Covariability in three dimensions of teenage driving risk behavior: impaired driving, risky and unsafe driving behavior, and secondary task engagement

From phone use to speeding and driving under influence: Identifying clusters of driving risk behaviors as an opportunity for targeted interventions

Identifying the profile of risky behaviors among drivers is central to propose effective interventions. Due to the multidimensional and overlapping aspects of risky driving behaviors, cluster analysis can provide additional insights in order to identify specific subgroups of risk. This study aimed to identify clusters of driving risk behavior (DRB) among car drivers, and to verify intra-cluster differences concerning clinical and sociodemographic variables. We approached a total of 12,231 drivers and we included 6392 car drivers. A cluster algorithm was used to identify groups of car drivers in relation to the DRB: driving without a seat belt (SB), exceeding the speed limit (SPD), using a cell phone while driving (CELL), and driving after drinking alcohol (DUI). The algorithm classified drivers within five different DRB profiles. In cluster 1 (20.1%), subjects with a history of CELL. In cluster 2 (41.4%), drivers presented no DRB. In cluster 3 (9.3%), all drivers presented SPD. In cluster 4 (12.5%), drivers presented all DRB. In cluster 5 (16.6%), all drivers presented DUI. Clusters with DUI-related offenses (4 and 5) comprised more men (81.9 and 78.8%, respectively) than the overall sample (63.4%), with more binge drinking (50.9 and 45.7%) and drug use in the previous year (13.5 and 8.6%). Cluster 1 had a high years of education (14.4 ± 3.4) and the highest personal income (Md = 3000 IQR [2000-5000]). Cluster 2 had older drivers (46.6 ± 15), and fewer bingers (10.9%). Cluster 4 had the youngest drivers (34.4 ± 11.4) of all groups. Besides reinforcing previous literature data, our study identified five unprecedented clusters with different profiles of drivers regarding DRB. We identified an original and heterogeneous group of drivers with only CELL misuse, as well as other significant differences among clusters. Hence, our findings show that targeted interventions must be developed for each subgroup in order to effectively produce safe behavior in traffic.

Association between a delay in driving licensure and driving while impaired and riding with an impaired driver among emerging adults

BACKGROUND

Teens who delay driving licensure may not be subject to graduated driver licensing restrictions that are known to reduce crash risk. We explored the association of delay in licensure with driving while impaired (DWI) and riding with an impaired driver (RWI) among emerging adults.

METHODS

Data from the NEXT Generation Health Study, starting with 10th grade (2009-2010), were analyzed. The outcome variables were Wave 7 (W7) self-reported DWI and RWI as dichotomous variables. The independent variable was delay in licensure. Covariates included sex, urbanicity, race/ethnicity, family structure, parent education, family affluence, teen's highest education, minimum legal drinking age laws, and onset age of alcohol use. Descriptive analysis and logistic regressions were conducted.

RESULTS

Of 2525 participants eligible for licensure, 887 reported a delay in licensure by 1-2 years (38.9%, weighted) and 1078 by > 2 years (30.3% weighted) across 7 waves. In W7, 23.5% (weighted and hereafter, 5.6% once, 17.8% ≥twice) of participants reported DWI and 32.42% (5.6% once, 25.4% ≥twice) reported RWI. Logistic regressions showed no overall significant association of delay in licensure with either W7 RWI or W7 DWI. However, in stratified analyses, among African American youth, delay in licensure was positively associated with DWI (OR = 2.41, p = 0.03) and RWI (OR = 2.72, p = 0.05). Among those with ≤ high school or lower education by W7, delayed licensure was positively associated with RWI (OR = 2.51, p < 0.01).

CONCLUSIONS

While in the overall sample, delayed licensure did not appear to be associated with DWI or RWI, our findings suggest that delayed licensure may be of concern to teen risk of DWI and RWI among African Americans and among those with lower educational attainment. Furthermore, as two-thirds of youth delayed licensure, more research is needed to determine whether this is more of a positive (i.e., protective) factor by reducing their exposure to crash risk or a negative (i.e., risk) factor due to their missing important driver safety stages of graduated driver licensing.

Factor Identification and Prediction for Teen Driver Crash Severity Using Machine Learning: A Case Study

Crashes among young and inexperienced drives are a major safety problem in the United States, especially in an area with large rural road networks, such as West Texas. Rural roads present many unique safety concerns that are not fully explored. This study presents a complete machine leaning pipeline to find the patterns of crashes involved with teen drivers no older than 20 on rural roads in West Texas, identify factors that affect injury levels, and build four machine learning predictive models on crash severity. The analysis indicates that the major causes of teen driver crashes in West Texas are teen drivers who failed to control speed or travel at an unsafe speed when they merged from rural roads to highways or approached intersections. They also failed to yield on the undivided roads with four or more lanes, leading to serious injuries. Road class, speed limit, and the first harmful event are the top three factors affecting crash severity. The predictive machine learning model, based on Label Encoder and XGBoost, seems the best option when considering both accuracy and computational cost. The results of this work should be useful to improve rural teen driver traffic safety in West Texas and other rural areas with similar issues.

Kinematic risky driving behavior among younger and older drivers: Differences over time by age group and sex

Objective: This research examined the incidence rates of elevated gravitational force events (kinematic risky driving, KRD) among 16- to 17-year-old drivers compared to those of 18- to 20-year-old, 21- to 25-year-old, and 35- to 55-year-old drivers over a 12-month period. Methods: Data were sampled from the Strategic Highway Research Program 2 (SHRP2) naturalistic driving study that recruited a U.S. national sample of study participants. General linear mixed models (GLIMMIX) for recurrent events were used to estimate KRD incident rates for age cohorts in 3-month periods. Results: KRD incidence rates for 16- to 17-year-old drivers were higher than the rates for older drivers at each 3-month period. Analyses of individual differences for the 12-month period indicated that incidence rates for the 16- to 17-year-old group were 1.84 times higher than the rates for 18- to 20-year-old drivers, 2.86 higher than those for 21- to 25-year-old drivers, and 4.92 times higher than those for 35- to 55-year-old drivers. The incident rate for 16- to 17-year-old males was 1.9 times higher than that for same-aged females in the first 3 months and 2.3 times higher over 12 months. Over the study period, KRD rates of 16- to 17-year-old participants declined 24.5% among females and 18.0% among males. Conclusions: KRD rates were higher among younger relative to older, more experienced drivers and did not decline over time, consistent with a protracted period of risky driving behavior. The persistently higher KRD rate among young drivers suggests the need to enhance crash prevention approaches, such as feedback about abrupt maneuvering, to young drivers and their parents.

Examination of associations between risky driving behaviors and hazardous drinking among a sample of college students

OBJECTIVE

Hazardous drinking is associated with other risky behaviors and negative health-related outcomes. This study examined covariation between hazardous drinking scores and the following risky driving behaviors: Falling asleep while driving, texting (receiving and sending) while driving, and driving after consuming alcohol.

METHODS

The participants in this study were a sample of undergraduate students (N = 1,298) who were enrolled in an introductory health course at a large Southeastern university in spring 2016 and completed an online health survey that assessed hazardous drinking, falling asleep while driving, texting while driving, and driving after consuming alcohol. We conducted a series of 2-step regression analyses to examine covariation between hazardous drinking scores (using the Alcohol Use Disorders Identification Test-Consumption [AUDIT-C]) and the risky driving behaviors of interest.

RESULTS

We found that the majority of participants did not drink and drive (91.8%), nor did they fall asleep while driving (80.4%); however, the majority did read a text while driving (81.7%) and sent a text while driving (75.3%). In the full multivariable model, hazardous drinking score was positively associated with drinking and driving (step 2 odds ratio [OR] = 1.28, 95% confidence interval [CI], 1.18, 1.38) and sending a text message while driving (step 2 b = 0.19, P = .01). Hazardous drinking score was also positively associated with reading a text while driving (step 1 b = 0.14, P < .001) and falling asleep while driving (step 1 b = 0.02, P = .02) in the model only controlling for demographic characteristics. The analyses also indicated covariability between the following risky driving behaviors: (1) drinking and driving/reading a text message while driving, (2) drinking and driving/falling asleep and driving, (3) reading a text while driving/sending a text while driving, and (4) sending a text while driving/sleeping and driving.

CONCLUSION

The findings of this study indicate a need for interventions discouraging both hazardous drinking and risky driving behaviors in college student populations; recommendations to reduce risky driving and hazardous drinking are discussed.

U.S. Adolescent Street Racing and Other Risky Driving Behaviors

PURPOSE

We examined demographic characteristics and risky driving behaviors associated with street racing among adolescents in the NEXT Generation Health Study (N = 2,395).

METHOD

Binomial logistic regression tested associations between demographics and driving in a street race (DSR) or being a passenger in a street race (PSR). Sequential logistic regression tested the robustness of the association between DSR and crashes.

RESULTS

Hispanic/Latino, non-Hispanic Black/African-American, and mixed-race participants were more likely to engage in DSR. Males were more likely and teens with moderate socioeconomic status were less likely to engage in DSR and PSR. DSR was associated with other risky driving behaviors in bivariate models but was not independently associated with crashes after sequential modeling.

CONCLUSIONS

Among adolescents, those who are male, racial/ethnic minorities, or low socioeconomic status may be at higher risk of DSR. However, overall driving risk might explain the association between DSR engagement and higher crash risk.

Emerging Adults Riding With Marijuana-, Alcohol-, or Illicit Drug-Impaired Peer and Older Drivers

OBJECTIVE

The purpose of this study was to examine the prevalence and covariates among emerging adults of riding with an impaired peer or older adult driver (RWI) because of marijuana (MJ), alcohol (ALC), or illicit drugs (ID).

METHOD

Data were from Waves 4 (W4, N = 2,085) and 5 (W5, N = 2,116) of the NEXT Generation Health Study, collected in 2013-2014, 1 and 2 years after high school. W5 RWI was specified for substance-specific impaired peer and older adult (peer/older adult) drivers. Multinomial logistic regressions estimated W5 association of substance-specific RWI with W4 RWI, and W5 heavy episodic drinking, MJ use, and school/residence/work status.

RESULTS

At W5, 33% of the participants reported RWI in the past year, including riding with ALC- (21%), MJ- (17%), and ID- (5%) impaired peer drivers and ALC- (2%), MJ- (4%), and ID- (0.7%) impaired older adult drivers. W4 RWI was associated with W5 RWI with impaired peer/older adult drivers for ALC- (adjusted odds ratio [AOR] = 4.28, 2.69) and MJ-RWI (AOR = 2.34, 3.56). W5 heavy episodic drinking was positively associated with W5 peer ALC-related RWI (AOR = 2.16) and peer/older adult MJ-related RWI (AOR = 2.38, 5.45). W5 MJ use was positively associated with W5 peer ALC-related RWI (AOR = 2.23), peer/older adult MJ-related (AOR = 10.89, 2.98), and peer/older adult ID-related (AOR = 9.34, 4.26) RWI. ID-related RWI was higher among those not attending 4-year college (AOR = 3.38), attending technology school (AOR = 16.23), living on their own (AOR = 6.85), or living on campus (AOR = 11.50).

CONCLUSIONS

The high prevalence of RWI among emerging adults occurred mostly with ALC- or MJ-impaired peer drivers. The findings support the need for precisely tailored programs to prevent impaired driving according to substance use and age.

Driving in search of analyses

Although transportation safety has greatly improved over the past 2 decades, motor vehicle crash injuries remain a leading cause of morbidity and mortality, particularly among young drivers. Driver errors and behaviors such as speeding and distraction contribute disproportionately to crashes among inexperienced novices, who develop safe driving judgment only with substantial driving experience, commonly described as the "young driver problem." Research on young drivers has applied a range of research methods, including analyses of national archival data (mainly from police reports), crash analyses, observation of driver behavior, surveys of driver behavior and dispositions, and experimental research on driver behavior and vehicle crash worthiness. Prominent research questions regarding young driver safety include what and how do novices learn to drive safely, what are the predictors of young driver crashes, what is the variability and overtime trajectories of young driver performance and outcomes, and to what extent is the young driver problem due mainly to average population risk or high-risk groups? Current research on young drivers is complicated by small sample sizes, relatively rare events, high within and between group variability, missing data, the need to estimate exposure, and the lack of longitudinal and experimental designs, problems that require complex analytic methods. In this paper, we provide an overview of driving research methods, examples of research addressing the young driver problem, and examples of statistical collaboration on young driver research, focusing particularly on estimating prediction of crash risk and estimating variability in young driver performance and outcomes.