Associations of Mental Health with Driving While Impaired and Risky Driving in Emerging Adults

Abstract

Objective:

Examined cross-sectional associations of driving while impaired (DWI) and risky driving with mental and psychosomatic health among U.S. emerging adults.

Methods:

Data were from years 1–4 after high school (waves 4–7) of the NEXT Generation Health Study, a nationally representative study starting with 10th grade (2009–2010). Outcome variables were DWI (dichotomous variable: $\ge$1 day vs. 0 days in the last 30 days) and risky driving Checkpoints Self-Reported Risky Driving Scale (C-RDS). Independent variables included depressive symptoms and psychosomatic symptoms. Multivariate logistic and linear regressions were conducted with complex survey features considered.

Results:

Higher depressive and psychosomatic symptoms were associated with modestly higher likelihood of DWI (Adjusted odds ratio [AOR] ranged from 1.02 to 1.03 and from 1.04 to 1.05, respectively) and higher C-RDS scores (b ranged from 0.06 to 0.12 and from 0.08 to 0.23, respectively) in years 1–4 after high school.

Conclusions:

Depressive and psychosomatic symptoms were associated with greater DWI and risky driving in all 4 years after high school. Negative mental and psychosomatic health should be targeted components of DWI and risky driving prevention to lower fatal motor vehicle crashes among emerging adults.

INTRODUCTION

Motor vehicle crashes remain the predominate cause of adolescent deaths in the U.S. In 2017, 2,364 U.S. teens aged 16–19 years were killed (six every day) in motor vehicle crashes. Risky driving behaviors (e.g., speeding, tailgating, and rolling through stops) are well known contributors to crash risk and are prevalent among adolescent drivers (Simons-Morton et al. 2013). Adolescent drivers are more likely to underestimate dangerous driving conditions and to participate in risky driving behaviors compared to their older counterparts (Simons-Morton et al. 2005).

Driving while alcohol- or drug-impaired (DWI) represents a particularly problematic type of risky driving behavior in relation to crash risk. Among high school students, the estimated national prevalence of drinking and driving in the past 30 days ranges from 9% in 2011 to 13% in 2010–2011 (Li et al. 2013). Moreover, alcohol related crashes are five-fold higher among adolescents than older drivers.

We need to identify factors that lead to risky driving and adolescent DWI to develop and implement strategies to effectively improve driving safety of adolescents. Previous studies demonstrated associations of other risky and unhealthy behaviors (e.g., tobacco use, physical fights, non-use of birth control, bullying behavior) with mental health (e.g., depression and stress) and psychosomatic health (e.g., headache, sleeping problems) (Fekkes et al. 2004) among adolescents, providing clear links between the other risk behaviors and mental health. Although the relations may be similar, few studies have examined associations of mental and psychosomatic health with risky driving (e.g., excessive speed may be related to stress relief and/or reduced self-control).

What evidence there is suggests experiencing psychosomatic symptoms is associated with poorer driving performance (Tepper et al. 2020). Psychosomatic symptoms are physical symptoms (e.g., headache, abdominal pain, musculoskeletal pain) related to psychosocial stressors. Psychosomatic symptoms such as chronic pain are common and the prevalence has increased among U.S. adolescents, with more high school students in the 2010s reporting symptoms compared to their counterparts in the 1980s (Twenge 2015). In the National Longitudinal Study of Adolescent Health, headache was the most frequently reported psychosomatic symptom (29%), followed by other symptoms such as musculoskeletal pain (27%), fatigue (21%) and stomach aches (18%) (Rhee et al. 2005). Demonstrated relations with driving performance include findings that anxious individuals were more likely to make errors (e.g., failing to check their mirrors before changing lanes) and perform aggressive violations (e.g., disregarding the speed limit) while driving (Shahar 2009). However, more studies are needed to examine the associations of psychosomatic symptoms with risky driving and DWI.

More is known about the association of driving-related behaviors and depression (Cunningham and Regan 2017). Depression is a substantial public health problem among U.S. adolescents with the prevalence of major depressive episodes having increased from 8.7% to 11.3% in less than 10 years (2005 to 2014) (Mojtabai et al. 2016). Major depressive disorder can negatively affect driving performance and safety as a result of, “slower reaction times to critical road events among participant between the ages of 18 and 65 (Bulmash et al. 2006).” A cross-sectional observational study found risky driving was associated with depressive symptoms in late adolescents and adults (McDonald et al. 2014). Furthermore, major depression disorder is a well-known significant predictor of alcohol-impaired driving among male adults (aged 21–64 years) (Pogue et al. 2017). However, more research is needed to characterize these relationships longitudinally and among larger and representative adolescent samples.

There are dynamic changes in trajectories of depressive symptoms and psychosomatic symptoms at different ages in adolescence (Ellis et al. 2017; Nummi et al. 2017) which may in turn contribute to risky driving behaviors, including DWI. Therefore, there is a need to examine the associations of risky driving and DWI with mental and psychosomatic health during the transition from adolescence to emerging adulthood.

We aimed to explore cross-sectional associations of risky driving and DWI with depressive and psychosomatic symptoms among U.S. emerging adults throughout the four years after high school. Findings will inform health specialists and practitioners in pediatric mental health care for the needs of early diagnosis and treatment of mental and psychosomatic health problems to avoid or alleviate immediate and lifelong behavioral consequences among emerging adults (Hawkins-Walsh and Van Cleve 2019). A broad understanding of the experience of late adolescents and emerging adults will be useful to practitioners engaged in transitioning youth with mental health conditions into adult care.

METHODS

Sampling

We used the data from four waves (W4-W7) of the NEXT Generation Health Study, a nationally representative cohort study, which started in the 2009–2010 school year. The sampling strategy for NEXT has been previously reported (Li et al. 2013). African American participants were oversampled to provide a large enough sample (N=687) to examine racial/ethnic differences. Surveys were administered in the spring semester of each wave. A total of 2785 unique participants participated in the NEXT study. Out of the total 2785 participants, 78% (N=2177), 79% (N=2202), 84% (N=2306), and 83% (N=2323) completed the survey in W4 through W7, respectively. Parental consent or participant assent were obtained. After turning 18, participant consent was individually obtained. The study protocol was approved by the Institutional Review Board of the National Institute of Child Health and Human Development.

Measures

Dependent Variables

Driving while alcohol- or drug-impaired:

DWI was assessed with three items asking participants how many days they drove in the past 30 days after: (1) drinking alcohol, (2) using marijuana, or (3) using illicit drugs. Due to non-normal distribution and a severe floor effect of the data, we collapsed and recoded responses across substances to derive a dichotomous variable, DWI≥1 day vs. 0 days in the past 30 days.

Checkpoints self-reported risky driving scale:

Risky driving was measured with the 21-item Checkpoints Risky Driving Scale (C-RDS) derived from previous studies (Simons-Morton et al. 2013) measuring risky driving behaviors (e.g., On how many days in the last 30 days have you: “…exceeded the speed limit in residential or school zones?”; “…purposely tailgated or followed another vehicle very closely?”, etc.). The DWI related measures were excluded from the C-RDS scale in this study to avoid redundancy and collinearity. In the current assessment the internal consistency of the C-RDS was acceptable (Cronbach alpha=0.90). We dichotomized responses on each of the 21 questions (1=≥1 day vs. 0=no days) and summed the 21 dichotomous variables, with possible counts ranging from 0 to 21 and used the C-RDS as a continuous variable.

Independent Variables

Depressive symptoms:

We measured depressive symptoms using the Pediatric Patient Reported Outcome Measurement Information System (PROMIS) depressive symptoms scale (Pilkonis et al. 2011). Participants were asked how often they felt each of 8 items was true including: (1) “I felt like I couldn’t do anything right”, (2) “I felt everything in my life went wrong”, (3) “I felt unhappy”, (4) “I felt lonely”, (5) “I felt sad”, (6) “I felt alone”, (7) “I thought that my life was bad”, and (8) “ I could not stop feeling sad” over the last 7 days, using a 5-point Likert scale, with responses ranging from (1) “Never” to (5) “Almost always”. The Cronbach’s alpha of the 8 items is 0.96 for all four waves (W4–7) indicating great internal consistency. Scores on the PROMIS were converted into T-scores based on distributions of scores in the general U.S. pediatric population (mean=50, SD=10). The T-scores were used in the linear and logistic regression analyses. Those scoring 60 or above, or one standard deviation higher than the reference population mean, were categorized as having higher depressive symptoms versus the rest of the sample. The dichotomous variable was used for the sample description of depressive symptoms.

Psychosomatic symptoms:

To measure psychosomatic symptoms participants were asked how often they have had any of the following symptoms: headache, stomach-ache, back ache, feeling low, irritability or bad temper, feeling nervous, difficulties in getting to sleep, and feeling dizzy in the last half year with 5 responses ranging from (1) “Rarely or never” to (5) “About every day” (Currie et al. 2008). Items were summed (range 8–40) to indicate psychosomatic symptoms as a continuous variable. The Cronbach’s alpha of the 8 items was 0.84 for all four waves (W4–7) indicating acceptable internal consistency. Those scoring in the 75^th percentile or above were categorized as having higher psychosomatic symptoms versus the rest of the sample. No clinical cut off is available for this measure. However, given that this is a nationally representative sample, it may be a reasonable that this cut off indicates a potentially problematic level of symptoms. The continuous variable was used in the linear and logistic regression analyses and the dichotomous variable was used for the sample description of psychosomatic symptoms.

Driving licensure:

Driving licensure was identified based on participants reporting if they had a license allowing independent and unsupervised driving. The analysis sample was limited to the subsample of participants who reported having an independent driver’s license.

Demographic and control variables:

Participants reported age at each wave, and sex, race/ethnicity, urbanicity, and family socioeconomic status at W1. One parent provided the highest education of both parents during the completion of the informed consent forms at W1. We estimated family socioeconomic status using the Family Affluence Scale (Harris et al. 2009), with measures including number of cars owned, number of computers owned, whether the student had his or her own bedroom, and the number of family vacations in the last 12 months. We categorized students as low, moderate, or high affluence. The highest education level of both parents was categorized as: 1) Less than a high school diploma, high school diploma or GED; 2) Some college, technical school or associate’s degree, or 3) Bachelor’s or graduate degree. We coded students’ living areas as urban, suburban, or rural.

Statistical Analysis

Although the NEXT project was a longitudinal study, we conducted a series of cross-sectional analyses rather than prospective analyses because many participants were still in the process of getting their driver’s license and sample sizes were not even across waves. All statistical analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC) and accounted for the complex survey design (i.e., stratification, clustering, and sampling weights). The standard errors were calculated based on the multistage stratified design of the survey. Linear regression was used for C-RDS and binary logistic regression was used for DWI. Unadjusted and adjusted models were conducted, the latter were conducted adjusting for sex, race/ethnicity, family affluence, highest parental education level, and urbanicity. All variables have four waves of data. Missing data were deleted listwise. The analysis was limited to the subsample of subjects with a driver’s license, therefore domain analysis was applied which provides accurate computation of statistics for subpopulations in addition to the computation of statistics for the entire study population. Interactions were conducted between demographic variables and depressive and psychosomatic symptoms for both outcome variables.

RESULTS

The weighted mean ages of participants at W4, W5, W6 and W7 were 19.15 years (SE=0.03), 20.25 years (SE=0.02), 21.24 years (SE=0.02), and 22.61 years (SE=0.03), respectively. Table 1 includes demographics, outcome variables and independent variables for all 4 waves (W4-W7). Along with the increasing proportion of participants with a driver’s license over four waves (76% [weighted and hereafter] at W4 to 87% at W7), DWI prevalence showed an increasing trend across the four waves, with prevalence increasing from W4 to W7 (14.63% at 4 to 23.40% at W7) based on trend analysis (W4 to W7, F value=63.32, p<.001) and paired t-test between W4 and W7 (t-value=9.85, p<.001). More specifically, prevalence of alcohol-related DWI, marijuana-related DWI, and other illicit drug-related DWI was 9.42%, 10.40%, and 3.24% at W4, 11.22%, 10.16%, and 3.02% at W5, 16.32%, 10.29%, and 2.84% at W6 and 18.99%, 11.32%, and 3.59% at W7. The levels of risky driving (C-RDS) and independent variables did not show a clear trend from W4-W7. The proportions of participants whose depressive symptom T-scores for the PROMIS were greater than 60 were 24% at W4, 24% at W5, 25% at W6, and 23% at W7. The proportions of participants whose psychosomatic symptoms scores were greater than the value at 75th percentile were 30% at W4, 33% at W5, 30% at W6, and 29% at W7.

Table 1.

Demographics, outcome variables and independent variables for waves 4 to 7

	Wave 4	Wave 5	Wave 6	Wave 7
	N (%)a	N (%)a	N (%)a	N (%)a
DWIb
No	1067 (85.4)	1200 (84.2)	1258 (79.0)	1311 (76.5)
Yes	185 (14.6)	227 (15.8)	336 (21.0)	438 (23.5)
Dep. symptomsb
T-score <60	1680 (76.3)	1678 (74.7)	1777 (77.1)	1791(78.4)
T-score ≥60	472 (23.7)	468 (25.3)	497 (22.9)	503 (21.6)
Psy. symptomsb
Score <75th percentile	1593 (69.6)	1573 (67.0)	1678 (69.8)	1709 (71.2)
Score ≥75th percentile	556 (30.4)	573 (33.0)	594 (30.2)	583 (28.8)
	N (Mean)	N (Mean)	N (Mean)	N (Mean)
Age (in years)	1374(19.15)	1560(20.25)	1723(21.24)	1869(22.61)
C-RDS (0–21)c	1253(7.6)	1429(20.25)	1595(7.5)	1751(7.7)
Dep. symptomsc,d	1370(49.6)	1561(51.1)	1722(50.5)	1872(50.1)
Psy. symptomsc,d	1368(15.9)	1562(16.3)	1719(50.5)	1869(15.7)

Notes. Please see full Table 1 including additional demographic variables and Confidence Intervals in the Appendix.

^aPercent values are weighted

^bDep. Symptoms and Psy. Symptoms were analyzed as dichotomous variables.

^cThe analysis was limited to participants who reported having an independent driver’s license by means of domain analysis.

^d.Dep. Symptoms and Psy. Symptoms were analyzed as continuous variables.

DWI: Driving while impaired; C-RDS: Checkpoints Self-Reported Risky Driving Scale; Dep. Symptoms: Depressive symptoms; Psy. Symptoms: Psychosomatic symptoms.

Table 2 shows the cross-sectional associations of risky driving (indicated by C-RDS) with depressive symptoms and psychosomatic symptoms. Higher depressive symptoms and higher psychosomatic symptoms were associated with higher C-RDS scores in all four waves (beta (β); Depressive symptoms: β and p value ranged from 0.06 to 0.11 and p<.001 to p<.05; Psychosomatic symptoms: β and p value ranged from 0.08 to 0.23 and p<.001 to p<.05). Table 3 shows the results of cross-sectional associations of DWI with depressive symptoms and psychosomatic symptoms. Higher depressive symptoms were associated with greater DWI likelihood among emerging adults in all four waves (adjusted odds ratio (AOR) ranged from 1.02–1.03 and p value ranged from p<.001 to p<.05). Higher psychosomatic symptoms were related with greater likelihood of DWI among emerging adults in three of the four waves (W5, W6, and W7) (AOR ranged from 1.04 to 1.05 and p value ranged from p<.001 to p<.01).

Table 2.

Linear regression of C-RDS on depressive symptoms and psychosomatic symptoms

	Wave 4	Wave 5	Wave 6	Wave 7
	βa	βa	βa	βa
Depressive symptomsb	0.11***	0.08*	0.06*	0.09***
Psychosomatic symptomse	0.23***	0.16**	0.08*	0.12***

Notes. Please see full Table 2 including Confidence Intervals in the Appendix.

^*p<.05

^**p<.01

^***p<.001

^aModels were adjusted for sex, race/ethnicity, and family affluence, parental highest education and urbanicity.

β = Beta.

^bDepressive symptoms: High score indicates higher depressive symptoms score

^cPsychosomatic symptoms: High score indicates higher psycho-somatic symptoms score.

Table 3.

Logistic regression of DWI on depressive symptoms and psychosomatic symptoms

	Wave 4	Wave 5	Wave 6	Wave 7
	AORa	AORa	AORa	AORa
Depressive symptomsb	1.02*	1.03***	1.03***	1.03**
Psychosomatic symptomsc	1.05	1.05*	1.05***	1.04*

Notes. Please see full Table 3 including Confidence Intervals in the Appendix.

^*p<.05

^**p<.01

^***p<.001

^aModels were adjusted for sex, race/ethnicity, and family affluence, parental highest education and urbanicity.

AOR-Adjusted Odds Ratio.

^bDepressive symptoms: High score indicates higher depressive symptoms score

^cPsychosomatic symptoms: High score indicates higher psycho-somatic symptoms score.

Significant interactions between three demographic variables (i.e., sex, urbanicity, and parental education) and depressive and psychosomatic symptoms are presented for C-RDS (Supplementary Table 1) and DWI (Supplementary Table 2), respectively. Not all interaction terms are statistically significant, and the patterns of those significant interactions are different across waves for each outcome variable.

DISCUSSION

The study findings show that higher levels of depressive symptoms and psychosomatic symptoms were significantly associated with greater likelihood of risky driving and DWI among emerging adults at each of the four years after high school, with the exception of psychosomatic symptoms and DWI at W4. Although we cannot draw a causal conclusion from the results of this cross-sectional study, it is plausible that depressive and psychosomatic symptoms could play a role in DWI and risky driving behaviors during the transitional period from adolescence to early adulthood. Risky driving and DWI may be distal outcomes of depressive and psychosomatic symptoms via higher use of alcohol and substances and more social and pyschological problems among those with depressive or psychosomatic symptoms.

Although there are few studies on the association between mental and psychosomatic health and risky driving among adolescents, the findings of this study are consistent with one study indicating that major depression was significantly predictive of alcohol-impaired driving among male adults (aged 21–64 years) (Pogue et al. 2017). It has been well documented that mental health disorders, especially depressive symptoms, have been shown to be highly related to clusters of unhealthy and risky behaviors (e.g., smoking, unhealthy diet, heavy drinking, and physical inactivity) (McDonald et al. 2014). Also, research demonstrates that psychosomatic health can be directly or indirectly associated with risky behaviors (e.g., repeated drunkenness, high tobacco consumption, and illicit drug use) (Choquet and Menke 1987) and psychiatric disorders (e.g., depression and anxiety disorders) in early adulthood (Bohman et al. 2012). Although future studies on the mechanism about how mental and psychosomatic symptoms are associated with DWI and risky driving are still underway, we propose that poor mental health may undermine one’s behavioral and cognitive functions such that mental health challenges may increase susceptibility to drinking and substance use to “self-medicate or suppress” (LaBrie et al. 2010) negative affect and/or lead to impaired decision-making (Rubinsztein et al. 2006). The increased susceptibility among individuals with poor mental health to drinking and substance use and impaired decision-making may in turn heighten the risk of DWI and risky driving.

We examined the moderation effects of demographic variables on the associations of depressive and psychosomatic symptoms with DWI and C-RDS. However, not all paired interaction terms are significant across all waves and we failed to see consistent patterns of those significant interactions in different waves. This indicates that demographic variables (e.g., sex, urbanicity, and parental education) may influence the associations of mental and psychosomatic health on DWI and risky driving. However, the effects may be dynamic over different ages. Further studies are needed to clarify the influence of individual characteristics on the associations of mental and psychosomatic health with teen DWI and risky driving.

There are four key limitations in our study. First, we examined DWI as a dichotomous outcome due to sparse and skewed responses. In addition, we used the overall DWI with combining alcohol, marijuana and illicit drugs specific DWI due to the overlapping responses to the three types of DWI. As a result, the transition probabilities and their correlates are only for moving from no DWI to any DWI, with no indication of changes in number/times or number/types of drugs used while driving. Second, the measurement of DWI and C-RDS were based on self-reported questions, which may be influenced by participants’ temporal mood and recent experiences. Third, we conducted a series of cross-sectional analyses over time rather than a longitudinal analysis due to the inconsistent sample sizes, along with more people obtaining independent driving licenses across waves. Fourth, although we observed statistically significant associations of depressive and psychosomatic symptoms with DWI and C-RDS among this population, it is unclear whether the associations will have any clinical significance and if so, to what extent. Given that this is a nationally representative sample, its plausible to consider that the associations identified here confirm the links between depressive and psychosomatic symptoms and DWI and risky driving to some measurable extent. However, more evidence is needed to justify the clinical significance of these links.

The high prevalence of mental (Mojtabai et al. 2016) and psychosomatic (Twenge 2015) health problems combined with high prevalence of teen DWI (Li et al. 2013) and risky driving (Simons-Morton et al. 2005) in U.S. young adults indicates the importance of understanding how mental health conditions and driving inter-relate. Most mental health disorder begin during the period of early adolescence to early adulthood 12 – 24 years old. However, mental health care needs in U.S. youth are not met due to the lack of resources available to adolescents with mental health issues (Cummings et al. 2013). Nurses and nurse practitioners can be the advocates and health educators for families, adolescents, schools and communities to promote and facilitate mental health assessment and early identification and diagnosis of mental health conditions as well as facilitating the successful transition from pediatric to adult care (Hawkins-Walsh and Van Cleve 2019). For example, available nurses and nurse practitioners at school- and college- based health centers can provide services for identification of mental conditions as well as timely treatment for the needs of adolescents in school and communities (Bains and Diallo 2016). Assuming possible causal associations, identifying as well as treating depressive symptoms and/or psychosomatic symptoms early may be effective methods for curbing DWI and risky driving among young adult drivers in the U.S. Reducing DWI and risky driving may lower the high rate of injury and fatal motor vehicle crashes among this vulnerable population.

Depressive and psychosomatic symptoms were associated with greater DWI and risky driving in all four post-high school years. Identifying and addressing poor mental and psychosomatic health should be taken into account when considering the development and implementation of DWI and risky driving prevention programs among youth.