Neurocognitive Correlates of Young Drivers' Performance in a Driving ...

Journal of Adolescent Health 58 (2016) 467e473

www.jahonline.org Original article

Neurocognitive Correlates of Young Drivers’ Performance in a Driving Simulator Stephanie A. Guinosso, Ph.D., M.P.H. a, *, Sara B. Johnson, Ph.D., M.P.H. a, b, Maria T. Schultheis, Ph.D. c, d, Anna C. Graefe, Ph.D. c, and David M. Bishai, Ph.D., M.D., M.P.H. a, e a

Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland Department of Pediatrics, Johns Hopkins School of Medicine, Baltimore, Maryland c Department of Psychology, Drexel University, Philadelphia, Pennsylvania d School of Biomedical Engineering, Sciences and Health Systems, Drexel University, Philadelphia, Pennsylvania e Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland b

Article history: Received June 12, 2015; Accepted December 11, 2015 Keywords: Adolescent; Driving; Executive function; Attention; Driving simulator

A B S T R A C T

Purpose: Differences in neurocognitive functioning may contribute to driving performance among young drivers. However, few studies have examined this relation. This pilot study investigated whether common neurocognitive measures were associated with driving performance among young drivers in a driving simulator. Methods: Young drivers (19.8 years (standard deviation [SD] ¼ 1.9; N ¼ 74)) participated in a battery of neurocognitive assessments measuring general intellectual capacity (Full-Scale Intelligence Quotient, FSIQ) and executive functioning, including the Stroop Color-Word Test (cognitive inhibition), Wisconsin Card Sort Test-64 (cognitive flexibility), and Attention Network Task (alerting, orienting, and executive attention). Participants then drove in a simulated vehicle under two conditionsda baseline and driving challenge. During the driving challenge, participants completed a verbal working memory task to increase demand on executive attention. Multiple regression models were used to evaluate the relations between the neurocognitive measures and driving performance under the two conditions. Results: FSIQ, cognitive inhibition, and alerting were associated with better driving performance at baseline. FSIQ and cognitive inhibition were also associated with better driving performance during the verbal challenge. Measures of cognitive flexibility, orienting, and conflict executive control were not associated with driving performance under either condition. Conclusions: FSIQ and, to some extent, measures of executive function are associated with driving performance in a driving simulator. Further research is needed to determine if executive function is associated with more advanced driving performance under conditions that demand greater cognitive load. Ó 2016 Society for Adolescent Health and Medicine. All rights reserved.

Conflicts of Interest: There are no conflicts of interest to disclose. * Address correspondence to: Stephanie A. Guinosso, Ph.D., M.P.H., 1157 Francisco Street, Berkeley, CA 94702. E-mail address: [email protected] (S.A. Guinosso). 1054-139X/Ó 2016 Society for Adolescent Health and Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jadohealth.2015.12.018

IMPLICATIONS AND CONTRIBUTION

Neurocognitive performance is associated with driving skills, but few studies have empirically examined this relationship. This study provides evidence that measures of general intelligence and cognitive inhibition are associated with more consistent driving in a driving simulator among young drivers. Further research on more challenging driving conditions is warranted.

Motor vehicle accidents are the most common cause of death among young drivers (15e24 years) in the United States with 6,510 young people of this age killed in 2013 [1]. Crash rates in young drivers are related to both immaturity and inexperience [2,3]. Despite progress due to graduated driver licensing and

468

S.A. Guinosso et al. / Journal of Adolescent Health 58 (2016) 467e473

other policy interventions, the first months of independent driving remain extremely dangerous. Crash rates are persistently higher for both the youngest drivers and the most inexperienced drivers; while 16 year olds have higher crash rates compared to 17 year olds, even drivers who begin driving after age 18 years exhibit higher crash rates for the first few months of licensure, highlighting the role of inexperience among young adults [4]. A growing body of research is focused on understanding developmental risk factors for crashes [5,6]. Based on police reports, it is estimated that 23% of crashes among drivers under 20 years are attributable to inattention [7,8]. Attention is a key component of executive functioningda set of supervisory cognitive functions involved in goal-directed behavior that includes working memory, response inhibition, planning, and delay of gratification [9]. Executive function steadily improves throughout adolescence and beyond [9]. Because driving requires significant attention to manage real-world distractions, including conversations with passengers, use of car controls or mobile phones, internal dialogue/mind wandering, and hazards that arise on the road [5,8], immature executive functioning may interfere with driving performance. Despite widespread agreement that cognitive functioning is central to driving competence [5,8,10], relatively little is known about the specific neurocognitive correlates of driving performance in healthy, young drivers [11]. Most studies have focused on older drivers or young drivers with clinical diagnoses that could impair driving performance. Adolescents with attentiondeficit hyperactivity disorder, and those with a history of attention problems in childhood have higher rates of citations, speeding, crashes, referrals to traffic school, and/or license suspensions than their nonaffected peers [12e15]. The relationship between attention-deficit hyperactivity disorder and driving risk appears to be mediated through poor choices, inability to modulate behavior in response to the environment, and/or failure to anticipate consequences [14,15]. Among studies of older adult drivers, neurocognitive testing performance may provide insight into crash risk [16,17]. A study comparing neurocognitive performance among men ages 65 years and older with a history of several recent crashes to those with no crash history found that crashing status could be predicted by performance on measures of executive function for 80% of individuals [18]. Similarly, men and women ages 65 years and older who scored in the bottom 10% on an assessment of cognitive functioning (including attention, reaction time, working memory, and mental flexibility) were 1.5 times more likely to crash over the subsequent three-year period than those who performed at the top 10% [16]. Although it is reasonable to expect that individual differences in neurocognitive performance could be related to driving performance in healthy adolescents and young adults, this area has received little study. In one study, three core aspects of executive functioning (working memory updating, inhibition, and shifting) were examined in relation to teenage driving performance; worse performance on only the working memory updating construct was associated with worse performance on a lane change task while counting backward to increase cognitive demand [19]. A second study showed that young drivers (ages 17e21 years) who were caught speeding, scored higher on a measure of impulsivity compared to nonoffenders [11]. Additional research supports the idea that older adolescents’ cognitive performance is comparable to that of adults under situations of low emotional salience but may be insufficient to override

distraction or social pressure (such as driving in the presence of peers), emotional arousal, or time pressure [10,20,21]. Thus, the neurocognitive correlates of driving performance in a simulator may likely differ based on the nature of the task and setting. In the current pilot study, we explored whether common measures of neurocognitive functioning, including both general intellectual capacity and executive functions, were associated with young driver’s performance in a driving simulator. We investigated whether the relationship between neurocognitive performance and driving differed in the presence of a driving challenge where drivers engaged in a verbal working memory task to increase cognitive load. There were no a priori assumptions about the relationship between intelligence and driving performance. However, we hypothesized that better executive functioning would be associated with better driving performance under both conditions. Methods Participants and recruitment Young drivers were recruited using fliers posted in public places, on a social media site, and in driving schools. Potential participants completed a brief online survey to determine eligibility. Eligible individuals were aged 16e24 years and had a valid learner’s permit or driver’s license. In all, 86 participants were eligible and consented to participate, and 74 participants (86%) completed the study. One participant did not complete the full evaluation due to a scheduling conflict. Eleven participants had corrupted data. Eighty-nine percent of participants were ages 22 years and younger (mean: 19.8; median: 19.0, SD ¼ 1.9). Fortythree percent were female. Participants reported between 0 and 9 years of driving experience (mean ¼ 3.0 years, SD ¼ 1.9). Design Participants 18 years or older provided verbal and written consent; verbal parental consent and participant assent was provided for subjects under 18 years. The Committee on Human Research of the Johns Hopkins University Bloomberg School of Public Health approved the study protocol, consent procedure, and study forms. Subjects completed demographic questionnaires, computer-based and paper and pencil measures of neurocognitive functioning, and drove in a driving simulator. Measures of neurocognitive functioning and simulated driving were counterbalanced. Driving simulator. This study used a high-fidelity virtual reality driving simulator, which comprised a desktop PC and three 32inch high-resolution displays to provide panoramic visual feedback (shown in Figure 1). The steering wheel, gas pedal, and brake pedal were manufactured by Extreme Competition Controls, Inc (ECCI, Minneapolis, MN), and the center console (shifter, cup holders, ashtray, and stereo system) was from a Ford Taurus sedan with an automatic transmission. Measures of real-time driving performance: driving speed, lane position, accelerator, and steering wheel inputs were sampled at 16 Hz. The software was custom-engineered by Digital Mediaworks, Inc (DMW, Ontario, Canada). The virtual environment was composed of daytime dry-pavement driving conditions with good visibility. Data reported here were collected during straight roadway segments of between 3,800 and 5,000 feet.


469

behavior [24]. Standard deviations of the mean, which capture the variability, or consistency, in driving performance, were calculated for repeated measures sampled every 50 milliseconds over the course of the segments of interest. Standard deviations were computed for the following four measures of driving performance: (1) velocity (mph); (2) accelerator position (a percentage of the level of depression, where 0 is not depressed and 1.0 is fully depressed); (3) lane position (distance from the vehicle’s centroid to the center of the driving lane, in inches); and (4) steering wheel position (measured in degrees from the center position). We use the term “variability” to refer to these measures of standard deviation. Variability in velocity and accelerator position are highly correlated, as are variability in lane and steering wheel position.

Figure 1. Virtual Reality Driving Simulator.

Procedure. The driving component consisted of a 10-minute training drive, a 25-minute baseline drive, and a 15-minute driving challenge, in that order. Drivers were not provided with any motive or incentive for their performance. During the training drive, participants learned to operate the simulated vehicle. For the baseline drive, participants were instructed to drive as if they were following normal traffic conventions, including responding appropriately to road and speed signs. The route included a residential, highway, commercial, school, and rural zone. During the driving challenge, participants drove the same rural driving segment as the baseline condition and followed the same instructions. However, participants were asked to complete a verbal working memory task at a predetermined section of the drive to the best of their ability. For this section, the speed was clearly marked 40 mph with no change in speed throughout the segment. We extracted the segment from a participant’s baseline drive that corresponded, in starting point and distance, to the segment in which they completed the verbal working memory task. Verbal working memory task. The verbal working memory task was employed during the driving challenge to mirror conversation while driving as a potential source of distraction and increased cognitive load on executive attention [22,23]. Participants were read two blocks of four sentences by the examiner. After each sentence, drivers were asked to report whether the sentence was congruent (e.g., did it make sense), and they were asked to report the last word of each sentence at the end of each block. This task did not require any motor/physical responses and was designed to reflect cognitive loading without physical demand. Measures Dependent variables: measures of driving performance. The virtual reality driving simulator data for both the baseline and challenge driving segments were analyzed using the Windshield software (DMW, Ontario, Canada), allowing for comparison of identical segments of the driving route “on” and “off” task. Speed control and lane management are commonly assessed in simulated driving studies and reflective of on-road driving

Independent variables: neurocognitive measures. General intelligence was measured by the Wechsler Abbreviated Scale of Intelligence (WASI). The test included block design, matrix reasoning, similarities, and vocabulary subtests, which generated scores for the WASI Full-Scale Intelligence Quotient (FSIQ) [25]. Several neurocognitive measures were used to assess executive function. Cognitive inhibition was evaluated using the Stroop Color-Word Test (Stroop), which assesses susceptibility to interference and the ability to shift perceptual set. Participants were presented with words for colors (e.g., red), displayed in a discordant color and were instructed to identify either the words or colors [26]. The color word t-score was used as an age-normed measure of task performance, with higher scores reflecting better performance. Cognitive flexibility was assessed by the Wisconsin Card Sort Test-64 (WCST-64), which measures problem solving, hypothesis testing, and set shifting [27]. Participants matched playing cards using feedback from the task to alter their behavior as the rules were changed covertly in subsequent matches. Higher scores on the age- and education-normed total errors T-score reflect better performance. Attention was assessed using the Attention Network Task (ANT) [28]. This test evaluates three attention networks: alerting, orienting, and executive control. Efficiency of the alerting network is examined by the changes in reaction time in response to a warning signal. Efficiency of the orienting network is examined by changes in reaction time that accompany cues indicating where a target will appear. Efficiency in executive control network is examined by requiring participants to press two keys indicating the direction (right or left) of a central arrow surrounded by two flanker arrows pointing in the same or opposite directions. For each of the ANT tasks, higher scores reflect worse cognitive performance (i.e., longer reaction time). Covariates. Based on previous literature, participant age (years), sex, driving experience (years), driving frequency (miles driven per week), family socioeconomic status (SES), and drowsiness were used as covariates. SES was measured using the Family Affluence Scale II, which assesses: car and computer ownership, having one’s own bedroom, and number of vacations in the last year. Responses were categorized into low, medium, and high SES [29]. Given that only two participants (2.7%) met the criteria for low SES, a dichotomized measure of high and low/medium SES was used for the analysis. Self-reported drowsiness was assessed using the Modified-Simulation Sickness Questionnaire following the simulator drive [30]. Responses ranged from none, slight, moderate, and severe.

470


Analysis We examined potential systematic reasons for missing data [31]. Missing values for driving experience (8.1% missing), driving frequency (8.1% missing), SES (2.7% missing), drowsiness (1.4% missing), and the WCST-64 (4.1% missing) were determined to be missing at random because factors associated with missingness were observed in the data set (i.e., missingness associated with younger drivers) and imputed with multiple imputation using chained equations. Separate, robust simple linear regressions were used to evaluate the unadjusted relationships between the continuous measures of neurocognitive functioning (independent variables) and the continuous measures of driving performance (dependent outcomes). Then covariates (age, sex, driving experience, driving frequency, SES, and drowsiness) were added to the models. Quantile regression models (using the 10th, 25th, 50th, 75th, and 90th quantiles of the driving outcomes) were used to validate ordinary least squares regression findings for the adjusted models due to non-normal distribution of regression residuals for some of the driving outcomes. Differences between the baseline and driving challenge conditions were examined with the post-estimation procedure, seemingly unrelated estimation [32]. Analyses were performed using STATA version 13 (StataCorp, College Station, TX). Results Descriptive statistics for key study measures are shown in Table 1. Our study sample had higher than average SES with 66% of the sample rated as high SES and an average school attainment of 13.6 years. Participants scored above average on the three agenormed neurocognitive assessments (FSIQ, Stroop, and WCST64) as expected for a high SES sample. Reaction times for the ANT were comparable to others studies of healthy adolescents and young adults [33]. Variability in the cognitive measures was not significantly associated with age. Driving outcomes remained similar between the baseline and verbal challenge segments; only variability in lane position increased significantly between the baseline (2.7 inches) and verbal driving challenge (16.4 inches; t ¼ 20.86; df ¼ 73; p < .000). Neurocognitive predictors of baseline driving performance The relationships between the neurocognitive measures and baseline driving performance are shown in Table 2. Findings were similar for the bivariate and adjusted analyses for both the baseline drive and driving challenge. Adjusted analyses are described in the following sections. Intelligence. After controlling for covariates, higher WASI FSIQ scores were associated with less variability in velocity (b ¼ .35; p < .01), accelerator position (b ¼ .37; p < .001), and lane position (b ¼ .28; p < .05) during the baseline drive. The WASI FSIQ was not significantly associated with variability in steering wheel position. Quantile regression showed the relationships between WASI FSIQ and variability in velocity and accelerator position were robust, with significant associations among the 75th and 90th quantiles of velocity variability and the 50the90th quantiles of accelerator variability. Executive functions. Among measures of executive function, better performance on the Stroop task was associated with less

Table 1 Summary of demographic characteristics, neurocognitive assessments, and driving outcomes of study participants (N ¼ 74) Mean (SD) or percent Demographics Age (years) Driving experience (years) Miles driven per week Education (years) Female Family socioeconomic status Lower Middle Higher History of ADHD Race Caucasian Black Asian Other Neurocognitive performancea Wechsler Abbreviated Scale of Intelligence Full-Scale IQ standard score (all) 16e17 years 18e24 years Stroop ColoreWord Test Color-word t-score (all) 16e17 years 18e24 years Wisconsin card sort task Total errors t-score (all) 16e17 years 18e24 years Attention network test Alerting (all) 16e17 years 18e24 years Orienting (all) 16e17 years 18e24 years Conflict executive control (all) 16e17 years 18e24 years Driving outcomesb Variability in velocity (mph) Baseline Verbal challenge Variability in accelerator position (% of pedal depression where 0 is not depressed and 1.0 is fully depressed) Baseline Verbal challenge Variability in lane position (inches) Baseline Verbal challenge Variability in steering wheel position (degrees from center position) Baseline Verbal challenge

19.8 3.0 107.6 13.6 43%

(1.9) (1.9) (138.0) (1.6)

3% 31% 66% 6.8% 55% 8% 30% 7%

115.0 (10.2) 112.2 (8.0) 115.2 (10.3) 53.9 (10.3) 56.4 (7.8) 53.7 (10.5) 53.7 (8.2) 56.2 (4.0) 53.5 (8.6) 42.8 39.4 43.0 48.1 40.7 48.6 119.6 97.6 121.2

(24.9) (12.7) (25.6) (28.2) (12.7) (29.1) (58.7) (42.6) (59.6)

2.3 (2.0) 2.5 (2.0)

.8 (.0) .7 (.0) 2.7 (4.8) 16.4 (4.8)

1.3 (.6) 1.3 (.5)

ADHD ¼ Attention Deficit Hyperactivity Disorder; IQ ¼ Intelligence Quotient. a Cognitive scores are presented for all ages and for younger and older participants separately. There were no significant differences between the age groups for any cognitive measure. b Variability is defined as the within-person standard deviation of the mean of repeated measures sampled every 50 ms for each driving outcome over the course of the driving segments of interest. The mean (SD) shown in the table represents the mean standard deviation across participants and the standard deviation of that mean.

variability in velocity (b ¼ .25; p < .01) and accelerator position (b ¼ .25; p < .05), and better performance on the alerting assessment was associated with less variability in accelerator position (b ¼ .32; p < .01). Quantile regression showed these


471

Table 2 Relationship between baseline driving performance and neurocognitive measures based on separate regression modelsa Baseline driving condition (N ¼ 74)

Intelligence Wechsler Abbreviated Scale of Intelligence Full-Scale IQ Executive functions Stroop Color-Word t-Score Wisconsin Card Sort t-Score Attention network test Alerting Orienting Executive

Variability in velocity

Variability in acceleration

Variability in lane position

Variability in steering

b (SE)

ß

b (SE)

ß

b (SE)

ß

b (SE)

ß

L.069** (.024)

L.35

L.0015*** (.0004)

L.37

L.131* (.053)

L.28

.010 (.006)

.17

L.049** (.016) .043 (.033)

L.25 .18

L.0010* (.0005) .0004 (.0007)

L.25 .07

.056 (.036) .124 (.099)

.12 .22

.009 (.006) .009 (.009)

.16 .13

.019 (.011) .020 (.012) .009 (.007)

.24 .29 .26

.0005** (.0002) .0002 (.0002)