Abstract
Purpose
Although motor vehicle crashes (MVC) are the leading cause of death for adolescents, there is a scarcity of research addressing adolescents’ lack of pre-licensure practical driving experience, which is theorized to increase their post-licensure crash risk.
Methods
Utilizing police-reported crashes and survey data from a randomized and quasi-randomized trial (n=458 adolescents, 16 or 17 years of age at enrollment), the impact of a parent-directed supervised practice driving intervention and a comprehensive on-road driver assessment (ODA) with feedback was evaluated on adolescent drivers’ MVC involvement.
Results
Compared to the control condition, a non-significant 20% relative reduction in risk was observed for the parent-directed intervention: aHR=0.80(95%CI: 0.44,1.43); the unadjusted absolute risk reduction was 1.1%(95% CI:−4.4%,7.1%). Exposure to the ODA resulted in an 53% relative reduction of risk: aHR=0.47(95% CI: 0.24,0.91); the unadjusted absolute risk reduction was 5.4%(95% CI: −0.3%,10.7%).
Conclusions
Comprehensive ODA might be protective for adolescents; however, additional research is needed.
Keywords: Motor vehicle crashes, Graduated Driver Licensing, Teen drivers, Adolescence, Expertise
Motor vehicle crashes (MVCs) are the leading cause of death and disability to adolescents in the United States.1 Adolescents’ MVC risk is due to multiple interactive factors2 (e.g., personality, maturation), but largely influenced by practical inexperience with the driving task.3,4 MVC involvement is highest during the transition from permit holder to licensed driver and then reduces as adolescent drivers gain experience.5,6 The most successful interventions are Graduated Driver Licensing policies (GDL). GDL policies phase adolescents into licensure by providing an opportunity to build practical experience while restricting adolescents from higher risk driving scenarios (e.g., driving with peers). GDL is a successful “one-size-fits-all” intervention.7 Individual-level interventions targeting inexperience directly are needed to complement universal policies.
Learner Period of GDL
GDL programs typically contain provisions requiring a minimum amount of supervised practice driving during a “learner’s permit period,” which is intended to serve as protected time for adolescents to gain practical skill under the supervision of a qualified adult, prior to independent licensure. It is common for parents to serve as their adolescents’ practice supervisor,8 but studies have shown that parents have difficulty providing an appropriately varied and challenging practice experience.9–12 Since MVC rates are highest when adolescents transition from supervised learners to independent license holders and because parents are responsible for most adolescents’ practice, 6,13,14 it follows that learner-period interventions are needed. Yet, there are few effective programs. The majority target parents’ knowledge of GDL, passively provide informational resources, and aim to increase the quantity of practice, without attention to a broader range of potential intervention targets (e.g., quality of supervised practice).13,15,16 Interventions that have demonstrated initial success on precursor behaviors (e.g., increased parent engagement, driving skills) were interactive (e.g., used active learning principles) and were more comprehensive in focus (i.e., targeted multiple psychological or behavioral factors).16–18 However, long-term effectiveness has been largely unevaluated.
Current Study
We evaluated the long-term effectiveness of a web-based intervention, the TeenDrivingPlan (TDP), administered during the learner period of GDL on adolescents’ involvement in police-reported MVCs for up to four years post-enrollment using a randomized-controlled trial design. A detailed description of the TDP can be found elsewhere.17,19,20 Briefly, the TDP was designed to improve the quality and quantity of practice driving. The TDP had three main components: (1) a practice tutorial library that consisted of short animated videos on how to practice specific environment-based goals (e.g., lane management on highways) and how to create a positive learning environment; (2) an interactive practice planner that families could use to plan drives ahead of time (e.g., pick a date and time, practice activity); and (3) a logging and rating tool. The logging and rating tool was the most used component of the TDP followed by the tutorial library; the planner was infrequently used.21 TDP use was assessed with user-specific log-in credentials. The TDP increased parent engagement, social support, and practice diversity and decreased the likelihood that adolescents would fail a comprehensive on-road driving assessment (ODA); no effect was observed on the quantity of practice.17,19
Because the ODA was not administered to every participant, the trial design presents a unique opportunity to determine if the ODA is associated with crash involvement. This is important because driver licensing evaluations in the US last on average only about 20 minutes and are generally undemanding of drivers.22 Therefore, there are direct GDL policy implications for identifying if a comprehensive ODA with feedback could be protective against crashes. As such, the primary objectives of the current analysis were to determine if the TDP and ODA influenced adolescent drivers’ involvement in police-reported MVCs. Secondary objectives were to evaluate if there was a difference in the amount of diversity and quantity of supervised driving practice between those participants who took the ODA and those who did not, and to determine if individual differences in these two practice variables were associated with adolescents’ MVC involvement.
The TDP was designed to improve parent-supervised practice explicitly. However, it is also possible that the ODA affected parents’ supervision by causing parents to “teach to the test.” In other words, the knowledge that their adolescent was going to participate in the ODA could have prompted parents to increase the quantity and diversity of supervised practice in preparation.
Hypotheses
We hypothesized that: (1) the TDP and ODA would reduce MVC involvement, (2) greater practice diversity would be associated with reduced MVC involvement, and (3) the ODA would increase both practice diversity and practice quantity. We did not make specific hypotheses about practice quantity and MVCs. Prior research has indicated weak support for practice quantity as a protective factor associated with future crashes, with the caveat that there have been methodological problems with most of the studies on this topic.23,24 We theorized that increasing diversity of practice (i.e., exposure to a greater variety of driving environments and conditions) could enhance the degree of correspondence between the pre-license supervised practice and the real-world task demands associated with independent driving, resulting in adolescent drivers who are more prepared to drive independently in a wider variety of settings and circumstances.
Methods
Description of the Trial Design
A stratified (ODA + survey vs. survey only) randomized (3:2) controlled trial design was used to determine how assignment to the TDP compared with a usual practice condition on the proportion of adolescent participants involved in motor vehicle crashes and the time to these events. As the primary foci of the initial trial were to evaluate the effect of the TDP on adolescents’ pre-license driving skill and to evaluate the relationship between supervised practice and driving skill, performance on the ODA served as the key outcome measure. Due to scheduling logistics, enrollment into the ODA stratum was prioritized based on the projected number of ODA time slots available, which varied weekly. When potential participants contacted the study team they were enrolled into the ODA stratum if slots were available and, if no slots were available, were enrolled into the survey only stratum. Slot availability was determined by the rehabilitation hospital that conducted the ODA. ODA stratum enrollment began in December 2011 and concluded in July 2012, and survey only enrollment began in January 2012 and ended in August 2012. Study procedures for the original trial concluded in January 2013. Therefore, TDP assignment was random and ODA assignment was quasi-random.
Description of the ODA
The ODA was administered by certified driving rehabilitation specialists in a dual control vehicle at 12 (±3) and 24 (±3) weeks after enrollment, serving as safety and primary outcome assessments for the initial trial, respectively. Evaluators were blinded to TDP assignment status. The ODA was 30.6 km (19.0 miles) long and consisted of 7 sequential modules in 2 sets: Set 1: parking lot, intermediate roads, suburban commercial district, and residential neighborhood and Set 2: urban commercial district, highway, and rural roads. The 12-week assessment consisted of the Set 1 modules only. Interrater reliability for the two evaluators who conducted the ODA was strong: κ = 0.85 [95% CI, 0.84–0.87].25 The ODA can discriminate between novice adolescent drivers and experienced adults,25,26 and dimensions of supervised practice are associated with better ODA test performance at 24 weeks (fewer errors, less likely to fail).19,26 Administrative procedures and scoring criteria are described elsewhere.25,26 At the conclusion of the 24-week ODA, and after scores were recorded, evaluators provided performance-related feedback and answered questions from the parents and adolescents. The content of these discussions was neither dictated nor documented by the study team. Providing feedback to the participants was deemed important for ethical reasons.
Eligibility Criteria
To be eligible for the initial trial, adolescent participants were required to: be 16 or 17 years of age at the time of enrollment; hold a learner’s permit with no more than 5 hours of behind-the-wheel practice at the time of enrollment; be fluent in written and spoken English; have an internet connection in the home; have one available vehicle at their primary residence; and have a parent or guardian at least 21 years of age to serve as the primary practice supervisor. Adolescents who were pregnant, anticipated needing a handicapped placard or license in order to drive, or who previously received driver education were ineligible to participate. Participants were recruited from the general community in the Philadelphia metro-area, which includes urban, suburban, and rural areas. The study procedures were approved by the IRB’s of the Children’s Hospital of Philadelphia and the University of Alabama at Birmingham.
Measures
Police-Reported Crashes
Crash records from December 2011 through December 2015 were requested from the Pennsylvania Department of Transportation (DOT). Reportable crashes in Pennsylvania are those that involve a death or injury and/or damage to any vehicle such that it must be towed from the scene. TDP trial data were individually joined with crash data provided by the DOT via learner permit numbers, which are the same as driver license numbers. The quality of the join was assessed by checking sociodemographic variables on the crash report against those provided at enrollment.
Surveys
A sociodemographic survey administered at enrollment was used to assess participants’ gender, date of birth, race, ethnicity, education, date of permit, and hours of prior behind-the-wheel practice. Four additional surveys were administered electronically at 6-week intervals for 24 weeks that collected data on a variety of social and behavioral variables; methodological details available elsewhere.19,27 For the current analysis, we utilized: (1) self-reported practice in the prior week in hours/minutes assessed at weeks 6, 12, 18, and 24. Parents and adolescents reported the number of hours of supervised practice completed in the past week and (2) self-reported practice diversity assessed only at week 24. Parents and adolescents used a six-point modified frequency scale: 1, “none”; 2, “less than 1 hour”; 3, “1–2 hours”; 4, “3–5 hours”; 5, “6–10 hours”; and 6, “more than 10 hours” to report how much supervised driving practice the adolescent completed in each of the following environments during the learner period: (1) empty parking lots; (2) residential neighborhoods; (3) one or two lane intermediate roads; (4) rural roads; (5) commercial roads (e.g., around shops and businesses); and (6) highways. Practice diversity was defined as the number of environments in which parent-adolescent dyads reported practicing for at least 1–2 hours. Responses were averaged for parents and adolescents.
Main Analytic Approach
For the effect of TDP, an intent-to-treat analytic approach was used, meaning that the randomization (enrollment) drives treatment delineation rather than treatment compliance. Time from enrollment to first crash was used as the primary outcome. For those participants without reported crashes, their event time was censored at December 31, 2015. Whether the proportion of participants having a crash differs by intervention group was assessed with a chi-square test. Cumulative incidence plots were used to visualize the time-to-crash outcome by treatment group and ODA assignment. Given the treatment and ODA assignment structure, all statistical models included both treatment group and ODA group. In order to jointly model effects of treatment and ODA assignment, Cox proportional hazards models were fit, and effect modification of both ODA assignment and gender were assessed separately through statistical interactions. Sensitivity analyses using time from intervention cessation (24 weeks) to crash were also performed. The analytic data set was de-identified prior to analysis. The level of statistical significance was set to 0.05. All analyses were conducted in SAS 9.4 (SAS Institute, Inc., Cary, NC) and R 3.3.1. (R Core Team, 2015).
Participants
The analytic sample consisted of 458 adolescents; this was 89% of the original study sample (Figure 1). Attrition was 5% in the survey only group and 18% in the ODA group; however, attrition was not selective to sociodemographic characteristics (p>.05 for all comparisons). Although the data for all 458 participants were obtained from the Pennsylvania DOT, one participant had crash data contradicting the age and gender from demographics obtained at enrollment; this case was excluded. Demographic characteristics of participants in the intervention and control groups did not differ, but compared to the ODA group, adolescent participants in the non-ODA group were more likely to be black, were slightly older, and parents engaged in less frequent internet use at work (Table 1).
Figure 1.
Enrollment diagram for the original trial
Table 1.
Characteristics of the Study Sample
| Teen Characteristics | Overall Sample | Non-ODA | ODA | |||
|---|---|---|---|---|---|---|
|
| ||||||
| Non-ODA (n=279) | ODA (n=178) | TDP Group (n=165) | Control (n=114) | TDP Group (n=96) | Control (n=82) | |
| Any crash, No. (%) | 34 (12.2) | 12 (6.7) | 18 (10.9) | 16 (14.0) | 7 (7.3) | 5 (6.1) |
| 0 Crash | 245 (87.8) | 166 (93.3) | 147 (89.1) | 98 (86.0) | 89 (92.7) | 77 (93.9) |
| 1 Crash | 32 (11.5) | 11 (6.2) | 17 (10.3) | 15 (13.2) | 6 (6.3) | 5 (6.1) |
| 2 Crashes | 2 (0.7) | 1 (0.6) | 1 (0.6) | 1 (0.9) | 1 (1.0) | 0 (0) |
| Age at enrollment, y, Median (IQR)* | 16.3 (16.1, 16.7)) | 16.2 (16.0, 16.5) | 16.3 (16.1, 16.6) | 16.3 (16.1, 16.9) | 16.1 (16.0, 16.4) | 16.2 (16.1, 16.6) |
| Behind-the-wheel practice hours at enrollment, Median (IQR) | 2.0 (0.3, 3.0) | 1.5 (0.5, 3.0) | 1.0 (0, 3.0) | 2.0 (0.5, 3.5) | 1.8 (0.5, 3.0) | 1.5 (0.5, 3.0) |
| Gender, No. (%) | ||||||
| Male | 134 (48.0) | 77 (43.3) | 87 (52.7) | 47 (41.2) | 41 (42.7) | 36 (43.9) |
| Female | 145 (52.0) | 101 (56.7) | 78 (47.3) | 67 (58.8) | 55 (57.3) | 46 (56.1) |
| Race* | ||||||
| Black or African American | 41 (14.7) | 9 (5.1) | 22 (13.3) | 19 (16.7) | 5 (5.2) | 4 (4.9) |
| White | 221 (79.2) | 159 (89.3) | 133 (80.6) | 88 (77.2) | 88 (91.7) | 71 (86.6) |
| Other/multiple race | 17 (6.1) | 9 (5.1) | 10 (6.1) | 7 (6.1) | 3 (3.1) | 6 (7.3) |
| Refused | 0 (0) | 1 (0.6) | 0 (0) | 0 (0) | 0 (0) | 1 (1.2) |
| Hispanic | ||||||
| Non-Hispanic | 270 (96.8) | 170 (95.5) | 158 (95.8) | 112 (98.2) | 93 (96.9) | 77 (93.9) |
| Hispanic | 7 (2.5) | 6 (3.4) | 5 (3.0) | 2 (1.8) | 2 (2.1) | 4 (4.9) |
| Unsure/refused | 2 (0.7) | 2 (1.1) | 2 (1.2) | 0 (0) | 1 (1.0) | 1 (1.2) |
| Parent characteristics | ||||||
| Relation to Teen | ||||||
| Mother | 218 (78.1) | 136 (76.4) | 130 (78.8) | 88 (77.2) | 72 (75.0) | 64 (78.0) |
| Father | 59 (21.2) | 42 (23.6) | 35 (21.2) | 24 (21.1) | 24 (25.0) | 18 (22.0) |
| Grandparent | 2 (0.7) | 0 (0) | 0 (0) | 2 (1.8) | 0 (0) | 0 (0) |
| Previous experience teaching a teen to drive | ||||||
| Yes | 124 (44.4) | 74 (41.6) | 78 (47.3) | 46 (40.4) | 41 (42.7) | 33 (40.2) |
| No | 155 (55.6) | 98 (55.1) | 87 (52.7) | 68 (59.6) | 52 (54.2) | 46 (56.1) |
| Missing | 0 (0) | 6 (3.4) | 0 (0) | 0 (0) | 3 (3.1) | 3 (3.7) |
| Parent characteristics con’t | Overall Sample | Non-ODA | ODA | |||
|---|---|---|---|---|---|---|
|
| ||||||
| Non-ODA (n=279) | ODA (n=178) | TDP Group (n=165) | Control (n=114) | TDP Group (n=96) | Control (n=82) | |
| Marital Status | ||||||
| Single | 44 (15.8) | 15 (8.4) | 27 (16.4) | 17 (14.9) | 8 (8.3) | 7 (8.5) |
| Living with partner | 3 (1.1) | 1 (0.6) | 2 (1.2) | 1 (0.9) | 0 (0) | 1 (1.2) |
| Married | 231 (82.8) | 159 (89.3) | 136 (82.4) | 95 (83.3) | 88 (91.7) | 71 (86.6) |
| Refused | 1 (0.4) | 1 (0.6) | 0 (0) | 1 (0.9) | 0 (0) | 1 (1.2) |
| Missing | 0 (0) | 2 (1.1) | 0 (0) | 0 (0) | 0 (0) | 2 (2.4) |
| Education | ||||||
| Some college or less | 73 (26.2) | 46 (25.8) | 46 (27.9) | 27 (23.7) | 28 (29.2) | 18 (22.0) |
| College graduate | 116 (41.6) | 75 (42.1) | 69 (41.8) | 47 (41.2) | 42 (43.8) | 33 (40.2) |
| Graduate degree | 90 (32.3) | 56 (31.5) | 50 (30.3) | 40 (35.1) | 26 (27.1) | 30 (36.6) |
| Refused | 0 (0) | 1 (0.6) | 0 (0) | 0 (0) | 0 (0) | 1 (1.2) |
| Internet/email home use | ||||||
| Several times a day | 207 (74.2) | 136 (76.4) | 123 (74.5) | 84 (73.7) | 75 (78.1) | 61 (74.4) |
| About once a day | 44 (15.8) | 31 (17.4) | 26 (15.8) | 18 (15.8) | 14 (14.6) | 17 (20.7) |
| 3–5 days a week | 15 (5.4) | 8 (4.5) | 10 (6.1) | 5 (4.4) | 4 (4.2) | 4 (4.9) |
| 1–2 days a week or fewer | 12 (4.3) | 3 (1.7) | 6 (3.6) | 6 (5.3) | 3 (3.1) | 0 (0) |
| Missing | 1 (0.4) | 0 (0) | 0 (0) | 1 (0.9) | 0 (0) | 0 (0) |
| Internet/email work use* | ||||||
| Several times a day | 163 (58.4) | 127 (71.3) | 96 (58.2) | 67 (58.8) | 70 (72.9) | 57 (69.5) |
| About once a day | 23 (8.2) | 7 (3.9) | 14 (8.5) | 9 (7.9) | 3 (3.1) | 4 (4.9) |
| 3–5 days a week | 5 (1.8) | 4 (2.2) | 4 (2.4) | 1 (0.9) | 1 (1.0) | 3 (3.7) |
| 1–2 days a week or fewer | 86 (30.8) | 36 (20.2) | 49 (29.7) | 37 (32.5) | 20 (20.8) | 16 (19.5) |
| Missing | 2 (0.7) | 4 (2.2) | 2 (1.2) | 0 (0) | 2 (2.1) | 2 (2.4) |
Note: To compare the ODA and Non-ODA groups, we used Wilcoxon ranked sum tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables.
Indicates p<.01. No other ODA vs. Non-ODA comparisons were significant.
Results
Of the 457 participants, 46 (10.1%) were involved as a driver in a reportable crash. In the non-ODA group 12.2% of participants crashed compared with 6.7% of participants in the ODA group (p=0.059), making the absolute unadjusted risk reduction 5.4% (95% CI:−0.3%,10.7%). Therefore, the estimated number needed to treat (i.e., the inverse of the absolute risk reduction) is: 1 out of 19 participants would be prevented from experiencing a crash if they were administered the ODA. In comparison, 10.7% of control participants crashed compared with 9.6% of participants in the TDP group, making the absolute risk reduction 1.1%(95% CI: −4.4%,7.1%). Therefore, we estimate that 1 out of 89 adolescents would be prevented from crashing if they received the TDP.
Accounting for ODA group, the hazard ratio for TDP was 0.86(95% CI: 0.48,1.54). Based on testing statistical interactions, there was no evidence that ODA (p=0.50) or gender (p=0.20) modified the effect of the TDP. Accounting for the TDP group, the hazard ratio for ODA was 0.53(95% CI:0.27,1.02). Adjusting for potential confounders (race, age at enrollment, and parent work internet use) did not change the point estimate for the effect of the ODA: 0.53(95% CI: 0.26,1.02). Given the emerging evidence that ODA may reduce crash risk, we performed an exploratory analysis examining whether gender modifies the effect of ODA assignment; it did not (p=0.49). Adjusting for TDP group and stratifying by gender, the following unadjusted hazard ratios for males and females were obtained: HRmales=0.39(95% CI: 0.15,1.06) and HRfemales=0.66(95% CI: 0.27,1.62). Comparisons of the overall proportion of adolescents who crashed versus those who did not showed no significant differences related to the parent and teen characteristics in Table 1 (see online supplement). Figure 2 depicts the cumulative proportion of crash-involved participants by group.
Figure 2.
Cumulative Incidence Plot
A Cox proportional hazards model was used to examine the independent effects of the TDP and ODA jointly with quantity and diversity of practice driving on adolescents’ crash risk. Wilcoxon rank-sum tests were used to assess for group-level differences in the practice variables by TDP and ODA assignment. Effect modification was assessed by creating multiplicative interaction terms (e.g., ODA*Diversity). For the Cox model, multiple imputation using chained equations was applied to account for those participants missing either practice diversity or practice quantity.29–31
As shown in Table 2, the ODA was associated with increased practice diversity (p=0.021) and the average amount of practice quantity across the 4 assessment points (p=0.004), while the TDP increased diversity (p=0.018) and slightly decreased practice quantity (p=0.019). Results of the multiple Cox proportional hazards regression model of crash risk accounting for practice quantity and practice diversity appear in Table 3. Adjusting for all other predictors in the model, the overall estimate of the effect of ODA was: HR=0.47(95% CI: 0.24,0.91). Practice quantity was not associated with crashes: HRquantity=1.00(95% CI: 0.71,1.41), but greater practice diversity was: HRdiversity=1.45(95% CI:1.02,2.06). There were no statistical interactions between ODA and practice diversity nor ODA and practice quantity (all comparisons p>.10).
Table 2.
Supervised Practice Variables (N=457)
| Non-ODA (n=279) | ODA (n=178) | p-value | TDP Group (n=261) | Control (n=196) | p-value* | |
|---|---|---|---|---|---|---|
| Practice Diversity, Median (IQR) | 5.0 (4.0, 6.0) | 5.5 (4.5, 6.0) | 0.021 | 5.0 (4.5, 6.0) | 5.0 (4.0, 6.0) | 0.018 |
| Observed n, | n=254 | n=171 | n=237 | n=188 | ||
| Practice Quantity (Hours/Week), Median (IQR) | 1.9 (1.3, 2.7) | 2.3 (1.6, 3.2) | 0.004 | 2.0 (1.4, 2.7) | 2.2 (1.5, 3.0) | 0.019 |
| Observed n, | n=244 | n=167 | n=231 | n=180 |
Note. The Wilcoxon rank-sum test used for each comparison.
Table 3.
Adjusted Hazard Ratio (aHR) Estimates of Adolescents’ Crash Risk in a Multiple Cox Proportional Hazards Model (N=457)
| Variable | aHR (95% CI) | p-value |
|---|---|---|
| TDP | 0.796 (0.442, 1.433) | 0.447 |
| ODA | 0.470 (0.241, 0.914) | 0.026 |
| Practice Quantity | 1.022 (0.866, 1.208) | 0.793 |
| Practice Diversity | 1.449 (1.022, 2.055) | 0.038 |
Note: Parameter estimates are adjusted for all other terms in the model: Intervention assignment, ODA assignment, Practice Quantity, and Practice Diversity.
To examine whether ODA assignment influenced practice quantity over time, a hierarchical approach to analyzing practice quantity across the 4 waves was used. Log-transformation of minutes of practice per week was necessary for statistical model validity, so the individual parameters from the linear mixed effects model are not straightforward to interpret on the original minutes per week scale. In light of that nuance, we have expressed the effect size as a relative percent change (that is, [exp(β) −1]*100%) between each of our comparison groups (ODA vs. non-ODA and Wave 4 vs. Wave 1). We found that participants experienced approximately a 43% increase in practice quantity by wave 4 compared to wave 1, holding ODA assignment and treatment group constant (p<0.0001). In addition, the ODA was associated with a 17% increase in minutes practiced per week over those not assigned to ODA, holding all other variables constant (at any given time, and across treatment groups) p=0.004, but that difference did not change over time (p=0.44); (see the online supplement for the graphical depiction).
Discussion
We explored the potential for a comprehensive on-road assessment with feedback and a web-based parent supervised driving program to improve practice behaviors during the learner period of GDL and to reduce adolescents’ motor vehicle crashes during the initial years of independent licensure. In our sample, the ODA increased practice quantity and diversity and reduced crash risk by an estimated 53%. The TDP reduced practice quantity, although this effect was quite small, and increased practice diversity. We did not observe strong evidence that the TDP reduced crashes. This pattern of results could be due to the salience of the ODA in comparison to the TDP (i.e., an on-road assessment in live traffic vs. a software application), the ODA’s timing in relation to the outcome of interest (i.e., exposure to the TDP was earliest in the beginning portion of the learner’s permit period in contrast to the ODA, which was last administered at the very end of the permit period), and/or that the initial positive effects of the TDP were largest for the psychological variables and weaker for the behavioral variables. Finally, the ODA directly targeted teens whereas the TDP primarily was designed for parents.
Further research on the potential effectiveness of the ODA as well as discerning the mechanisms between ODA and crash reduction is needed. In contrast to parents and driver educators who emphasize verbal, instructional didactics, and maintaining safety margins,12,32 the ODA evaluators did not provide instruction; they created an opportunity to observe drivers’ performance in a variable and challenging traffic environment. This might have created a platform for spontaneous strategy discovery. Improved performance on higher-order problem solving tasks can result from explicit instruction about optimal strategies or through spontaneous strategy discovery, and occur abruptly in a phase shift, as opposed to gradually over time.33–37 Consistent with prior research on this topic (see Williams, 2007) our study did not provide strong evidence that practice quantity is protective.23 This counterintuitive, but consistent, finding could be further evidence that learning-to-drive safely may not be best characterized by only gradual learning processes, the dominant framework that is currently used to describe learning-to-drive safely.38,39 There is a scarcity of research on theoretical models of learning-to-drive safely and the results of this study highlight that much more research is needed on this topic.
We observed an interesting pattern of associations among the practice variables, crash risk, and the ODA. Although the ODA was protective against crashes and reduced crash risk, practice diversity was positively associated with future crashes, after adjusting for any ODA and TDP effects, which was contrary to our hypothesis. There may be a critical difference between self-selected and induced practice diversity. Self-selected practice diversity, in-part, might serve as an early proxy variable for future exposure to the driving environment. The positive association between exposure and crash risk is well documented.40 Adolescents with a stronger interest in - or practical need for - diverse driving may be more likely to practice in diverse environments when they have their permits and then go on to become high exposure drivers later, which increases their opportunity to be involved in a motor vehicle crash. Therefore, it may not be the diversity of practice per se, but rather other unmeasured psychosocial or practical factors that are accounting for some of the positive association between self-selected practice diversity and future crash risk.
It is also possible that experimentally inducing adolescents to experience a more diverse driving environment during the learner period (i.e., exposing them to a more diverse environment than what they would have chosen on their own) could have sensitized them to their strengths and weaknesses, which could have influenced their post-license behaviors (e.g., avoidance of challenging environments that exceed their developing skill), as well as directly increased their driving skills and strategies, accounting for some of the positive effects of the ODA. We did not, however, directly assess these variables in the current study.
Limitations
The study had important strengths such as the use of the statewide crash database, which served to minimize recall bias and selective reporting bias. However, the study did have several limitations: sample homogeneity (e.g., race, education), lack of statistical power, potential sample selection bias at enrollment, and lack of true randomization of ODA assignment. As ODA assignment was only quasi-random, we cannot rule out the presence of unmeasured confounders associated both with enrollment in the ODA stratum and with crashes; however, there is not strong empirical or theoretical guidance that suggests that exposure would be different in these groups or what factors, other than the ODA, could account for the magnitude of the effect observed. Finally, because countries vary in their policies and programs used to transition adolescents into licensure, we cannot be certain how the ODA might fit into non-US systems.
Conclusions
Most road test examinations that serve as gateways for the intermediate period of GDL are less comprehensive than the ODA and do not entail a detailed feedback process with parents and adolescents. This might be a missed opportunity.
Supplementary Material
Implications and Contribution.
Identifying interventions to reduce adolescent drivers’ risk for motor vehicle crashes is critical for adolescent health and for public safety. In this study, preliminary evidence suggests that comprehensive on-road driver assessment in variable conditions and feedback administered prior to licensure may reduce crash risk.
Acknowledgments
Funding: Research reported in this publication was supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R03HD082664. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Dennis R. Durbin, Megan C. Fisher Thiel, Maria T. Schultheis, Chris Gantz, and Flaura Winston are gratefully acknowledged for their contributions and support at earlier stages of this project. We also thank Bryn Mawr Rehabilitation Hospital, the Pediatric Research Consortium (PeRC) at the Children’s Hospital of Philadelphia, and the Pennsylvania Department of Transportation for their support of the study. Finally, we would like to express our immense gratitude to the adolescents and their parents who agreed to participate in this research.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.National Center for Injury Prevention and Control (NCIP) Web-based Injury Statistics Query and Reporting System(WISQARS) [Google Scholar]
- 2.Shope JT, Bingham CR. Teen Driving. Motor-Vehicle Crashes and Factors That Contribute. Am J Prev Med. 2008;35(3 SUPPL) doi: 10.1016/j.amepre.2008.06.022. [DOI] [PubMed] [Google Scholar]
- 3.Braitman KA, Kirley BB, McCartt AT, Chaudhary NK. Crashes of novice teenage drivers: characteristics and contributing factors. J Safety Res. 2008;39(1):47–54. doi: 10.1016/j.jsr.2007.12.002. [DOI] [PubMed] [Google Scholar]
- 4.Curry AE, Hafetz J, Kallan MJ, Winston FK, Durbin DR. Prevalence of teen driver errors leading to serious motor vehicle crashes. Accid Anal Prev. 2011;43(4):1285–1290. doi: 10.1016/j.aap.2010.10.019. [DOI] [PubMed] [Google Scholar]
- 5.Chapman EA, Masten SV, Browning KK. Crash and traffic violation rates before and after licensure for novice California drivers subject to different driver licensing requirements. J Safety Res. 2014;50:125–138. doi: 10.1016/j.jsr.2014.05.005. [DOI] [PubMed] [Google Scholar]
- 6.Curry AE, Pfeiffer MR, Durbin DR, Elliott MR. Young driver crash rates by licensing age, driving experience, and license phase. Accid Anal Prev. 2015;80:243–250. doi: 10.1016/j.aap.2015.04.019. [DOI] [PubMed] [Google Scholar]
- 7.Williams AF, Tefft BC, Grabowski JG. Graduated Driver Licensing Research, 2010-Present. J Safety Res. 2012;43(3):195–203. doi: 10.1016/j.jsr.2012.07.004. [DOI] [PubMed] [Google Scholar]
- 8.Ginsburg KR, Winston FK, Durbin DR. Parents Teaching Teens to Drive: The Adolescent Perspective. Philadelphia: 2011. [Google Scholar]
- 9.Mirman JH, Kay J. From Passengers to Drivers: Parent Perceptions About How Adolescents Learn to Drive. J Adolesc Res. 2012;27(3) doi: 10.1177/0743558411409934. [DOI] [Google Scholar]
- 10.Sherman K, Lapidus G, Gelven E, Banco L. New teen drivers and their parents: What they know and what they expect. Am J Health Behav. 2004;28(5):387–396. doi: 10.5993/AJHB.28.5.1. [DOI] [PubMed] [Google Scholar]
- 11.Goodwin AH, Foss RD, Margolis LH, Harrell S. Parent comments and instruction during the first four months of supervised driving: An opportunity missed? Accid Anal Prev. 2014;69:15–22. doi: 10.1016/j.aap.2014.02.015. [DOI] [PubMed] [Google Scholar]
- 12.Tronsmoen T. Differences between formal and informal practical driver training as experienced by the learners themselves. Transp Res Part F Traffic Psychol Behav. 2011;14(3):176–188. doi: 10.1016/j.trf.2010.11.009. [DOI] [Google Scholar]
- 13.Jacobsohn LJ, Garcia-Espana JF, Durbin DR, Erkoboni D, Winston FK. Adult-supervised practice driving for adolescent learners: The current state and directions for interventions. J Safety Res. 2012;43(1):21–28. doi: 10.1016/j.jsr.2011.10.008. [DOI] [PubMed] [Google Scholar]
- 14.Mayhew DR, Simpson HM, Pak A. Changes in collision rates among novice drivers during the first months of driving. Accid Anal Prev. 2003;35(5):683–691. doi: 10.1016/S0001-4575(02)00047-7. [DOI] [PubMed] [Google Scholar]
- 15.Goodwin AH, Waller MW, Foss RD, Margolis LH. Parental supervision of teenage drivers in a graduated licensing system. Traffic Inj Prev. 2006;7(3):224–231. doi: 10.1080/15389580600665194. [DOI] [PubMed] [Google Scholar]
- 16.Curry AE, Peek-Asa C, Hamann CJ, Mirman JH. Effectiveness of Parent-Focused Interventions to Increase Teen Driver Safety: A Critical Review. J Adolesc Heal. 2015;57:S6–14. doi: 10.1016/j.jadohealth.2015.01.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mirman JH, Curry AE, Winston FK, et al. Effect of the teen driving plan on the driving performance of teenagers before licensure: A randomized clinical trial. JAMA Pediatr. 2014;168(8) doi: 10.1001/jamapediatrics.2014.252. [DOI] [PubMed] [Google Scholar]
- 18.Taubman-Ben-Ari O, Lotan T. The contribution of a novel intervention to enhance safe driving among young drivers in Israel. Accid Anal Prev. 2011;43(1):352–359. doi: 10.1016/j.aap.2010.09.003. [DOI] [PubMed] [Google Scholar]
- 19.Mirman JH, Albert WD, Curry AE, Winston FK, Fisher Thiel MC, Durbin DR. TeenDrivingPlan effectiveness: The effect of quantity and diversity of supervised practice on teens’ driving performance. J Adolesc Heal. 2014;55(5) doi: 10.1016/j.jadohealth.2014.04.010. [DOI] [PubMed] [Google Scholar]
- 20.Mirman JH, Lee Y-C, Kay J, Durbin D, Winston FK. Development of Web-Based Parent Support Program to Improve Quantity, Quality, and Diversity of Teens’ Home-Based Practice Driving. 2012. [DOI] [Google Scholar]
- 21.Winston FK, Mirman JH, Curry AE, Pfeiffer MR, Elliott MR, Durbin DR. Engagement with the TeenDrivingPlan and diversity of teens’ supervised practice driving: lessons for internet-based learner driver interventions. Inj Prev. 2014:1–6. doi: 10.1136/injuryprev-2014-041212. [DOI] [PubMed] [Google Scholar]
- 22.Mayhew DR. The learner’s permit. J Safety Res. 2003;34(1):35–43. doi: 10.1016/S0022-4375(02)00078-6. [DOI] [PubMed] [Google Scholar]
- 23.Williams AF. Contribution of the components of graduated licensing to crash reductions. J Safety Res. 2007;38(2):177–184. doi: 10.1016/j.jsr.2007.02.005. [DOI] [PubMed] [Google Scholar]
- 24.Twisk DAM, Stacey C. Trends in young driver risk and countermeasures in European countries. J Safety Res. 2007;38(2):245–257. doi: 10.1016/j.jsr.2007.03.006. [DOI] [PubMed] [Google Scholar]
- 25.Mirman JH, Curry AE, Schultheis MT, et al. Development of On-Road Driving Assessment for Learner Teen Drivers. Transp Res Rec. 2014;2465:64–72. doi: 10.3141/2465-09. [DOI] [Google Scholar]
- 26.Durbin DR, Mirman JH, Curry AE, et al. Driving errors of learner teens: Frequency, nature and their association with practice. Accid Anal Prev. 2014;72C:433–439. doi: 10.1016/j.aap.2014.07.033. [DOI] [PubMed] [Google Scholar]
- 27.Mirman JH, Curry AE, Wang W, Fisher Thiel MC, Durbin DR. It takes two: A brief report examining mutual support between parents and teens learning to drive. Accid Anal Prev. 2014:69. doi: 10.1016/j.aap.2013.10.006. [DOI] [PubMed] [Google Scholar]
- 28.R: A language and environment for statistical computing. 2015 https://www.r-project.org/
- 29.van Buuren S. Multiple imputation of discrete and continuous data by fully conditional specification. Stat Methods Med Res. 2007;16(3):219–242. doi: 10.1177/0962280206074463. [DOI] [PubMed] [Google Scholar]
- 30.White IR, Royston P. Imputing missing covariate values for the Cox model. Stat Med. 2009;28(15):1982–1998. doi: 10.1002/sim.3618. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med. 2011;30(4):377–399. doi: 10.1002/sim.4067. [DOI] [PubMed] [Google Scholar]
- 32.Ehsani JP, Klauer SG, Zhu C, Gershon P, Dingus TA, Simons-Morton BG. Naturalistic assessment of the learner license period. Accid Anal Prev. 2017;106(June):275–284. doi: 10.1016/j.aap.2017.06.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Riley MA, Holden JG. Dynamics of cognition. WIRES Cogn Sci. 2012;3(6):593–606. doi: 10.1002/wcs.1200. [DOI] [PubMed] [Google Scholar]
- 34.Haider H, Frensch PA. Why aggregated learning follows the power law of practice when individual learning does not: comment on Rickard (1997, 1999), Delaney et al. (1998), and Palmeri (1999) J Exp Psychol Learn Mem Cogn. 2002;28(2):392–406. doi: 10.1037/0278-7393.28.2.392. [DOI] [PubMed] [Google Scholar]
- 35.Welsh MC, Huizinga M. Tower of Hanoi disk-transfer task: Influences of strategy knowledge and learning on performance. Learn Individ Differ. 2005;15(4):283–298. doi: 10.1016/j.lindif.2005.05.002. [DOI] [Google Scholar]
- 36.Stephen DG, Boncoddo RA, Magnuson JS, Dixon JA. The dynamics of insight: mathematical discovery as a phase transition. Mem Cognit. 2009;37(8):1132–1149. doi: 10.3758/MC.37.8.1132. [DOI] [PubMed] [Google Scholar]
- 37.Siegler RS, Shipley C. In: Variation, Selection, and Cognitive Change. Simon T, Halford G, editors. Hillsdale, NJ: Erlbaum; 1995. [Google Scholar]
- 38.Simons-Morton BG, Ehsani JP. Learning to drive safely: reasonable expectations and future directions for the learner period. Safety. 2016;2:19p. doi: 10.3390/safety2040020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Foss RD, Martell CA, Goodwin AH, O’Brien NP. Measuring Changes in Teenage Driver Crash Characteristics During the Early Months of Driving. Washington, DC: AAA Foundation for Traffic Safety; 2011. [DOI] [Google Scholar]
- 40.Chapman R. The concept of exposure. Accid Anal Prev. 1973;5(2):95–110. doi: 10.1016/0001-4575(73)90018-3. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.