Traffic Crashes, Violations, and Suspensions Among Young Drivers With ADHD

Subjects were identified from the 6 New Jersey primary care practices within the CHOP network, which serves a socioeconomically, ethnically, and racially diverse population. Full details on the study’s design are available elsewhere.⁶ Briefly, we queried CHOP’s EHR database to select individuals who (1) were born in 1987–1997; (2) were patients at a New Jersey CHOP primary care practice; and (3), to establish New Jersey residency, had a CHOP network visit as a New Jersey resident within 4 years of becoming age eligible to drive (at 16 years) and maintained a New Jersey address through their last CHOP visit. We identified a total of 19 588 individuals. We then excluded individuals with a diagnosed intellectual disability (n = 73); individuals with only 1 primary care visit (n = 676), to minimize ADHD misclassification; and individuals who had their last primary care visit before age 12 years (n = 317), to ensure that individuals were seen at an old enough age to confirm ADHD status.¹ The underlying cohort included 18 522 patients. We limited this study to adolescents who obtained an intermediate (initial) driver’s license during the study period and had at least 1 month of post-licensure follow-up (n = 14 936; see Fig 1).

We classified subjects as having ADHD if their EHR indicated a 314.x International Classification of Diseases, Ninth Revision, Clinical Modification code either at a CHOP network visit or on the list of known conditions. A total of 1769 subjects were identified; 94% were classified on the basis of visits. We conducted a formal internal validation study of this classification scheme (sensitivity = 0.96; specificity = 0.98).¹⁶ 

Details on the process and validation of data linkages are available in previous publications.^6,17 Briefly, we obtained records for individuals who received a New Jersey license through December 2014 from the New Jersey Motor Vehicle Commission; data included exact dates of licensure, license suspensions and restorations, and traffic violations issued. We also obtained data on all police-reported crashes in New Jersey from January 2004 to December 2014. We conducted a hierarchical deterministic linkage; 98% of New Jersey drivers involved in a crash linked to a licensing record. We then used similar methods to link this licensing-crash database with CHOP EHR data. We estimated the true-match rate (true matches divided by original matches) to be 99.95% and the false nonmatch rate (true matches not found) to be 1.5%.

All subjects acquired an intermediate license in New Jersey, where the minimum licensure age is 17 years. In New Jersey, intermediate drivers licensed at age <21 years can drive independently (without adult supervision) but are restricted from driving between 11:01 pm and 4:59 am, while using electronic equipment, or with >1 peer passenger for the first year. They are also prohibited from driving with any detectable amount of alcohol in their system.¹⁸ Notably, New Jersey drivers age <21 years cannot plea bargain a point-carrying moving violation.

Primary crash outcomes for adolescent drivers were defined a priori and included (1) all crashes, reportable to police if an injury or >$500 in property damage occurred¹⁹; (2) injury crashes, in which at least 1 person had a moderate or greater severity injury (noted on crash report); (3) at-fault crashes, defined in previous work as those with a crash-contributing driver action (eg, inattention or unsafe speed)²⁰; (4) night crashes, including crashes that occurred during New Jersey’s restricted 11:01 pm to 4:59 am period (late) and between 9:00 pm and 11:00 pm (early), a period also identified to be higher risk²¹; (5) passenger crashes, previously defined as crashes with only passengers aged 14 to 20 years (peer) or ≥2 passengers of any age (multiple)²²; (6) single-vehicle crashes, the majority of which have been shown to involve speeding or traveling too fast for conditions²³; and (7) alcohol-related crashes, in which the driver was issued a violation for alcohol use or noted on the crash report to have a blood alcohol content level of ≥0.01 for drivers age <21 years and ≥0.08 for drivers age ≥21 years. We calculated monthly rates per 10 000 driver-months; the numerator was the number of crashes among validly licensed drivers, and person-time was calculated by summing for all drivers the proportion of the month that the driver had a valid license. Average monthly rates for the first 12 and 48 months were estimated. Follow-up time concluded at 48 months after licensure, at death, or at end of the study period, whichever occurred first.

Using similar methods, we calculated the monthly rate of violations issued for (1) all traffic offenses; (2) point-carrying moving violations, which include (a) speeding, (b) careless driving, (c) alcohol and/or drug use, and (d) electronic equipment use; (3) seat belt nonuse; and, (4) for intermediate drivers age <21 years, violation of GDL restrictions (eg, nighttime or passenger restrictions). Finally, exact periods of license suspension were identified; suspension rates were calculated as the number of days in which a driver had a suspended license per year of follow-up since the date of licensure.

Demographic variables were ascertained from the EHR. Two co-occurring conditions that may affect crash risk were categorized via International Classification of Diseases, Ninth Revision, Clinical Modification codes: seizure disorder (345.x) and disruptive behavior disorder (DBD) (312.x or 313.81). DBD was further verified through EHR review by study authors (T.J.P. and B.E.Y.) and trained abstractors; confirmation included ≥3 visits for DBD, independent sources (eg, provider notes), or symptoms consistent with DBD. We used 2010 Census Gazetteer Files and 2007−2011 American Community Survey data to categorize residence zip code at last visit into quintiles of population density and median household income, respectively.^24,25 

We compared bivariate distributions of demographic and clinical characteristics among drivers with and without ADHD using χ² and Wilcoxon rank-sum tests. Crash and violation risk within the first 12 and 48 months of licensure were compared by using χ² statistics; estimates were restricted to drivers followed to the specified post-licensure month. We estimated adjusted rate ratios (adjRRs) and 95% confidence intervals (CIs) using generalized estimating equation models with a log link (ie, Poisson distribution). For models, we accounted for correlation within individual drivers using an independent correlation structure. Potential covariates were chosen a priori on the basis of known or suspected association with ADHD (or its diagnosis) and outcomes, including sex, race and/or ethnicity, insurance payer, DBD and seizure disorders, licensing age (17 years 0 months, 17 years 1 month−17 years 11 months, 18 years, and ≥19 years), primary care practice, and birth year. Models also included indicators for zip code–level household income and population density as well as linear and quadratic terms for month since licensure to control for temporal trends. Analyses of GDL violations were restricted to drivers licensed at age <21 years and only through the first year of licensure. For rarer outcomes (alcohol-related crashes, violations for alcohol and/or drug use [12 months after licensure], and licensure suspension), fully adjusted models failed to converge; thus, models were adjusted only for sex and licensing age. Analyses were conducted in SAS version 9.4 (SAS Institute, Inc, Cary, NC). This study was approved by the CHOP Institutional Review Board.

Twelve percent of the cohort had a diagnosis of ADHD (Table 1). The majority of subjects were long-term CHOP primary care patients and were last seen at a median age of 18.1 years. Drivers with ADHD were licensed a median of 3.6 months later than those without ADHD and were more likely to be male and non-Hispanic white.

A higher proportion of young drivers with ADHD crashed within the first month (2.8% vs 1.9%; P = .007; not shown), 12 months, and 4 years of licensure (Table 2). The adjusted crash rate in the first month after licensure was 62% higher than that among drivers without ADHD (294.0 vs 187.6 per 10 000 driver-months; adjRR: 1.62 [95% CI: 1.18−2.23]; Fig 2A). After adjusting for potential covariates and temporal trends, the 4-year crash rate of novice adolescent drivers with ADHD was 37% higher than that of drivers without ADHD (95% CI: 1.26−1.48; Fig 3, Supplemental Table 3). In addition, in analyses limited to drivers with ADHD, we found that crash rates of novice drivers who were licensed older versus younger did not differ; for example, there did not appear to be a difference in the 12-month rate for drivers licensed at age ≥18 years and those licensed at age 17 years (adjRR: 1.10 [95% CI: 0.83−1.44]).

Drivers with ADHD also experienced higher rates of crash subtypes (crashes involving passengers and at-fault, single-vehicle, injury, and alcohol-related crashes; Fig 3, Supplemental Table 3). For example, in the first 48 months after licensure, drivers with ADHD had a 62% higher rate of injury crashes (95% CI: 1.23−2.14) and a 109% higher rate of alcohol-related crashes (95% CI: 1.16−3.76). Notably, these 2 outcomes were rare events; within the 48-month study period, 4.3% of young drivers with ADHD were involved in an injury crash, and 1.2% were involved in an alcohol-related crash (Table 2).

Among drivers with ADHD, 35.6% were issued a traffic violation, and 26.8% were issued a moving violation within their first year of driving (compared with 25.3% and 18.6%, respectively, among drivers without ADHD; Table 2). Rates of moving violations were consistently higher for drivers with ADHD over the study period (at 48 months adjRR: 1.47 [95% CI 1.36−1.58]; Figs 2B and 4). Similarly, rates for specific violations, including careless driving and speeding, were higher among drivers with ADHD. Violation rates for other risky driving behaviors were also elevated, including for alcohol and/or drug violations, which were an estimated 3.57 and 1.61 times higher over the first year and over 4 years of licensure, respectively. GDL violations did not appear to be elevated among those with ADHD (adjRR: 1.06 [95% CI: 0.80−1.40]). Finally, 17.1% of drivers with ADHD and 10.1% of drivers without ADHD had their license suspended at least once in the 4-year period, with average suspension rates of 13.5 and 7.0 days per driver-year, respectively (adjRR: 1.32 [95% CI: 1.08−1.61]; Table 2, Fig 4, Supplemental Table 3).

With this study, we provide the first longitudinal assessment of crash, violation, and suspension risk among adolescent drivers with ADHD. Findings indicate that adolescent drivers with ADHD have a moderately increased crash risk, a finding consistent with several recent population-based studies of adolescent and adult drivers with ADHD^12,13 but lower than estimates in early small studies of adolescents more severely affected.¹¹ In addition, with our study, we uniquely identify the early licensure period as a time of particularly high risk and the potential contribution of preventable factors to increased risk. These findings both offer important implications for families and for professionals working with them and highlight the need to develop comprehensive preventive approaches to reduce these adolescents’ crash risk.

First, this study suggests that risky driving behaviors may underlie elevated crash risk of adolescent drivers with ADHD. Their risk of alcohol-related crashes (albeit low on an absolute scale) remained considerably higher over time, and the relative risk of several risky driving–related crash types was comparatively higher than that of all crash types. In addition, drivers with ADHD had a higher risk of violations for speeding, seat belt nonuse, alcohol and/or drug use, and electronic equipment use. This is consistent with literature revealing that individuals with ADHD are more likely to engage in risk-taking behaviors such as risky sexual activity and substance abuse.^26,27 Importantly, many risky driving behaviors may be amenable to change. Although medication may improve driving performance for drivers with ADHD^28,29 (and thus is a primary recommendation to families³⁰), there is not firm evidence that medication reduces adolescents’ engagement in certain risky behaviors.³¹ In addition, there are serious medication adherence issues in adolescence, and our previous study revealed that few adolescents with ADHD are medicated at the time of licensure and detected no differences in crash risk by medication status at licensure.^6,32 Thus, novel nonpharmacologic preventive approaches that capitalize on adolescents’ desire for independence and that incorporate skill-training approaches for ADHD (which reveal evidence of effectiveness³³) should be strongly considered to promote safe driving. Research suggests that risky behaviors among adolescents with ADHD may be mediated by suboptimal decision-making.^34,35 As such, approaches may include training teenagers in decision-making skills in the context of driving and parent training to promote independent, responsible driving behavior. In addition, shared decision-making involving the health care provider, parent, and adolescent (strongly affirmed by the Institute of Medicine) is recommended for the development of intervention plans that balance best provider practices with the goals and preferences of adolescents and their families.³⁶ Additional research is needed to shape shared decision-making practices for youth with ADHD as they learn to drive. Finally, vehicles are increasingly equipped with advanced safety features specifically designed to reduce young drivers’ risk (eg, preset speed limits). Future studies to investigate the effect of these features on driving behaviors, and further, how adolescents interact with these features, are needed.

Second, in this study, we reported that crash risk is particularly heightened in the initial month of licensure and is elevated for novice drivers regardless of age at licensure. Our finding of a 62% increased rate for adolescent drivers with ADHD in their first month of driving reveals that they may be at their most vulnerable just after licensure. This is of particular concern because in general, teenagers are at their highest risk during the first 6 to 12 months after licensure.^37,38 Although parents of adolescents with ADHD are frequently guided to delay licensure,³⁰ our results reveal that (contrary to the general population of adolescent drivers³⁸) those with ADHD who delay licensure do not experience a lower crash risk once licensed. This finding may have important implications. In the vast majority of US states, adolescents who delay licensure until age ≥18 years are licensed outside the protective benefits of a GDL system (the most proven intervention to reduce young drivers’ crash risk³⁹). Thus, providers may encourage families inquiring about readiness to drive to seek the assistance of a certified driving rehabilitation specialist (educator specializing in training drivers with medical conditions) during the learner’s permit phase and to adopt GDL-like household rules for teenagers who are not under GDL purview.

Strengths of this study include its longitudinal nature, use of objective measures of outcomes, ability to account for potential covariates (sex, licensing age, time since licensure, and co-occurring conditions), and increased generalizability to the general population of adolescents with ADHD by using a community-identified cohort. A primary limitation is that diagnoses relied on assessment by primary care providers; however, we previously validated ADHD diagnosis codes with high sensitivity and specificity.¹⁶ Nevertheless, our sample may more appropriately be characterized as individuals with a lifetime history of ADHD because diagnoses occurred before adolescence and because some individuals may not have had ADHD-related impairments at the time of licensure.⁴⁰ Additionally, driving exposure was not directly measured. Several previous studies found that those with ADHD reported driving more miles^8,41; however, these studies were limited in the authors’ assessment of exposure (eg, usual driving at time of survey) and reliance on self-reports, which might have been subject to positive illusory bias.¹⁵ In our analyses, we accounted for time since licensure, a proxy for exposure and itself critically important given that crash risk declines over the initial years of driving. In this study, we also used crash and violation events as proxies for engagement in risky driving behavior; future studies in which naturalistic driving methodologies are used would provide additional crucial insights.⁴² Adolescents with ADHD may be less likely than adolescents without ADHD to leave the state for postsecondary education^43,44; this may lead to an overestimation of true rate ratios. However, rates were elevated in the 12-month period after licensure, before most subjects completed high school. In addition, generalizability may be limited as a result of New Jersey’s licensing age (the oldest in the United States) and high urbanization. Finally, the 12% prevalence of ADHD in our study cohort was higher than estimates of 8% to 9% ever diagnosed for New Jersey children⁴⁵; this may be influenced by CHOP’s reputation and corresponding care-seeking behavior of parents.

Adolescent drivers with ADHD are at particularly high crash risk in their initial months of licensure, and engagement in preventable risky driving behaviors likely underlies this increased risk. Prospective studies to objectively measure risky driving behaviors among novice drivers with ADHD and examine the extent to which these behaviors mediate driving outcomes are vital to inform prevention strategies. The development of comprehensive preventive approaches to reduce crash risk is critically needed.

We thank the New Jersey Department of Transportation, the New Jersey Motor Vehicle Commission, the New Jersey Office of Information Technology, and Children’s Hospital of Philadelphia Department for Biomedical Informatics for their assistance in providing data. We thank Melissa R. Pfeiffer, MPH, for her work on the New Jersey Traffic Safety Outcomes data warehouse. Drs Curry and Metzger had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

ADHD

attention-deficit/hyperactivity disorder

adjRR

adjusted rate ratio

CHOP

Children’s Hospital of Philadelphia

CI

confidence interval

DBD

disruptive behavior disorder

EHR

electronic health record

GDL

graduated driver licensing