and driving (64.7% vs. 33.5%, 2 = 135.5, p < 0.001) were more likely to be unbelted than drivers obeying speed limit laws or not drinking, respectively. Drivers who crashed over the social weekend had a 29% higher proportion of their crashes involve alcohol compared to weekday crashes (14.7% vs.
8.8%, 2 = 29.6, p < 0.001); however, seat belt wearing did not differ significantly between weekend and weekday crashes (37.3% vs. 36.7% unbelted, 2 = 0.13, p = 0.71). Drivers carrying more than two
Number of Fatalities
definitions, respectively, adjusted for the number of hours captured by each definition. The social definition captures 183 more deaths than the traditional definition, even though they are both 48 h time periods. Over one-half (53.5%) of all non-compliant crashes took place between 6:00 pm on Friday night and 5:59 am on Saturday morning or between 6:00 pm on Saturday night and 5:59 am on Sunday morning (278/520). Friday night to Saturday morning had 2.3 times more non-compliant crashes (126 vs. 55) than the period from Sunday night to Monday morning. Of note, the majority of non-compliant crashes that occurred between Sunday night and Monday morning took place during summer or holiday weekends (i.e. when Sunday night was not a school/weeknight). Figs. 1 and 2 compare teen driver fatalities on Sunday to other weekday and weekend days and nights.
Hour (Military Time) Fig. 1. 15–17-year-old driver fatalities by hour of the day.
114
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
160 Friday Sunday
Number of Fatalities
140 120 100 80 60
4.3.5. Weather, day of the week, and seat belt laws Non-compliant drivers were 20% more likely to crash on urban roads (50.2% vs. 39.8%, 2 = 19.7, p < 0.001), 22% more likely to crash on interstate highways (18.0% vs. 14.1%, 2 = 5.4, p = 0.02), and one-third more likely to crash on the weekend (54.0% vs. 36.1%, 2 = 59.7, p < 0.001) compared to compliant drivers. Neither weather conditions (2 = 1.2, p = 0.28) nor having a primary statelevel seatbelt law targeting teen drivers (2 = 0.0002, p = 0.99) were associated with compliance (Table 1).
40 20
4.4. Multivariable logistic regression m
m 5a
4a
m
m
m 3a
2a
1a
am
pm
m
pm
12
11
10
9p
m
m 8p
7p
6p
m
0
Hour Fig. 2. Friday night vs. Sunday night.
passengers below the age of 25 were more likely to drink and drive (13.5% vs. 10.3%, 2 = 6.7, p = 0.009) and speed (51.9% vs. 30.9%, 2 = 34.2, p < 0.001) compared to drivers who were carrying one or fewer passengers, but increased passengers was not associated with driver seat belt use (2 = 0.2, p = 0.89). 4.3. Nighttime compliance and risk factors 4.3.1. Age Compliance with nighttime GDL provisions was not statistically different by age (18.3%, 14.6% and 14.1%, 2 for trend = 5.0, p = 0.081), except for the difference between 15- and 17-year olds (2 = 4.6, p=.03). Irrespective of age, non-compliant drivers were 24% more likely to die in their crash compared to compliant drivers (53.0% vs. 40.3%, p < 0.001). 4.3.2. Gender Given coverage by a nighttime GDL provision, 16.8% of all male drivers and 11.3% of all female drivers were non-compliant with their states’ nighttime GDL (2 = 18.3, p < 0.001). Also of note, 15-year-old drivers were significantly more likely to be male than 16–17-year-old drivers, especially in non-compliant crashes (2 = 15.6, p < 0.001) (Table 1). 4.3.3. Alcohol, speeding, and seatbelts Compliant and non-compliant drivers differed by several behavioral factors (Table 1). Non-compliant drivers were more likely to be drinking and driving, speeding, and unbelted compared to compliant drivers. 4.3.4. Passengers Crashes with two or more youth passengers present in the vehicle were 30% more likely to be non-compliant than crashes with 1 or fewer youth passengers (17.6% vs. 14.1%, 2 = 5.9, p = 0.02). For drivers covered by both nighttime and passenger GDL restrictions, drivers non-compliant with the passenger GDL were 40% more likely to be non-compliant with the nighttime GDL compared to teen drivers compliant with the passenger restriction (19.7% vs. 14.9%, 2 = 5.49, p = 0.02). Of note, 27% of crashes were non-compliant with either the nighttime or the passenger GDL restriction (when teens were covered by both types of law and compliance with each coverage could be assessed). The odds of nighttime non-compliance increased 13% with each additional passenger present in the vehicle. Presence of an adult passenger greater than 25 years of age was associated with higher nighttime GDL compliance (2 = 5.1, p = 0.02).
The unadjusted and fully adjusted ORs, with 95% confidence intervals and associated p-values for each of the covariates are reported in Table 3. After controlling for the other listed covariates, non-compliant drivers were more likely to have been drinking alcohol, speeding, unbelted, crash on an urban road and interstate highway, crash during the weekend, and die in their crash compared to compliant drivers. Although not statistically significant in the multivariable model, non-compliant drivers also tended to be younger, male, speeding, and carrying two or more youth passengers below 25 years of age compared to compliant drivers. Risk factors varied by gender and age for drivers covered by a nighttime GDL. Therefore, several interaction terms were assessed in the final multivariable model. Despite the differences between characteristics of teen drivers and the circumstances surrounding their crashes, all of the interaction terms were insignificant when entered sequentially into the final multivariable model. All analyses were performed in Stata version 11 (StataCorp., 2009). 5. Discussion The goal of graduated driver licensing legislation is to decrease injury and fatality rates by reducing teens’ exposure when they are most at risk, while valuable driving experience is accumulated under lower-risk conditions (Hyde et al., 2005; Waller, 2003). Past studies have shown that GDLs are effective in lowering statelevel mortality rates (Chen et al., 2006; Dee et al., 2005; Ferguson, 2003; Fohr et al., 2005; Kirley et al., 2008; Males, 2007; Margolis et al., 2007; Masten and Foss, 2010; Neyens et al., 2008; Williams et al., 2005; Zhu et al., 2009), with stronger laws producing larger reductions in fatal crash rates compared to weaker laws (Mccartt et al., 2010; Morrisey and Grabowski, 2006; Pressley et al., 2009; Traynor, 2009). However, teen fatality rates remain significantly higher than those of other age groups. Despite this, GDL laws remain heterogeneous, with many states having vastly different nighttime restrictions and passenger-carrying limitations. A major difficulty in writing and legislating teenage traffic laws is retaining the balance between teens’ increased risk of death, which is well-documented, and teens’ need for mobility. This is reflected in the fact that the most rural states, whose teens often have the fewest public transportation options, tend to have the least strict GDL laws (Morrisey and Grabowski, 2006). However, there is a small but growing pool of data that suggests, at least through age 17, that nighttime and passenger laws do not have a significantly negative impact on teens’ quality of life (Williams, 2009). This suggests that the benefits of GDL laws likely outweigh the societal costs. Many states have successfully implemented stricter GDL laws without compromising teens’ quality of life (Williams, 2009). Survey evidence from California found that nearly three-quarters of teens reported not having their quality of life significantly affected by their nighttime or passenger restrictions (Williams et al., 2002), and a study in New Jersey, which is the sole state to enforce an intermediate license through the age of 17, found similar
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
115
Table 3 Univariate and multivariate analysis for non-compliance with nighttime GDL. Univariate Odds ratio (95% C.I.)
Multivariable p-Value
Adjusted odds ratio (95% C.I.)
p-Value
Personal characteristics Age 16 vs 17 15 vs 17 Male Fatality
0.95 (0.79–1.14) 1.33 (1.03–1.72) 1.59 (1.28–1.96) 1.67 (1.39–2.02)
0.585 0.027 <0.001 <0.001
1.26 (0.996–1.60) 1.66 (1.21–2.30) 1.27 (0.998–1.60) 1.31 (1.04–1.65)
Environmental characteristics Bad weather Rural Interstate highway
0.84 (0.60–1.16) 0.66 (0.54–0.79) 1.34 (1.05–1.72)
0.282 <0.001 0.02
− 0.55 (0.44–0.68) 1.43 (1.08–1.90)
− <0.001 0.014
6.25 (4.97–7.86) 2.18 (1.79–2.66) 1.61 (1.34–1.95) 1.30 (1.05–1.60) 1.14 (1.06–1.23) 0.79 (0.59–1.08) 0.62 (0.40–0.94) 1.13 (1.06–1.21) 1.40 (1.05–1.86)
<0.001 <0.001 <0.001 0.0163 <0.001 0.14 0.025 <0.001 0.021
4.97 (3.85–6.40) 1.58 (1.25–1.99) 1.18 (0.94–1.47) 1.16 (0.91–1.49) − − 0.80 (0.51–1.27) − −
<0.001 <0.001 0.145 0.224 − − 0.347 − −
2.08 (1.72–2.51)
<0.001
1.82 (1.47–2.24)
<0.001
1.00 (0.83–1.21)
0.99
Behavioral characteristics Drinking and driving Not wearing seatbelt Speeding Increased youth pass (≥2 pass) Number of youth passengers (≤25) Number of adult passengers (>25) Adult passenger present (>25) Number of passengers Non-compliance with passenger GDL Contextual characteristics Social weekend (5 pm Fri to 5 pm Sun) State-level laws Primary seatbelt law
results (Williams, 2009). Furthermore, surveys of affected teens’ parents from Pennsylvania, North Dakota, New Jersey, California and Connecticut have suggested widespread parental support, with large majorities supporting current level or even more stringent restrictions, with few parents neutral or opposed to their child’s restrictions (Mckay et al., 2008; Vachal et al., 2009; Williams et al., 2002; Williams and Chaudhary, 2008). Nonetheless, survey evidence also suggests that compliance with restrictions is far from 100%. Our study documented the contribution of nighttime noncompliance to teen motor vehicle mortality and associated risk factors. In our study, given coverage with a nighttime GDL law, nearly one-fifth of all 15–17-year old drivers’ fatalities (18.8%) occurred while they were non-compliant. Of all the risk factors associated with non-compliance, the behavioral-level factors such as drinking and driving, speeding, and seatbelt use showed the strongest association within our study population. Non-compliant crashes predominately occurred on the weekend, especially Friday and Saturday nights, making the examination of non-compliant crashes with respect to the school year more effective in capturing teen crash risk than using traditional definitions of the weekend. These results indicate that while weekend driving is associated with increased numbers of fatal MVCs for teen drivers, it is especially associated with non-compliant driving, in which a 24 h period (Friday night to Saturday morning and Saturday night to Sunday morning) accounted for over one-half of all fatal non-compliant crashes. These findings further suggest that compliance enforcement dollars would be best spent targeting drivers on non-school nights. The proportion of non-compliant crashes we found comports with survey evidence. In North Carolina, 23% of teens reported violating their nighttime GDL restrictions either with or without parental permission; however, most teens admitted to only infrequent violations (Goodwin and Foss, 2004). In Nova Scotia, 12% of teens admitted to frequently violating their nighttime GDL restriction (Mayhew et al., 2003), which accords with data from California, in which 15% of teens admitted to violating their nighttime restrictions more than 10 times (Williams et al., 2002).
−
0.054 0.002 0.052 0.023
−
In our study population (i.e. teens covered by a nighttime GDL restriction), when a passenger GDL restriction was also in place, 21.8% of teen drivers were violating it at the time of fatal crash occurrence. Other studies have also reported finding that teens disobey passenger restrictions more frequently than nighttime restrictions. For example, in North Carolina 34% of teens reported violating their passenger restriction either with or without parental permission (Goodwin and Foss, 2004), and, in California, 60% of teens admitted to violating it more than 10 times (Williams et al., 2002). Differences in compliance in relation to nighttime and passenger GDL restrictions are likely due to the heterogeneity of passenger GDL restrictions, in which some states forbid any number of passengers whereas others allow up to 3 passengers. Thus, some states allow for teen vehicles to be filled with other teens as passengers and still be considered compliant with their passenger GDL, while, in other states, teen vehicles could have a single passenger and be deemed non-compliant. After controlling for the number of hours included in the definitions, our newly defined weekend definition exhibited greater ability to capture risk than the traditional and expanded weekend definitions. Comparing the social definition versus the traditional definition, we see that the difference results from substituting the lower crash risk on Sunday night for the higher risk of Friday night. We found that incorporating Friday night more accurately reflects the teenage social landscape, in which Friday night is a popular night for group gatherings, including high-school football games, house parties, and other events. This is reflected by the fact that Friday has a much higher crash count compared to Sunday night. We feel this is due in part to at least two phenomena: (1) that driving on Friday night produces more exposure through larger numbers of teens on roadways, and (2) other risky behaviors like drinking alcohol are more prevalent on Friday nights. We excluded Sunday night from our definition because we felt it is traditionally used to rest and recharge for another schoolweek, unlike Friday night. The expanded definition of the weekend includes Friday night; however, it also extends to 5:59 am on Monday morning. Including this latter time period dilutes the increased risk we see when only including Friday night in the definition.
116
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
Examining drivers who were covered by nighttime GDL laws, the behavioral factors were most strongly related to noncompliance – especially drinking and driving, driving on the weekend, and being unrestrained. These factors’ relationship to nighttime GDL non-compliance suggests several potential points of intervention. Strengthening seat belt and alcohol laws targeting teen drivers, as well as their enforcement, in conjunction with increased GDL enforcement on the weekends has the potential to increase compliance with nighttime GDLs and to mitigate the effects of non-compliance. Previous studies have found that the passage of GDL laws has been associated with increased seat belt usage (Hyde et al., 2005), and our findings suggest that being unbelted is significantly associated with nighttime GDL non-compliance. These findings serve to underscore the importance of the Centers for Disease Control and Prevention’s recommendations for what can be done to combat motor vehicle mortality in teen drivers, including consistent and strong three-phase GDL laws in every state, primary seatbelt laws targeting teen drivers, as well as minimum legal drinking age and zero tolerance laws for drivers under the age of 18 years (http://www.cdc.gov/Motorvehiclesafety/teenbrief/index.html) (CDC, 2011). This study has several limitations. First, there was a small number of drivers for whom there was an inability to determine compliance with nighttime GDL laws, and they were therefore excluded from the study. We feel it is unlikely that this greatly biased our findings since there is no evidence that compliance for the indeterminate category varied systematically from the teens in the study. However, to the extent that excluded teens differ from teens in the study, this could be a source of bias. One state may have included a small number of misclassified teens due to our including all 17 year olds even though its restriction was lifted at 17 years and 11 months of age. Although studies have shown that teens from urban and rural areas have different crash risks (Donaldson et al., 2006), our variable for crash location was unable to differentiate between crash location and teen residence in urban or rural areas. It is possible that non-compliant crashes were significantly more urban than compliant crashes because areas considered “urban” were more likely weekend destinations for all teens. Furthermore, because non-compliant crashes were more likely to occur during the weekend, they may appear to more greatly impact urban teens, when in fact many of these crashes were actually sustained by rural drivers, who travel into urban areas to congregate socially. We suspect that this effect explains at least a portion of the discrepancy observed in crash location; however, the magnitude of this misclassification is unknown. Insurance claims data suggest that most crashes occur within five miles of home (Progressive, 2002), however, it is unclear if this is the case for teen drivers or if this varies by day of the week. We were also unable to determine if teens were driving with members of their family at the time of their crash, which may have resulted in some teens inappropriately being considered non-compliant in some states. Furthermore, some rural teens are exempt from certain GDL restrictions, which could not be captured in our study. Finally, this study captures the proportion of teen drivers and teen fatalities that were involved in a fatal MVC, not the actual rate of non-compliance.
6. Conclusion Non-compliance with state-specific nighttime GDLs is a significant source of mortality for U.S. teens. Continued action by parents, police, researchers and policy-makers is needed to adequately address this challenge. Adoption of the social definition
of the weekend would more accurately describe teenagers’ social landscape and capture the risk associated with teen driving . A more developed understanding of the level and characteristics of GDL non-compliance as well as the accurate assessment of teen drivers’ weekend risk would allow for more focused policy analysis and facilitate decision-making with regard to targeting potential GDL improvements. The findings of this study suggest that GDL enforcement dollars would be best spent targeting weekend driving, alcohol and seat belt use in teen drivers.
Acknowledgments This manuscript was awarded a “Best Student Paper Award” from the Injury Control and Emergency Health Services Section of the American Public Health Association, 2011. The award was sponsored, in part, by the Liberty Mutual Research Institute for Safety. Contents of this effort are solely the responsibility of the authors and do not necessarily represent the official view of the American Public Health Association or the Liberty Mutual Research Institute for Safety. The authors would like to thank Dr. Barbara Barlow, M.D., who graciously read several drafts of an earlier work related to this paper.
References Centers for Disease Control and Prevention, 2010a. Drivers aged 16 or 17 years involved in fatal crashes—United States, 2004–2008. MMWR Morb. Mortal. Wkly. Rep. 59 (41), 1329–1334. Centers for Disease Control and Prevention, 2010b. Web-Based Injury Statistics Query and Reporting System (WISQARS). National Center for Injury Prevention and Control, Centers for Disease Control and Prevention (online). Centers for Disease Control and Prevention, 2011. Policy Impact: Teen Driver Safety. National Center for Injury Prevention and Control, Division of Unintentional Injury Prevention. Chen, L.H., Baker, S.P., Braver, E.R., Li, G., 2000. Carrying passengers as a risk factor for crashes fatal to 16- and 17-year-old drivers. JAMA 283 (12), 1578–1582. Chen, L.H., Baker, S.P., Li, G., 2006. Graduated driver licensing programs and fatal crashes of 16-year-old drivers: a national evaluation. Pediatrics 118 (1), 56–62. Corden, T.E., Tonellato, D.J., Frisch, K.B., Laud, P.W., 2009. Graduated driver licensing policy in the great lakes states: current benefits and future potential. WMJ 108 (8), 393–397. Dee, T.S., Grabowski, D.C., Morrisey, M.A., 2005. Graduated driver licensing and teen traffic fatalities. J. Health Econ. 24 (3), 571–589. Donaldson, A.E., Cook, L.J., Hutchings, C.B., Dean, J.M., 2006. Crossing county lines: the impact of crash location and driver’s residence on motor vehicle crash fatality. Accid. Anal. Prev. 38 (4), 723–727. Ferguson, S.A., 2003. Other high-risk factors for young drivers—how graduated licensing does, doesn’t, or could address them. J. Saf. Res. 34 (1), 71–77. Fohr, S.A., Layde, P.M., Guse, C.E., 2005. Graduated driver licensing in wisconsin: does it create safer drivers? WMJ 104 (7), 31–36. Gonzales, M.M., Dickinson, L.M., Diguiseppi, C., Lowenstein, S.R., 2005. Student drivers: a study of fatal motor vehicle crashes involving 16-year-old drivers. Ann. Emerg. Med. 45 (2), 140–146. Goodwin, A.H., Foss, R.D., 2004. Graduated driver licensing restrictions: awareness, compliance, and enforcement in north carolina. J. Saf. Res. 35 (4), 367–374. Goodwin, A.H., Wells, J.K., Foss, R.D., Williams, A.F., 2006. Encouraging compliance with graduated driver licensing restrictions. J. Saf. Res. 37 (4), 343–351. Hosmer, D.W., Lemeshow, S., 2010. Applied Logistic Regression, 2nd ed. WileyInterscience Publication. Hyde, L.K., Cook, L.J., Knight, S., Olson, L.M., 2005. Graduated driver licensing in Utah: is it effective? Ann. Emerg. Med. 45 (2), 147–154. Institute for Highway Safety, Young driver licensInsurance ing systems in the U.S.: intermediate & unrestricted stages, http://www.iihs.org/laws/gdl intermediate.aspx (retrieved). Karaca-Mandic, P., Ridgeway, G., 2010. Behavioral impact of graduated driver licensing on teenage driving risk and exposure. J. Health Econ. 29 (1), 48–61. Kirley, B.B., Feller, A., Braver, E., Langenberg, P., 2008. Does the maryland graduated driver licensing law affect both 16-year-old drivers and those who share the road with them? J. Saf. Res. 39 (3), 295–301. Lin, M.L., Fearn, K.T., 2003. The provisional license: nighttime and passenger restrictions—a literature review. J. Saf. Res. 34 (1), 51–61. Males, M., 2007. California’s graduated driver license law: effect on teenage drivers’ deaths through 2005. J. Saf. Res. 38 (6), 651–659. Margolis, L.H., Masten, S.V., Foss, R.D., 2007. The effects of graduated driver licensing on hospitalization rates and charges for 16-and 17-year-olds in north Carolina. Traffic Inj. Prev. 8 (1), 35–38.
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117 Masten, S.V., Foss, R.D., 2010. Long-term effect of the north carolina graduated driver licensing system on licensed driver crash incidence: a 5-year survival analysis. Accid. Anal. Prev. 42 (6), 1647–1652. Masten, S.V., Hagge, R.A., 2004. Evaluation of California’s graduated driver licensing program. J. Saf. Res. 35 (5), 523–535. Mayhew, D.R., Simpson, H.M., Desmond, K., Williams, A.F., 2003. Specific and longterm effects of Nova Scotia’s graduated licensing program. Traffic Inj. Prev. 4 (2), 91–97. Mccartt, A.T., Teoh, E.R., Fields, M., Braitman, K.A., Hellinga, L.A., 2010. Graduated licensing laws and fatal crashes of teenage drivers: a national study. Traffic Inj. Prev. 11 (3), 240–248. Mckay, M.P., Coben, J.H., Larkin, G.L., Shaffer, A., 2008. Attitudes of teenagers and their parents to Pennsylvania’s graduated driver licensing system. Traffic Inj. Prev. 9 (3), 217–223. Morrisey, M.A., Grabowski, D.C., 2006. Graduated drivers license programs and rural teenage motor vehicle fatalities. J. Rural Health 22 (4), 300–307. National Highway Traffic Safety Administration, 2009. Fatality Analysis Reporting System (FARS). National Center for Statistics and Analysis, National Highway Traffic Safety Administration, available from URL http://www.nhtsa.gov/FARS (online). Neyens, D.M., Donmez, B., Boyle, L.N., 2008. The Iowa graduated driver licensing program: effectiveness in reducing crashes of teenage drivers. J. Saf. Res. 39 (4), 383–390. Pressley, J.C., Benedicto, C.B., Trieu, L., Kendig, T., Barlow, B., 2009. Motor vehicle injury, mortality, and hospital charges by strength of graduated driver licensing laws in 36 states. J. Trauma 67 (1 Suppl.), S43–S53. Progressive Insurance, 2002. Claims survey finds there’s no place like home–for vehicle crashes, http://www.progressive.com/newsroom (retrieved). Shope, J.T., 2007. Graduated driver licensing: review of evaluation results since 2002. J. Saf. Res. 38 (2), 165–175.
117
StataCorp., 2009. Stata Statistical Software: Release 11. StataCorp. LP, College Station, TX. Traynor, T.L., 2009. The impact of state level behavioral regulations on traffic fatality rates. J. Saf. Res. 40 (6), 421–426. Vachal, K., Tumuhairwe, E.K., Berwick, M., 2009. Underride safety protection: benefit-cost assessment of rear-impact guards for the north Dakota farm truck fleet. Traffic Inj. Prev. 10 (2), 178–183. Wagenaar, A.C., Finnegan, J.R., Wolfson, M., Anstine, P.S., Williams, C.L., Perry, C.L., 1993. Where and how adolescents obtain alcoholic beverages. Public Health Rep. 108 (4), 459–464. Waller, P.F., 2003. The genesis of GDL. J. Saf. Res. 34 (1), 17–23. Williams, A.F., 2003. Teenage drivers: patterns of risk. J. Saf. Res. 34 (1), 5–15. Williams, A.F., 2007. Contribution of the components of graduated licensing to crash reductions. J. Saf. Res. 38 (2), 177–184. Williams, A.F., 2009. Licensing age and teenage driver crashes: a review of the evidence. Traffic Inj. Prev. 10 (1), 9–15. Williams, A.F., Chaudhary, N.K., 2008. Views of connecticut parents of teens and other adults about graduated licensing upgrades. Traffic Inj. Prev. 9 (6), 503–507. Williams, A.F., Ferguson, S.A., 2002. Rationale for graduated licensing and the risks it should address. Inj. Prev. 8 (Suppl. 2), ii9–ii14 (discussion pp. ii14–ii6). Williams, A.F., Ferguson, S.A., Wells, J.K., 2005. Sixteen-year-old drivers in fatal crashes, United States, 2003. Traffic Inj. Prev. 6 (3), 202–206. Williams, A.F., Nelson, L.A., Leaf, W.A., 2002. Responses of teenagers and their parents to California’s graduated licensing system. Accid. Anal. Prev. 34 (6), 835–842. Williams, A.F., Preusser, D.F., 1997. Night driving restrictions for youthful drivers: a literature review and commentary. J. Public Health Policy 18 (3), 334–345. Zhu, M., Chu, H., Li, G., 2009. Effects of graduated driver licensing on licensure and traffic injury rates in upstate New York. Accid. Anal. Prev. 41 (3), 531–535.
Contents lists available at SciVerse ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Graduated driver license nighttime compliance in U.S. teen drivers involved in fatal motor vehicle crashes Dustin Carpenter a,b,∗ , Joyce C. Pressley a,c a b c
Department of Epidemiology, Columbia University, Mailman School of Public Health, 722 West 168th Street, New York, NY 10032, USA Saint Louis University School of Medicine, 1402 South Grand Blvd., Saint Louis, MO 63104, USA Department of Health Policy and Management, Columbia University, 722 W 168th Street, New York, NY 10032, USA
a r t i c l e
i n f o
Article history: Received 13 December 2011 Accepted 15 December 2011 Keywords: Motor vehicle crash Graduated driver licensing Teen driver
a b s t r a c t Background: Examination of teen driver compliance with graduated driver licensing (GDL) laws could be instrumental in identifying factors associated with persistently high motor vehicle mortality rates. Methods: Fatality analysis reporting system (FARS) data from the years 2006 to 2009 were used in this nation-wide cross-sectional study of drivers covered by a state nighttime GDL law (n = 3492). A new definition of weekend, based on the school night in relation to the teenage social landscape, redefined Friday night as a weekend night and Sunday night as a weekday/school night and compared it to previous weekend definitions. Multiple logistic regression was used to examine independent effects of demographic, behavioral, environmental, contextual, and other factors on compliance with nighttime GDL laws. All analyses were performed in Stata version 11. Results: Given coverage under nighttime GDL laws, drivers aged 15–17 years were non-compliant in 14.9% of the fatal MVCs in which they were involved, and nearly one-fifth (18.8%) of all fatalities aged 15–17 years were associated with non-compliance. Mortality risk was 10% higher using a revised social (school night) versus traditional (Sat–Sun) weekend definitions. In multivariable analysis, drivers noncompliant with nighttime GDL laws were more likely to be drinking (OR = 4.97, 3.85–6.40), unbelted (OR = 1.58, 1.25–1.99), driving on the weekend (OR = 1.82, 1.47–2.24), and killed (OR = 1.31, 1.04–1.65). Conclusion: GDL non-compliance contributes to teen motor vehicle mortality. Legislative and enforcement efforts targeting non-school night driving, seatbelt nonuse and alcohol have potential to further reduce teen driving mortality. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Motor vehicle crashes (MVCs) are the number one killer of U.S. teens (Chen et al., 2000; Corden et al., 2009; Gonzales et al., 2005; Shope, 2007), taking the lives of approximately 3000 teens aged 15–19 and accounting for 33–40% of deaths for this age group each year (CDC, 2010b; Gonzales et al., 2005; Karaca-Mandic and Ridgeway, 2010). Two of the most important contributors to teen MVC morbidity and mortality include driving at night and presence of other teen passengers (Williams, 2003). Approximately 15% of all vehicle-miles traveled by 16–17-year-old drivers occur between 9:00 pm and 5:59 am, but this time period accounts for approximately 40% of fatal crashes in this age group (Lin and Fearn, 2003; Williams and Preusser, 1997; Williams, 2003). Nearly 50% of teen
∗ Corresponding author at: Floraville Road, Millstadt, IL 62260, USA. Tel.: +1 618 567 5560. E-mail addresses: [email protected], [email protected] (D. Carpenter), [email protected] (J.C. Pressley). 0001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2011.12.014
MVC fatalities occur with other teen passengers and no supervising adult present (Williams and Ferguson, 2002; Williams, 2003), with fatal crash risk increasing with each additional teen passenger (Chen et al., 2000). In order to mitigate the excess risk posed to and by teen drivers, every state has passed some form of graduated driver licensing (GDL) legislation. The core of GDL licensure consists of nighttime driving restrictions and underage passenger-carrying limitations. Nighttime driving restrictions decrease fatal nighttime crashes (Williams, 2007), and passenger limitations reduce mortality and the proportion of crashes with multiple teen passengers present (Masten and Hagge, 2004; Williams, 2007). The combination of laws may be more effective than either law in isolation (Chen et al., 2006; Williams, 2007), with stronger laws producing larger reductions than weak ones (Mccartt et al., 2010; Morrisey and Grabowski, 2006; Pressley et al., 2009; Traynor, 2009). Despite the dramatic lowering of teen MVC morbidity and mortality by GDL laws, the crash rate for teens remains high. Individual, behavioral and contextual factors such as alcohol involvement, restraint use, weather, and the degree of urbanization
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
111
affect driving risk in all age groups (CDC, 2010a). Recently, national studies have begun to explore how the relationship between some of these factors and GDL laws impacts teen MVC morbidity and mortality (Morrisey and Grabowski, 2006). Additionally, a limited number of studies using surveys have attempted to investigate teens’ level of reported compliance with nighttime and passenger restrictions (Goodwin and Foss, 2004; Goodwin et al., 2006; Williams et al., 2002). These studies suggest non-compliance is a significant issue and that teens are more likely to be non-compliant with passenger limitations than nighttime driving restrictions (Williams, 2007), but they do not directly reveal non-compliance’s impact and contribution to teen MVC mortality. Because passenger restrictions vary, ranging from states prohibiting the carrying of any teen passengers to allowing three or more teen passengers, the effect of non-compliance and its relationship to teen motor vehicle mortality is less clear than for nighttime driving restrictions. This study examines factors associated with nighttime GDL compliance in fatal crashes involving a teen driver. It also explores whether traditional definitions of the weekend fully capture teen weekend driving risk. An improved understanding of the relation between individual, behavioral, environmental, contextual, and state-level factors with GDL non-compliance could be instrumental in further reducing teen driving fatalities.
2.3.1. Compliance with state GDL Compliance with nighttime driving restrictions is individually coded for each state because the times when teens can legally drive varies across states (e.g., Idaho – sunrise to sunset (lifted at 16 years of age); Alaska – 1–5 am (lifted at 16 years and 6 months of age); Vermont – no nighttime restriction (thus excluded); see Insurance Institute for Highway Safety (2011) for further details). FARS data contains variables to determine compliance: age of driver, state of driver residence, and time and date of crash. States that passed laws during the study timeframe had teens who were listed as covered or not covered in relation to when the law went into effect. State-specific nighttime restrictions and variables for time of crash, and, in some cases, day of the week of the crash were used to assess whether a crash occurred during restricted hours. A crash was categorized as non-compliant with nighttime GDL laws if it occurred outside of legally permitted driving times. A teen driver was considered non-compliant with passenger GDL restrictions if more passengers of the specified ages were present than allowed by the state-specific law.
2. Materials and methods
2.3.2. Demographic variables Age and gender were obtained for each driver and passenger involved in each MVC using uniform FARS coding.
2.1. Data sources Analyses for traffic fatalities and relevant covariates for the years 2006–2009 utilized data obtained from the National Highway Traffic Safety Administration’s (NHTSA) fatality analysis reporting system (FARS). FARS is a motor vehicle crash (MVC) census database that contains every traffic crash on public roads that results in a fatality of a motorist or non-motorist within 30 days of the crash in all 50 states, Washington D.C., and Puerto Rico. FARS is a publically accessible, de-identified database (NHTSA, 2009). Intercensal denominator estimates were acquired from the Center for Disease Control and Prevention’s (CDC) web-based injury statistics query and reporting system (WISQARS) for total population by age and state. WISQARS population data are taken from the United States Census Bureau and the BridgedRace Post-censal Population Estimates of the United States, which is a collaboration of the U.S. Census Bureau and the National Center for Health Statistics (NCHS) updated annually (www.cdc.gov/nchs/about/major/dvs/popbridge/popbridge.htm). The years 2006–2007 are the most recently available data for WISQARS. In order to obtain population figures for the years 2008 and 2009, the percent change from 2006 to 2007 was used to estimate the population for 2008–2009 (CDC, 2010a,b).
2.2. Study population The study population was comprised of fatal MVCs involving at least one teen driver between the ages of 15 and 17 years who was covered by a state nighttime GDL law (n = 3492). Since FARS includes data on all MVCs that result in a fatality, the drivers who comprise our study population did not necessarily die in the crash (i.e. the death could have been a passenger in the driver’s vehicle, who was not included in our study population). FARS data provides age in increments of one year (i.e. integers), but some states lift their GDL provisions at non-integer time-points. At the time of this
study, eleven states had laws that precluded inclusion of at least one age group, resulting in exclusion of 544 teen driver crashes.1 2.3. Variable classications
2.3.3. Crash fatality In order to determine driver case fatality rates (CFR), given involvement in a crash, a variable was dichotomized into fatal and non-fatal injury. Fatal injuries comprised the numerators and were used with population census data to establish age-specific mortality rates per 100,000 population. 2.3.4. Teen residence/crash location The variable, “road function”, identifies crash location as rural or urban based on population density data in the area surrounding the crash. 2.3.5. Interstate highways FARS provides uniform coding for dichotomizing crashes that occur on an interstate or non-interstate highway, providing an indicator for roadways with varying levels of access. 2.3.6. Weather The FARS variable for adverse road conditions such as rain, sleet, snow, fog, mist, and severe crosswinds or other conditions that existed at the time of crash was dichotomized, indicating presence or absence of adverse weather conditions. 2.3.7. Day of the week FARS coding for day of the week and time of day was used to create a dichotomous variable for weekday/weekend. Three definitions of weekend were examined in this study to assess which more accurately reflects crash risk among teen drivers: (1) “traditional weekend” was coded as “weekend” for the days Saturday and Sunday, while the remainder of the week was coded as “weekday”; (2) “expanded weekend” (from the 2009 FARS codebook) defines the weekend as the time period from 6:00 pm on Fridays until 5:59 am
1 Of note, Missouri lifts its GDL at 17 years and 11 months, and, in order not to lose its data, all 17-year olds were considered covered under its provision.
112
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
on Monday mornings; and (3) “social weekend” defines weekends as the period from Friday at 5:00 pm through 4:59 pm Sunday. 2.3.8. Alcohol involvement The presence of alcohol was assessed using a derived composite variable, “dr drink”, that incorporates charges filed by police, police-reported alcohol involvement and alcohol tests that note presence of sufficient evidence to consider alcohol as a contributing factor in the crash. Presence of alcohol is coded as a dichotomous variable. 2.3.9. Seatbelt restraints The FARS variable, “restraint use”, indicates seat belt use as present or absent at the time of crash.
“weekend + no adult passenger present”, and “interstate highway + driving over the speed limit”. Interaction terms were added one at a time into the final multivariable model. Significance was assessed using the likelihood ratio test and the associated p-value of the interaction term with the cut-off = 0.05. 3.1. Traditional, expanded, and social weekend In order to assess the impact that each weekend definition has for explaining the risk of MVC death among teenagers, an adjustment for differences in length of time covered by the weekend variables was made to facilitate comparisons between the three definitions. All data was coded, transformed, and analyzed using Stata version 11.
2.3.10. Speeding A variable indicating whether or not a driver was speeding at the time of the crash utilized FARS’ default coding.
4. Results
2.3.11. Number of passengers The number of passengers in the vehicle, besides the driver, is derived from the FARS variable for number of vehicle occupants, with the number of occupants minus one to account for the driver. Youth passengers are defined as those aged 25 years or younger, and adult passengers are those aged 25 years and older. This definition was used because studies have shown that older teens and young adults increase access to alcohol for younger teens (Wagenaar et al., 1993) and increase crash risk (Chen et al., 2000). Insurers use a similar definition for driver age because of the aforementioned risk associated with younger drivers and passengers (www.iihs.org).
There were 20,967 teens aged 15–17 years involved in fatal MVCs between 2006 and 2009, of which 7415 were drivers. Of the drivers, coverage with state nighttime GDL laws could be determined for 6871 (92.7%) cases. Of these, 3492 (50.8%) were covered by a state nighttime GDL law. These teens comprise the study population. The population characteristics of teen drivers covered by a nighttime GDL law are shown by compliance status in Table 1. Coverage by a nighttime GDL provision differed significantly by age, with coverage decreasing with each increasing year of age for 15-, 16-, and 17-year olds (90.1% vs. 84.4% vs. 29.5%). Of the 3492 teens covered by a nighttime GDL law, 520 (14.9%) were in violation of nighttime GDL restrictions at the time of the fatal crash. Of the 1464 (41.9%) who died in crashes, 275 (18.8%) were non-compliant with their nighttime GDL restriction.
2.3.12. Seat belt Law A variable was dichotomized to indicate whether or not drivers were covered by a primary seat belt law as reported by the Insurance Institute of Highway Safety (www.iihs.org).
4.1. Study population
4.2. Risk factors 2.3.13. Fatality rate in person-years Fatality rates for the three definitions of weekend were calculated using person-years as the denominator. The age-specific population for each year was summed to determine the number of person-years each age group contributed over the study period. 3. Statistical analysis Given coverage by a nighttime GDL provision, the proportion of teen drivers and teen fatalities aged 15–17 involved in fatal MVCs while drivers were non-compliant with their state’s GDL law was calculated. These proportions are reported for the nighttime driving restriction by key covariates. Chi-squared tests, chi-squared test for trends and Student’s t-test were used to assess the statistical significance of categorical variables and group means, respectively, between compliant and non-compliant drivers. Significance is defined as p-values ≤0.05. We based our model building strategy on the recommendations in Hosmer and Lemeshow (2010)’s Applied Logistic Regression. Variables hypothesized to have an association with compliance with GDL laws were analyzed univariately with the outcome of interest, non-compliance with the nighttime GDL. Variables with p-values ≤0.25 were considered for the final multivariable logistic regression model. Correlations were then assessed for all predictor variables of interest. Two variables, “not wearing seat belt” and “death in the crash”, had an ‘r’ of 0.30, and were ultimately left in the final model because of their epidemiological significance. Several interaction terms were considered, including “weekend + drinking and driving”, “gender + drinking and driving”, “drinking and driving + not wearing a seat belt”, “gender + not wearing a seat belt”,
4.2.1. Age Drivers aged 17 years were more likely to drink and drive than 15- or 16-year-old drivers (12.6%, 11.5%, and 9.8%, 2 for trend by age = 6.0, p = 0.049). Younger teens were more likely to be unbelted (47.9%, 34.0% and 36.8%, 2 for trend by age = 29.6, p < 0.001), speeding (39.1%, 34.7% and 29.8%, 2 for trend by age = 15.7, p < 0.001), involved in single vehicle crashes (83.4%, 83.2% and 79.2%, 2 for trend by age = 11.3, p = 0.024), and die in their crashes (44.9%, 43.7% and 38.8%, 2 for trend by age = 9.0, p = 0.011) compared to older teen drivers. Younger drivers were also more likely to have a passenger older than 25 at the time of crash (16.0%, 6.8% and 4.8%, 2 for trend = 71.4, p < 0.001), but driver age was not related to weekend vs. weekday crash (2 for trend = 1.3, p = 0.53) or driving with more than two youth passengers (2 for trend = 1.8, p = 0.41). 4.2.2. Gender Male drivers were more likely than female drivers to be drinking and driving (13.0% vs. 7.1%, 2 = 27.7, p < 0.001), unbelted (41.1% vs. 28.8, 2 = 47.5, p < 0.001), and speeding (37.1% vs. 26.3%, 2 = 40.9, p < 0.001). However, male drivers were not significantly more likely to die in their crashes (2 = 2.5, p = 0.12), crash on the weekend (2 = 2.5, p = 0.12) or to have been driving with more than two youth passengers (2 = 0.76, p = 0.38) compared to female drivers. 4.2.3. Traditional, expanded, and social weekend Three weekend definitions and their effect on measuring teen mortality risk are examined in Table 2. In our study population, the social definition of the weekend had a 10.1% and 10.3% higher mortality rate compared to the traditional and expanded weekend
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
113
Table 1 Population characteristics for 15–17-year-old drivers complaint and non-compliant with nighttime GDL. Total covered by nighttime GDL (n = 3478) Non-compliant (n = 520) Compliant (n = 2958)
Total (n = 3478)
p-Value
Personal characteristics Age 15 16 17 Male (male) Fatality (died)
388 (13.1%) 1489 (50.3%) 1081 (36.6%) 1935 (65.4%) 1189 (40.3%)
87 (16.7%) 255 (49.1%) 178 (34.2%) 390 (75.0%) 275 (53.0%)
475 (13.7%) 1744 (50.1%) 1259 (36.2%) 2325 (66.9%) 1464 (42.2%)
<0.001 <0.001
Environmental characteristics Weather (bad weather) Rural (rural) Interstate highway
302 (10.2%) 1766 (60.2%) 410 (14.1%)
45 (8.7%) 258 (49.8%) 93 (18.0%)
347 (10.0%) 2024 (58.7%) 503 (14.7%)
0.282 <0.001 0.02
171 (32.9%) 253 (53.2%) 224 (43.1%) 1.26 (1.43) 1.11 (1.32) 145 (27.9%) 0.07 (0.36) 25 (4.8%) 79 (27.0%)
386 (11.1%) 1196 (37.0%) 1168 (33.6%) 1.06 (1.31) 0.93 (1.20) 824 (23.7%) 0.08 (0.36) 249 (7.2%) 401 (21.8%)
<0.001 <0.001 <0.001 <0.001 <0.001 0.015 0.138 0.024 0.02
Behavioral characteristics Drinking and driving Not wearing seatbelt Speeding Mean number of passengers (SD) Mean number pass ≤25 (SD) Increased youth pass (GE 2 pass ≤25) Mean number of adult pass >25 (SD) Adult passenger present (>25) Non-compliance with passenger GDL
215 (7.3%) 943 (34.2%) 944 (31.9%) 1.04 (1.25) 0.9 (1.18) 679 (23.0%) 0.09 (0.36) 224 (7.6%) 322 (20.9%)
0.081
Contextual characteristics Social weekend (5 pm Fri to 5 pm Sun) Traditional weekend (Sat–Sun)
1069 (36.1%) 919 (31.1%)
281 (54.0%) 274 (52.7%)
1350 (38.8%) 1193 (34.3%)
<0.001 <0.001
State-level laws Primary seatbelt law
1247 (42.9%)
219 (42.9%)
1466 (42.9%)
0.99
Table 2 Impact of different definitions of the weekend on fatality rates. Social weekend
Traditional weekend
Expanded weekend
Crashes involving 15–17-year-old drivers Number of 15–17-year-old fatalities over study period Person-years for 15–17-year olds for 2006–2009/100,000 Proportion of time each definition captures Fatality rate
1808 (39.8%) 523 0.286 12.1/100,000/person-year
1625 (35.8%) 523 0.286 10.9/100,000/person-year
2027 (44.7%) 523 0.357 10.9/100,000/person-year
Crashes involving all-age drivers Number of 15–17-year-old fatalities over study period Person-years for 15–17-year olds for 2006–2009/100,000 Proportion of time each definition captures Fatality rate
2905 (40.9%) 523 0.286 19.1/100,000/person-year
2651 (37.4%) 523 0.286 17.7/100,000/person-year
3322 (46.8%) 523 0.357 17.8/100,000/person-year
160 140 120
Friday Saturday Mon-Thurs Sunday
100 80 60 40 20 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
4.2.4. Alcohol, speeding, and seatbelts Drivers who were speeding (50.8% vs. 30.1%, 2 = 132.9, p < 0.001) or drinking and driving (64.7% vs. 33.5%, 2 = 135.5, p < 0.001) were more likely to be unbelted than drivers obeying speed limit laws or not drinking, respectively. Drivers who crashed over the social weekend had a 29% higher proportion of their crashes involve alcohol compared to weekday crashes (14.7% vs.
8.8%, 2 = 29.6, p < 0.001); however, seat belt wearing did not differ significantly between weekend and weekday crashes (37.3% vs. 36.7% unbelted, 2 = 0.13, p = 0.71). Drivers carrying more than two
Number of Fatalities
definitions, respectively, adjusted for the number of hours captured by each definition. The social definition captures 183 more deaths than the traditional definition, even though they are both 48 h time periods. Over one-half (53.5%) of all non-compliant crashes took place between 6:00 pm on Friday night and 5:59 am on Saturday morning or between 6:00 pm on Saturday night and 5:59 am on Sunday morning (278/520). Friday night to Saturday morning had 2.3 times more non-compliant crashes (126 vs. 55) than the period from Sunday night to Monday morning. Of note, the majority of non-compliant crashes that occurred between Sunday night and Monday morning took place during summer or holiday weekends (i.e. when Sunday night was not a school/weeknight). Figs. 1 and 2 compare teen driver fatalities on Sunday to other weekday and weekend days and nights.
Hour (Military Time) Fig. 1. 15–17-year-old driver fatalities by hour of the day.
114
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
160 Friday Sunday
Number of Fatalities
140 120 100 80 60
4.3.5. Weather, day of the week, and seat belt laws Non-compliant drivers were 20% more likely to crash on urban roads (50.2% vs. 39.8%, 2 = 19.7, p < 0.001), 22% more likely to crash on interstate highways (18.0% vs. 14.1%, 2 = 5.4, p = 0.02), and one-third more likely to crash on the weekend (54.0% vs. 36.1%, 2 = 59.7, p < 0.001) compared to compliant drivers. Neither weather conditions (2 = 1.2, p = 0.28) nor having a primary statelevel seatbelt law targeting teen drivers (2 = 0.0002, p = 0.99) were associated with compliance (Table 1).
40 20
4.4. Multivariable logistic regression m
m 5a
4a
m
m
m 3a
2a
1a
am
pm
m
pm
12
11
10
9p
m
m 8p
7p
6p
m
0
Hour Fig. 2. Friday night vs. Sunday night.
passengers below the age of 25 were more likely to drink and drive (13.5% vs. 10.3%, 2 = 6.7, p = 0.009) and speed (51.9% vs. 30.9%, 2 = 34.2, p < 0.001) compared to drivers who were carrying one or fewer passengers, but increased passengers was not associated with driver seat belt use (2 = 0.2, p = 0.89). 4.3. Nighttime compliance and risk factors 4.3.1. Age Compliance with nighttime GDL provisions was not statistically different by age (18.3%, 14.6% and 14.1%, 2 for trend = 5.0, p = 0.081), except for the difference between 15- and 17-year olds (2 = 4.6, p=.03). Irrespective of age, non-compliant drivers were 24% more likely to die in their crash compared to compliant drivers (53.0% vs. 40.3%, p < 0.001). 4.3.2. Gender Given coverage by a nighttime GDL provision, 16.8% of all male drivers and 11.3% of all female drivers were non-compliant with their states’ nighttime GDL (2 = 18.3, p < 0.001). Also of note, 15-year-old drivers were significantly more likely to be male than 16–17-year-old drivers, especially in non-compliant crashes (2 = 15.6, p < 0.001) (Table 1). 4.3.3. Alcohol, speeding, and seatbelts Compliant and non-compliant drivers differed by several behavioral factors (Table 1). Non-compliant drivers were more likely to be drinking and driving, speeding, and unbelted compared to compliant drivers. 4.3.4. Passengers Crashes with two or more youth passengers present in the vehicle were 30% more likely to be non-compliant than crashes with 1 or fewer youth passengers (17.6% vs. 14.1%, 2 = 5.9, p = 0.02). For drivers covered by both nighttime and passenger GDL restrictions, drivers non-compliant with the passenger GDL were 40% more likely to be non-compliant with the nighttime GDL compared to teen drivers compliant with the passenger restriction (19.7% vs. 14.9%, 2 = 5.49, p = 0.02). Of note, 27% of crashes were non-compliant with either the nighttime or the passenger GDL restriction (when teens were covered by both types of law and compliance with each coverage could be assessed). The odds of nighttime non-compliance increased 13% with each additional passenger present in the vehicle. Presence of an adult passenger greater than 25 years of age was associated with higher nighttime GDL compliance (2 = 5.1, p = 0.02).
The unadjusted and fully adjusted ORs, with 95% confidence intervals and associated p-values for each of the covariates are reported in Table 3. After controlling for the other listed covariates, non-compliant drivers were more likely to have been drinking alcohol, speeding, unbelted, crash on an urban road and interstate highway, crash during the weekend, and die in their crash compared to compliant drivers. Although not statistically significant in the multivariable model, non-compliant drivers also tended to be younger, male, speeding, and carrying two or more youth passengers below 25 years of age compared to compliant drivers. Risk factors varied by gender and age for drivers covered by a nighttime GDL. Therefore, several interaction terms were assessed in the final multivariable model. Despite the differences between characteristics of teen drivers and the circumstances surrounding their crashes, all of the interaction terms were insignificant when entered sequentially into the final multivariable model. All analyses were performed in Stata version 11 (StataCorp., 2009). 5. Discussion The goal of graduated driver licensing legislation is to decrease injury and fatality rates by reducing teens’ exposure when they are most at risk, while valuable driving experience is accumulated under lower-risk conditions (Hyde et al., 2005; Waller, 2003). Past studies have shown that GDLs are effective in lowering statelevel mortality rates (Chen et al., 2006; Dee et al., 2005; Ferguson, 2003; Fohr et al., 2005; Kirley et al., 2008; Males, 2007; Margolis et al., 2007; Masten and Foss, 2010; Neyens et al., 2008; Williams et al., 2005; Zhu et al., 2009), with stronger laws producing larger reductions in fatal crash rates compared to weaker laws (Mccartt et al., 2010; Morrisey and Grabowski, 2006; Pressley et al., 2009; Traynor, 2009). However, teen fatality rates remain significantly higher than those of other age groups. Despite this, GDL laws remain heterogeneous, with many states having vastly different nighttime restrictions and passenger-carrying limitations. A major difficulty in writing and legislating teenage traffic laws is retaining the balance between teens’ increased risk of death, which is well-documented, and teens’ need for mobility. This is reflected in the fact that the most rural states, whose teens often have the fewest public transportation options, tend to have the least strict GDL laws (Morrisey and Grabowski, 2006). However, there is a small but growing pool of data that suggests, at least through age 17, that nighttime and passenger laws do not have a significantly negative impact on teens’ quality of life (Williams, 2009). This suggests that the benefits of GDL laws likely outweigh the societal costs. Many states have successfully implemented stricter GDL laws without compromising teens’ quality of life (Williams, 2009). Survey evidence from California found that nearly three-quarters of teens reported not having their quality of life significantly affected by their nighttime or passenger restrictions (Williams et al., 2002), and a study in New Jersey, which is the sole state to enforce an intermediate license through the age of 17, found similar
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
115
Table 3 Univariate and multivariate analysis for non-compliance with nighttime GDL. Univariate Odds ratio (95% C.I.)
Multivariable p-Value
Adjusted odds ratio (95% C.I.)
p-Value
Personal characteristics Age 16 vs 17 15 vs 17 Male Fatality
0.95 (0.79–1.14) 1.33 (1.03–1.72) 1.59 (1.28–1.96) 1.67 (1.39–2.02)
0.585 0.027 <0.001 <0.001
1.26 (0.996–1.60) 1.66 (1.21–2.30) 1.27 (0.998–1.60) 1.31 (1.04–1.65)
Environmental characteristics Bad weather Rural Interstate highway
0.84 (0.60–1.16) 0.66 (0.54–0.79) 1.34 (1.05–1.72)
0.282 <0.001 0.02
− 0.55 (0.44–0.68) 1.43 (1.08–1.90)
− <0.001 0.014
6.25 (4.97–7.86) 2.18 (1.79–2.66) 1.61 (1.34–1.95) 1.30 (1.05–1.60) 1.14 (1.06–1.23) 0.79 (0.59–1.08) 0.62 (0.40–0.94) 1.13 (1.06–1.21) 1.40 (1.05–1.86)
<0.001 <0.001 <0.001 0.0163 <0.001 0.14 0.025 <0.001 0.021
4.97 (3.85–6.40) 1.58 (1.25–1.99) 1.18 (0.94–1.47) 1.16 (0.91–1.49) − − 0.80 (0.51–1.27) − −
<0.001 <0.001 0.145 0.224 − − 0.347 − −
2.08 (1.72–2.51)
<0.001
1.82 (1.47–2.24)
<0.001
1.00 (0.83–1.21)
0.99
Behavioral characteristics Drinking and driving Not wearing seatbelt Speeding Increased youth pass (≥2 pass) Number of youth passengers (≤25) Number of adult passengers (>25) Adult passenger present (>25) Number of passengers Non-compliance with passenger GDL Contextual characteristics Social weekend (5 pm Fri to 5 pm Sun) State-level laws Primary seatbelt law
results (Williams, 2009). Furthermore, surveys of affected teens’ parents from Pennsylvania, North Dakota, New Jersey, California and Connecticut have suggested widespread parental support, with large majorities supporting current level or even more stringent restrictions, with few parents neutral or opposed to their child’s restrictions (Mckay et al., 2008; Vachal et al., 2009; Williams et al., 2002; Williams and Chaudhary, 2008). Nonetheless, survey evidence also suggests that compliance with restrictions is far from 100%. Our study documented the contribution of nighttime noncompliance to teen motor vehicle mortality and associated risk factors. In our study, given coverage with a nighttime GDL law, nearly one-fifth of all 15–17-year old drivers’ fatalities (18.8%) occurred while they were non-compliant. Of all the risk factors associated with non-compliance, the behavioral-level factors such as drinking and driving, speeding, and seatbelt use showed the strongest association within our study population. Non-compliant crashes predominately occurred on the weekend, especially Friday and Saturday nights, making the examination of non-compliant crashes with respect to the school year more effective in capturing teen crash risk than using traditional definitions of the weekend. These results indicate that while weekend driving is associated with increased numbers of fatal MVCs for teen drivers, it is especially associated with non-compliant driving, in which a 24 h period (Friday night to Saturday morning and Saturday night to Sunday morning) accounted for over one-half of all fatal non-compliant crashes. These findings further suggest that compliance enforcement dollars would be best spent targeting drivers on non-school nights. The proportion of non-compliant crashes we found comports with survey evidence. In North Carolina, 23% of teens reported violating their nighttime GDL restrictions either with or without parental permission; however, most teens admitted to only infrequent violations (Goodwin and Foss, 2004). In Nova Scotia, 12% of teens admitted to frequently violating their nighttime GDL restriction (Mayhew et al., 2003), which accords with data from California, in which 15% of teens admitted to violating their nighttime restrictions more than 10 times (Williams et al., 2002).
−
0.054 0.002 0.052 0.023
−
In our study population (i.e. teens covered by a nighttime GDL restriction), when a passenger GDL restriction was also in place, 21.8% of teen drivers were violating it at the time of fatal crash occurrence. Other studies have also reported finding that teens disobey passenger restrictions more frequently than nighttime restrictions. For example, in North Carolina 34% of teens reported violating their passenger restriction either with or without parental permission (Goodwin and Foss, 2004), and, in California, 60% of teens admitted to violating it more than 10 times (Williams et al., 2002). Differences in compliance in relation to nighttime and passenger GDL restrictions are likely due to the heterogeneity of passenger GDL restrictions, in which some states forbid any number of passengers whereas others allow up to 3 passengers. Thus, some states allow for teen vehicles to be filled with other teens as passengers and still be considered compliant with their passenger GDL, while, in other states, teen vehicles could have a single passenger and be deemed non-compliant. After controlling for the number of hours included in the definitions, our newly defined weekend definition exhibited greater ability to capture risk than the traditional and expanded weekend definitions. Comparing the social definition versus the traditional definition, we see that the difference results from substituting the lower crash risk on Sunday night for the higher risk of Friday night. We found that incorporating Friday night more accurately reflects the teenage social landscape, in which Friday night is a popular night for group gatherings, including high-school football games, house parties, and other events. This is reflected by the fact that Friday has a much higher crash count compared to Sunday night. We feel this is due in part to at least two phenomena: (1) that driving on Friday night produces more exposure through larger numbers of teens on roadways, and (2) other risky behaviors like drinking alcohol are more prevalent on Friday nights. We excluded Sunday night from our definition because we felt it is traditionally used to rest and recharge for another schoolweek, unlike Friday night. The expanded definition of the weekend includes Friday night; however, it also extends to 5:59 am on Monday morning. Including this latter time period dilutes the increased risk we see when only including Friday night in the definition.
116
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117
Examining drivers who were covered by nighttime GDL laws, the behavioral factors were most strongly related to noncompliance – especially drinking and driving, driving on the weekend, and being unrestrained. These factors’ relationship to nighttime GDL non-compliance suggests several potential points of intervention. Strengthening seat belt and alcohol laws targeting teen drivers, as well as their enforcement, in conjunction with increased GDL enforcement on the weekends has the potential to increase compliance with nighttime GDLs and to mitigate the effects of non-compliance. Previous studies have found that the passage of GDL laws has been associated with increased seat belt usage (Hyde et al., 2005), and our findings suggest that being unbelted is significantly associated with nighttime GDL non-compliance. These findings serve to underscore the importance of the Centers for Disease Control and Prevention’s recommendations for what can be done to combat motor vehicle mortality in teen drivers, including consistent and strong three-phase GDL laws in every state, primary seatbelt laws targeting teen drivers, as well as minimum legal drinking age and zero tolerance laws for drivers under the age of 18 years (http://www.cdc.gov/Motorvehiclesafety/teenbrief/index.html) (CDC, 2011). This study has several limitations. First, there was a small number of drivers for whom there was an inability to determine compliance with nighttime GDL laws, and they were therefore excluded from the study. We feel it is unlikely that this greatly biased our findings since there is no evidence that compliance for the indeterminate category varied systematically from the teens in the study. However, to the extent that excluded teens differ from teens in the study, this could be a source of bias. One state may have included a small number of misclassified teens due to our including all 17 year olds even though its restriction was lifted at 17 years and 11 months of age. Although studies have shown that teens from urban and rural areas have different crash risks (Donaldson et al., 2006), our variable for crash location was unable to differentiate between crash location and teen residence in urban or rural areas. It is possible that non-compliant crashes were significantly more urban than compliant crashes because areas considered “urban” were more likely weekend destinations for all teens. Furthermore, because non-compliant crashes were more likely to occur during the weekend, they may appear to more greatly impact urban teens, when in fact many of these crashes were actually sustained by rural drivers, who travel into urban areas to congregate socially. We suspect that this effect explains at least a portion of the discrepancy observed in crash location; however, the magnitude of this misclassification is unknown. Insurance claims data suggest that most crashes occur within five miles of home (Progressive, 2002), however, it is unclear if this is the case for teen drivers or if this varies by day of the week. We were also unable to determine if teens were driving with members of their family at the time of their crash, which may have resulted in some teens inappropriately being considered non-compliant in some states. Furthermore, some rural teens are exempt from certain GDL restrictions, which could not be captured in our study. Finally, this study captures the proportion of teen drivers and teen fatalities that were involved in a fatal MVC, not the actual rate of non-compliance.
6. Conclusion Non-compliance with state-specific nighttime GDLs is a significant source of mortality for U.S. teens. Continued action by parents, police, researchers and policy-makers is needed to adequately address this challenge. Adoption of the social definition
of the weekend would more accurately describe teenagers’ social landscape and capture the risk associated with teen driving . A more developed understanding of the level and characteristics of GDL non-compliance as well as the accurate assessment of teen drivers’ weekend risk would allow for more focused policy analysis and facilitate decision-making with regard to targeting potential GDL improvements. The findings of this study suggest that GDL enforcement dollars would be best spent targeting weekend driving, alcohol and seat belt use in teen drivers.
Acknowledgments This manuscript was awarded a “Best Student Paper Award” from the Injury Control and Emergency Health Services Section of the American Public Health Association, 2011. The award was sponsored, in part, by the Liberty Mutual Research Institute for Safety. Contents of this effort are solely the responsibility of the authors and do not necessarily represent the official view of the American Public Health Association or the Liberty Mutual Research Institute for Safety. The authors would like to thank Dr. Barbara Barlow, M.D., who graciously read several drafts of an earlier work related to this paper.
References Centers for Disease Control and Prevention, 2010a. Drivers aged 16 or 17 years involved in fatal crashes—United States, 2004–2008. MMWR Morb. Mortal. Wkly. Rep. 59 (41), 1329–1334. Centers for Disease Control and Prevention, 2010b. Web-Based Injury Statistics Query and Reporting System (WISQARS). National Center for Injury Prevention and Control, Centers for Disease Control and Prevention (online). Centers for Disease Control and Prevention, 2011. Policy Impact: Teen Driver Safety. National Center for Injury Prevention and Control, Division of Unintentional Injury Prevention. Chen, L.H., Baker, S.P., Braver, E.R., Li, G., 2000. Carrying passengers as a risk factor for crashes fatal to 16- and 17-year-old drivers. JAMA 283 (12), 1578–1582. Chen, L.H., Baker, S.P., Li, G., 2006. Graduated driver licensing programs and fatal crashes of 16-year-old drivers: a national evaluation. Pediatrics 118 (1), 56–62. Corden, T.E., Tonellato, D.J., Frisch, K.B., Laud, P.W., 2009. Graduated driver licensing policy in the great lakes states: current benefits and future potential. WMJ 108 (8), 393–397. Dee, T.S., Grabowski, D.C., Morrisey, M.A., 2005. Graduated driver licensing and teen traffic fatalities. J. Health Econ. 24 (3), 571–589. Donaldson, A.E., Cook, L.J., Hutchings, C.B., Dean, J.M., 2006. Crossing county lines: the impact of crash location and driver’s residence on motor vehicle crash fatality. Accid. Anal. Prev. 38 (4), 723–727. Ferguson, S.A., 2003. Other high-risk factors for young drivers—how graduated licensing does, doesn’t, or could address them. J. Saf. Res. 34 (1), 71–77. Fohr, S.A., Layde, P.M., Guse, C.E., 2005. Graduated driver licensing in wisconsin: does it create safer drivers? WMJ 104 (7), 31–36. Gonzales, M.M., Dickinson, L.M., Diguiseppi, C., Lowenstein, S.R., 2005. Student drivers: a study of fatal motor vehicle crashes involving 16-year-old drivers. Ann. Emerg. Med. 45 (2), 140–146. Goodwin, A.H., Foss, R.D., 2004. Graduated driver licensing restrictions: awareness, compliance, and enforcement in north carolina. J. Saf. Res. 35 (4), 367–374. Goodwin, A.H., Wells, J.K., Foss, R.D., Williams, A.F., 2006. Encouraging compliance with graduated driver licensing restrictions. J. Saf. Res. 37 (4), 343–351. Hosmer, D.W., Lemeshow, S., 2010. Applied Logistic Regression, 2nd ed. WileyInterscience Publication. Hyde, L.K., Cook, L.J., Knight, S., Olson, L.M., 2005. Graduated driver licensing in Utah: is it effective? Ann. Emerg. Med. 45 (2), 147–154. Institute for Highway Safety, Young driver licensInsurance ing systems in the U.S.: intermediate & unrestricted stages, http://www.iihs.org/laws/gdl intermediate.aspx (retrieved). Karaca-Mandic, P., Ridgeway, G., 2010. Behavioral impact of graduated driver licensing on teenage driving risk and exposure. J. Health Econ. 29 (1), 48–61. Kirley, B.B., Feller, A., Braver, E., Langenberg, P., 2008. Does the maryland graduated driver licensing law affect both 16-year-old drivers and those who share the road with them? J. Saf. Res. 39 (3), 295–301. Lin, M.L., Fearn, K.T., 2003. The provisional license: nighttime and passenger restrictions—a literature review. J. Saf. Res. 34 (1), 51–61. Males, M., 2007. California’s graduated driver license law: effect on teenage drivers’ deaths through 2005. J. Saf. Res. 38 (6), 651–659. Margolis, L.H., Masten, S.V., Foss, R.D., 2007. The effects of graduated driver licensing on hospitalization rates and charges for 16-and 17-year-olds in north Carolina. Traffic Inj. Prev. 8 (1), 35–38.
D. Carpenter, J.C. Pressley / Accident Analysis and Prevention 56 (2013) 110–117 Masten, S.V., Foss, R.D., 2010. Long-term effect of the north carolina graduated driver licensing system on licensed driver crash incidence: a 5-year survival analysis. Accid. Anal. Prev. 42 (6), 1647–1652. Masten, S.V., Hagge, R.A., 2004. Evaluation of California’s graduated driver licensing program. J. Saf. Res. 35 (5), 523–535. Mayhew, D.R., Simpson, H.M., Desmond, K., Williams, A.F., 2003. Specific and longterm effects of Nova Scotia’s graduated licensing program. Traffic Inj. Prev. 4 (2), 91–97. Mccartt, A.T., Teoh, E.R., Fields, M., Braitman, K.A., Hellinga, L.A., 2010. Graduated licensing laws and fatal crashes of teenage drivers: a national study. Traffic Inj. Prev. 11 (3), 240–248. Mckay, M.P., Coben, J.H., Larkin, G.L., Shaffer, A., 2008. Attitudes of teenagers and their parents to Pennsylvania’s graduated driver licensing system. Traffic Inj. Prev. 9 (3), 217–223. Morrisey, M.A., Grabowski, D.C., 2006. Graduated drivers license programs and rural teenage motor vehicle fatalities. J. Rural Health 22 (4), 300–307. National Highway Traffic Safety Administration, 2009. Fatality Analysis Reporting System (FARS). National Center for Statistics and Analysis, National Highway Traffic Safety Administration, available from URL http://www.nhtsa.gov/FARS (online). Neyens, D.M., Donmez, B., Boyle, L.N., 2008. The Iowa graduated driver licensing program: effectiveness in reducing crashes of teenage drivers. J. Saf. Res. 39 (4), 383–390. Pressley, J.C., Benedicto, C.B., Trieu, L., Kendig, T., Barlow, B., 2009. Motor vehicle injury, mortality, and hospital charges by strength of graduated driver licensing laws in 36 states. J. Trauma 67 (1 Suppl.), S43–S53. Progressive Insurance, 2002. Claims survey finds there’s no place like home–for vehicle crashes, http://www.progressive.com/newsroom (retrieved). Shope, J.T., 2007. Graduated driver licensing: review of evaluation results since 2002. J. Saf. Res. 38 (2), 165–175.
117
StataCorp., 2009. Stata Statistical Software: Release 11. StataCorp. LP, College Station, TX. Traynor, T.L., 2009. The impact of state level behavioral regulations on traffic fatality rates. J. Saf. Res. 40 (6), 421–426. Vachal, K., Tumuhairwe, E.K., Berwick, M., 2009. Underride safety protection: benefit-cost assessment of rear-impact guards for the north Dakota farm truck fleet. Traffic Inj. Prev. 10 (2), 178–183. Wagenaar, A.C., Finnegan, J.R., Wolfson, M., Anstine, P.S., Williams, C.L., Perry, C.L., 1993. Where and how adolescents obtain alcoholic beverages. Public Health Rep. 108 (4), 459–464. Waller, P.F., 2003. The genesis of GDL. J. Saf. Res. 34 (1), 17–23. Williams, A.F., 2003. Teenage drivers: patterns of risk. J. Saf. Res. 34 (1), 5–15. Williams, A.F., 2007. Contribution of the components of graduated licensing to crash reductions. J. Saf. Res. 38 (2), 177–184. Williams, A.F., 2009. Licensing age and teenage driver crashes: a review of the evidence. Traffic Inj. Prev. 10 (1), 9–15. Williams, A.F., Chaudhary, N.K., 2008. Views of connecticut parents of teens and other adults about graduated licensing upgrades. Traffic Inj. Prev. 9 (6), 503–507. Williams, A.F., Ferguson, S.A., 2002. Rationale for graduated licensing and the risks it should address. Inj. Prev. 8 (Suppl. 2), ii9–ii14 (discussion pp. ii14–ii6). Williams, A.F., Ferguson, S.A., Wells, J.K., 2005. Sixteen-year-old drivers in fatal crashes, United States, 2003. Traffic Inj. Prev. 6 (3), 202–206. Williams, A.F., Nelson, L.A., Leaf, W.A., 2002. Responses of teenagers and their parents to California’s graduated licensing system. Accid. Anal. Prev. 34 (6), 835–842. Williams, A.F., Preusser, D.F., 1997. Night driving restrictions for youthful drivers: a literature review and commentary. J. Public Health Policy 18 (3), 334–345. Zhu, M., Chu, H., Li, G., 2009. Effects of graduated driver licensing on licensure and traffic injury rates in upstate New York. Accid. Anal. Prev. 41 (3), 531–535.