Participants
Sixty-four participants aged 18 to 54 years with a UK driving licence were recruited to this study. Participants in the ADHD group (N = 29, 12 females) were patients undergoing assessment or receiving care from the Adult ADHD Clinic, Nottinghamshire Healthcare NHS Trust, U.K. Of 74 potential participants initially contacted about the study, 21 did not drive, 18 were not interested or did not feel able to take part, 1 was outside the age criteria (18 to 55 years) and 5 withdrew after initially agreeing to take part, providing a sample of 29. Participants in the ADHD group had a confirmed lifetime and current diagnosis of ADHD according to DSM-IV-TR criteria. All were recruited through an adult ADHD service led by a Consultant Child & Adolescent Psychiatrist (CH) with extensive experience of assessment and diagnosis of ADHD across the lifespan. Diagnosis was made as part of thorough clinical assessment prior to referring patients for participation in the study. Current and lifetime ADHD diagnosis was established by conducting the Diagnostic Interview for ADHD in Adults (DIVA 2.0 [33]), childhood developmental history and comprehensive psychiatric assessment.Scores were also obtained from the Conners Adult ADHD Rating Scales (CAARS self- and observer-report) [34] and Autism Quotient (AQ) [35]. All participants met current and lifetime criteria for ADHD diagnosis.
Of the 29 participants with ADHD, 17 were taking stimulant medication, two were taking non-stimulant medication (atomoxetine), one was taking bupropion and nine were not taking any medication. The study was approved by the Nottingham Research Ethics Committee and by the Research & Development department of Nottinghamshire Healthcare Trust. All participants provided fully informed consent.
Control participants (N = 35, 11 females) were recruited through posters displayed on the University of Nottingham campus, U.K. and in local community centres. Volunteers were eligible to take part if they were aged 18 to 55 years, held a driving licence and had never received a diagnosis of ADHD or autism spectrum disorder (ASD).
All participants completed the CAARS (self-report) [34] to assess presence and severity of ADHD symptoms and the AQ [35] to measure signs of ASD. These scales are well-established tools for screening for ADHD and ASD symptoms, demonstrating good test-retest reliability (correlation .89 for CAARS and .7 for AQ) and moderate to excellent internal consistency (Cronbach’s alpha .86 to .92 for CAARS and .63 to 77 for AQ) [35, 36]. Of 29 participants in the ADHD group, 6 met/exceeded the threshold for risk of ASD. These participants were not excluded from the study due to the high rates of ASD symptoms in the adult ADHD population. Instead, the possible influence of ASD symptoms on the dependent variables of interest was examined during statistical analysis. Five potential control participants were excluded from the study as their score on the CAARS self-report exceeded the screening cut-off point for ADHD (score >65 on the CAARS ADHD Index). Of the remaining 30 control participants, none met the threshold for ASD (score >32 on the AQ).
Apparatus and stimuli
Driving simulator and Eye-tracking apparatus
The driving simulator used was the Nottingham Integrated Transport and Environment Simulation facility’s high fidelity system (NITES 1). This driving simulator consists of a fully instrumented BMW Mini housed within a 360° projection dome mounted on a Bosch Rexroth six degrees-of-freedom motion platform. The wing mirrors of the vehicle contain LCD screens with a graphical representation of the rear view and a sound system provides realistic vehicle and traffic noise.
The driving scenario consisted of three different road types. The scenario started in a built-up urban area, which required constant shifting of attention and monitoring and evaluation of performance. After driving 2.1 miles in the urban area, participants reached a single carriageway, which they drove on for 4.6 miles until they reached a three lane motorway which was followed for 9.8 miles. The motorway section contained little traffic, therefore providing a monotonous and low stimulation environment. The single carriageway section was included only to improve the realism of driving from the urban area onto a motorway. All three parts of the drive contained short sections with a speed limit that was lower than the default speed limit for that road type (i.e. 20 mph in the urban area and 40 mph on the carriageway and the motorway). Along the route, five events were programmed to occur, three in the urban area and two on the motorway. Examples of these are pedestrians stepping onto the road and a car suddenly pulling out in the urban area and sudden slowing down of traffic due to an accident on the motorway. Continuous driving performance measures obtained from the simulator were average speed, the proportion of the distance travelled in excess of the speed limit (in excess of 10 % of the speed limit for that road section plus two miles), the coefficient of variation of velocity and the standard deviation of lateral position.
Eye movements were recorded simultaneously by two Seeing Machines FaceLAB 5 eye tracking systems using four cameras. Measures obtained were mean fixation duration and the standard deviation of gaze coordinates (spread of search) for both the horizontal and vertical axes.
Self-report measures of driving performance
The Manchester Driving Behaviour Questionnaire (DBQ) [37], one of the most widely used questionnaires in driving research, was used to assess self-reported driving. It consists of items measuring three different components: errors, violations and lapses. Errors reflect mistakes due to misjudgements and failures of observation such as failing to check mirrors before changing lanes or attempting to overtake someone that you hadn’t noticed to be signalling a right turn. Violations are deliberate deviations from what is considered to be safe driving (illegal or not) such as speeding, tailgating, undertaking and jumping a red light. Lapses refer to less serious failures of attention or memory such as taking the wrong exit on a roundabout after misreading the signs or having no recollection of the road you have just been travelling or where you parked your car. Violations and errors are significant predictors of self-reported accidents [38].
Observational measures of driving performance
To obtain a more detailed indication of participants’ driving behaviour, a coding scheme consisting of 23 items in seven categories was developed to score driving errors and violations (see Table 3). One independent observer naïve to the purpose of the study and blinded to group allocation coded all 43 videos showing a reconstruction of the participant’s drive with their eye movements overlaid as well as a video (including audio) of the participant. Additionally, responses to events in the urban section were categorised as ‘crash or near miss’, ‘appropriate response’ (slowing down or changing lane) or ‘missed hazard due to fast driving’. Responses to events on the motorway were coded as ‘did not slow down’, ‘slowed but not below speed limit’ or ‘slowed below speed limit’. A second observer coded 55 % of the urban and motorway driving scenarios. Inter–reliability scores were calculated and Kappa ranged from 0.70 (observed behaviour in the seven categories) to 0.87 (responses to events). Only the data of the first observer are reported. Finally, the start and duration of ocular fixations following an event were coded, however the inter-rater reliability was poor and so the data were not entered into statistical analysis.
To gain an additional measure of driving behaviour spontaneous comments made during the drive were transcribed and processed by the Linguistic Inquiry and Word Count (LIWC) text analysis software [39]. This software calculates the degree to which different categories of words occur based on a dictionary of around 4500 words and word stems with 82 language dimensions. Previous research has shown excellent levels of internal consistency (e.g. Cronbach’s alpha = .97 for emotion words) (measured as the consistency between words within a category) of this software and acceptable external validity [39]. The categories used in the present study were positive emotion, negative emotion, anger, swearing and anxiety. The frequency of each category of words was recorded for each participant (words can be counted in more than one category). In addition we computed the total number of comments made and the number of words per comment.
Procedure
All participants in the ADHD group who were on stimulant medication withdrew from medication for 24–36 h (depending on their medication regime) before the study simulator session. Before taking part in the study, participants were asked to fill in an informed consent sheet, a driving and health questionnaire, the CAARS (self-report), AQ and DBQ. After filling out the forms, participants were seated in the simulator and were given instructions about its use and safety procedures. Following calibration of the eye trackers a 5-min practice drive was completed with the experimenter present in the dome. Automated verbal directions were given throughout the route. After completion of this drive, participants completed the 16-item Kennedy Simulator Sickness checklist Questionnaire (SSQ) [40] to ensure they were not experiencing too many symptoms of simulator sickness at this point. In case of a high score, participants were withdrawn from the remainder of the study. The experimenter then left the dome and after a short break during which the motion platform was activated, the participants drove the experimental part of the route, which took about 25 min. After completing the experimental drive, participants were asked to fill in the simulator sickness checklist again as well as a post-trial consent form to ensure that any feelings of discomfort they may have experienced had subsided. Participants received an inconvenience allowance for taking part.
Data analysis
Self-report measures of driving behaviour and driving history
Scores on each factor of the DBQ (errors, lapses, violations) and items assessing driving history were compared between ADHD and control groups using independent-samples t-tests.
Simulator driving performance and eye movement measures
The continuous driving performance measures (average speed, proportion of distance travelled in excess of speed limit, coefficient of variation of velocity and standard deviation of lateral position) and the eye movement measures (mean fixation duration, standard deviation of gaze coordinates (spread of search) for both the horizontal and vertical axes) were each entered into separate mixed design ANOVAs. Each ANOVA was designed to assess the between-subject effect of Group (Control, ADHD) and the within-subject effect of Road Type (Urban, Motorway) on these variables, and the Group*Road Type interaction.
Observational coding of driving behaviours
Driving performance
The categories identified from the observational coding of driving behaviours were compared between groups by first computing the total frequency of each behaviour within each category and then comparing the group mean frequencies using multivariate ANOVA across items within each category.
Responses to events
To compare the type of response to urban and motorway events between groups, the chi-square statistic was computed to determine whether there were group differences in allocation of participants to categories.
Emotional speech
The frequencies of each category of verbal responses measured using the LIWC software was compared between groups using univariate ANOVAs.
In all analyses the threshold for significance was .05 two-tailed. Where group differences are reported, secondary analyses were performed to examine correlations between the dependent variable and scores on the hyperactivity/impulsivity and inattentiveness sub-scales of the CAARS in the ADHD group, computing Pearson’s r or Spearman’s rho depending on the data type. These analyses were performed to determine whether each aspect of impaired driving was explained by variability in hyperactivity-impulsivity or inattention symptoms or both.
Participants on atomoxetine (n = 2) were not withdrawn from medication for the study. To ensure this did not influence the findings, all analyses were re-run excluding these participants. The findings remained the same and the participants were not group outliers (defined as >2.5 SD from the group mean) on any measure. The results are therefore reported with these participants included.
The possible influence of comorbid ASD symptoms on the pattern of results was checked by re-running all ANOVAs for which there was a main effect of Group, with AQ scores included as a covariate, or by computing correlations between dependent variables and AQ scores. The analyses confirmed that AQ scores were not significantly related to any DV and did not alter the pattern of group effects reported in the results section. These secondary analyses are therefore not included in the Results section.