Scoping Review of the Driving Behaviour of and Driver Training Programs for People on the Autism Spectrum

Gaining a driver's licence represents increased independence and can lead to improved quality of life for individuals and their families. Learning to drive a motor vehicle and maintaining safe on-road skills are often more difficult for people on the autism spectrum. Many countries currently have no autism-specific licencing requirements for learner drivers, and there is a general lack of ASD-specific support and training packages for individuals, their families, and driving instructors. This review synthesises the peer-reviewed literature about the driving characteristics of drivers on the spectrum and driver training available for the cohort. The evidence in this review showed that individuals on the autism spectrum drive differently from their neurotypical counterparts. There are shortcomings in tactical skills of drivers on the autism spectrum, but the extent to which this affects their own safety or the safety of other road users is unclear. Tactical skills can be improved through training programs. There are few autism spectrum-specific learner training programs available. Development of an effective training program will benefit individuals on the spectrum to learn to drive, be independent, and be safe on the road.


Introduction
People on the autism spectrum experience problems with social and communication skills that are associated with executive function deficits affecting working memory, motor coordination, attention, planning, mental flexibility, and visual perception [1,2]. This means that learning to drive a motor vehicle and maintaining safe on-road skills are often more difficult for people on the autism spectrum. Many people on the autism spectrum face barriers to community inclusion, which can create lifelong obstacles to experiencing individual independence. In most countries, gaining a driver's licence represents increased independence and can lead to improved quality of life for individuals and their families [3]. In fact, a recent retrospective longitudinal study showed that fewer people on the autism spectrum had a licence and generally, they obtained their licence later when compared to their peers [4]. Although recent estimates suggest that prevalences of people on the autism spectrum in Western countries range from 0.6% to 2.2% [5][6][7]; as far as we are aware, in the Australian context, there are currently no evidence-based autism-specific driving interventions available to people on the spectrum to help counter community mobility limitations that affect social and community participation. This review synthesises the peer-reviewed literature about the driving characteristics of drivers on the spectrum and driver training available for the cohort.
Driving plays a critical developmental role during the transition to adulthood and is associated with greater chances of securing paid employment, accessing education, training and services, and maintaining social relationships [8,9]. Learner drivers on the autism spectrum demonstrate problems with driving skill acquisition and performance and typically require more, and longer, driving lessons and need to take the practical driving test more often when compared to their neurotypical peers [2,3]. Further, research shows that people on the autism spectrum are also at greater risk of being involved in motor vehicle accidents, which poses risks not only for them but also for other road users in the community [10,11].
Driving requires higher-order executive function mechanisms to work in a coordinated fashion to respond to, manage, and cope with multiple demands and unexpected and unpredictable driving situations [10]. Research has shown that people on the autism spectrum may experience executive function deficits [2,9], which will directly impact on their ability to develop driving skills and drive safely. Research investigating the driving behaviour of people on the autism spectrum found that they notice driving hazards but they displayed a delay in response to social hazards that required interactions with other road users, such as failing to give way to pedestrians who show clear intent to cross the road [12][13][14]. This is thought to be related to a combination of reduced processing of social stimuli and attention deficits, which is very problematic when learning to drive a motor vehicle-a process which requires collaborative problemsolving skills [3]. These challenges are further confounded by research supporting the notion that people on the autism spectrum have atypical eye-gaze patterns, such as prolonged fixation of the speedometer to check the car speed and scanning repeatedly on traffic-irrelevant roadside objects such as advertisement billboards [2].
Three reviews of the literature covering various aspects of driving and use of public transportation in people on the autism spectrum have been published. The first review by Classen and Monahan [15] presented an overview of evidence-based intervention and predictor studies of driving performance in young people with attention deficit hyperactivity disorder (ADHD) or autism spectrum disorder (ASD). A total of ten papers, published between 1995 and 2011, met the criteria for inclusion in this review with nine of the ten papers focussing on driving in people with ADHD. Only one observational case-control study included ASD participants and focussed on hazard recognition in young people with and without ASD, using a virtual reality video display interface [14]. Thus, this first review highlighted the paucity of studies in the area and the urgent need for more empirical studies. The second review focussed on barriers and facilitators associated with driving and public transportation in people with ASD [9]. The review included 14 studies and narratively summarised findings related to (1) difficulties faced in obtaining a driver's licence (e.g., handling unexpected changes, sustaining attention for long drives, merging into traffic, and limited ability to read facial expressions and gestures), (2) driver confidence (e.g., due to lower reaction time and anxiety), (3) driving behaviours (e.g., avoiding heavy traffic or highways or driving at night, adherence to speed regulation, and lane maintenance), and (4) strategies advocated to improve driving skills in ASD (e.g., direct communication, minimal verbal correction, and-short duration lessons; introduction to driving in low noise density/quiet areas; and strategies to address anxiety). A third review focussed on the likely effects that ASD has on driving abilities in young people [16]. The authors reported that drivers on the autism spectrum (1) were less likely to identify social hazards (i.e., males with ASD), (2) had increased reaction times, (3) had more tactical driving difficulties, (4) reported more traffic crashes, citations, and intentional driving violations, and (5) had poorer situational awareness skills than drivers without autism. A major implication of this review was the need for strategic and systematic approaches to train people on the autism spectrum to drive.
In summary, the body of knowledge on driving and ASD needs further consolidation. A comprehensive review of both empirical and grey literatures that focuses on barriers and enablers to obtaining a licence that are related to (a) people on the autism spectrum (skills, behaviours), (b) the learning environment (policy, driver-training practice, instructor, and physical/built environments), and (c) within-the-person-environment context is needed to guide the design of a training intervention on the person with autism spectrum disorder, as well as their driving instructors [1,9]. Many countries currently have no autism-specific licencing requirements for learner drivers and there is a general lack of ASD-specific support and training packages for individuals, their families, and driving instructors [3,4,17].

Methodology
2.1. Aim. The aim of this scoping review was to synthesise the current peer-reviewed research literature about the driving behaviours of individuals on the autism spectrum and explore the available training programs for this cohort with a view to suggest future research that promotes better training outcomes for learner drivers on the autism spectrum.

Design.
A scoping review of contemporary peer-reviewed research articles published in English between 2000 and August 2017 was undertaken. Seven databases were searched: PubMed, Scopus, ProQuest, Embase, Medline, CINAHL, and TRID.

Search Strategy.
The following search terms were used as appropriate, for each database: "Child development disorders, pervasive" OR "child developmental disorder * " OR "pervasive developmental disorder * " OR autism OR Asperger OR "autism spectrum disorder * " OR "autistic disorder * " OR "Developmental disabil * " AND "Automobile Driving" OR driving or "driv * training" OR "driver behaviour" OR "driver performance" OR "driver characteristic" OR "driv * education" OR "driv * testing" OR "driv * procedures" OR "driving hazard * " OR "car driv * "OR "automobile driv * " OR "driv * packages".

Study Selection.
The search yielded 1389 results from all seven databases. After removing 609 duplicates, a total of 780 articles were screened at an abstract level for inclusion. Studies were excluded if they did not involve people with ASD and if they did not focus on an aspect of driver training or driving skill development. After two authors reviewed the abstracts, 56 articles were deemed to be suitable for review at full-text level. Of the 56 studies reviewed at full-text level, 28 were excluded based on the following exclusion criteria: (1) not peer-reviewed article, (2) no full text available, (3) focus on other disabilities (not ASD), and (4) focus on transportation rather than independent driving. This resulted in a total of 28 studies for analysis; Figure 1 provides an illustration of the review process.

Results
All articles were grouped into four broad categories based on all author-agreed interpretation of each study's findings. These categories included (1) on-road driving behaviours and transport statistics reports on drivers on the spectrum (n = 7), (2) performance in driving simulators (n = 9), (3) performance in virtual reality driving (n = 9), and (4) barriers to obtaining a licence and training of drivers on the spectrum (n = 3). Although driving simulators represent a virtual reality context, the simulator category differs from the virtual reality driving category as the context of the latter was using devices such as a desktop computer, rather than a fully functional simulator. Between the two assessment regimes, the low-cost computer-based driving simulator has a better face validity in observing driving behaviours. Descriptive information was extracted from all articles and included the authors, year of publication, study design, objectives, population, and key findings. Study designs were described using the decision matrix recommended by the Centre for Evidence-Based Medicine [18]. This information is summarised in Table 1.
There was not a wide distribution of empirical studies from around the globe, with over 70% of the studies from the USA (n = 20), followed by Australia (n = 3), the UK (n = 3), Canada (n = 1), and Sweden (n = 1). Research on the topic area was mainly conducted in high-income countries. A total of 1481 participants on the autism spectrum were included across all 28 studies, with 79% of those participants being male. In terms of research designs, there were a total of 23 observational studies, two case studies, two review articles, and one driving simulator study with a quasiexperimental design. The published case study [19] used the same single male subject on the autism spectrum that was used in a comparison with a neurotypical male peer of the same age [20]. In the quasi-experimental study, the intervention group consisted of novice drivers on the autism spectrum (n = 51) and the comparison group (n = 333) was experienced licenced drivers without autism aged over 25 years [21].

On-Road Driving Behaviours and Transport Statistics
Reports on Drivers on the Spectrum. Drivers on the autism spectrum were reported to have proportionally more traffic offences than their neurotypical counterparts (p < 0 05) [22]. This study also reported that drivers on the autism spectrum obtained their licence later than other drivers and self-rated their driving skills on a 10-point Likert scale as lower. Cox et al. [21] reported on a greater proportion of traffic offences in drivers on the autism spectrum. Other studies reported that drivers on the autism spectrum demonstrated decreased manoeuvring ability, particularly in leftand right-hand turns and an increased response time to traffic hazards, particularly in circumstances that required interaction with other road users (e.g., being hesitant to merge into another lane when other drivers had already gestured and reduced speed to allow the manoeuvre to happen) [14,23]. Chee et al. [23] also found that drivers on the autism spectrum performed better than neurotypical drivers in rulefollowing aspects of driving, such as using the indicator and checking for traffic when approaching an intersection.
Performing complex driving functions that required multitasking skills (e.g., merging and using roundabouts) were often more difficult for the person on the autism spectrum [23,24]. Driving tasks, or situations, that were sometimes difficult for drivers on the autism spectrum included driving in heavy traffic, night driving, maintaining the correct speed, lane maintenance, judging distance, and undertaking long journeys. In particular, drivers on the autism spectrum sometimes struggled to interpret the driving actions of other road users and found slight deviations from traffic rules of other drivers-a challenge and anxiety provoking [23].

Performance in Driving Simulators.
All of the studies in this category used a comparison group of drivers not on the autism spectrum, except for one study by Monahan et al. [19]. Bishop et al. [12] and Brooks et al. [1] reported that in comparing drivers on the spectrum and the drivers in the control groups, there were no between-group differences in reaction time to hazard perception and motor response time during predriving assessments. In Brooks et al.'s [1] driving simulator study, when compared with the neurotypical controls, participants on the autism spectrum required an additional 30-35 minutes to complete 18 rounds of steering and pedal skill exercises.
Other driving simulator studies reported no betweengroup differences in (1) errors of maintenance of lane position and speed, (2) adjustment to distractions and poorer right-sided visual acuity [10], (3) response time in braking and overall driving ability [2], and (4) interpreting the gap between the front car, the speed, and the traffic flow of specific traffic scenarios [19]. Reimer et al. [25] also reported that participants on the autism spectrum showed different eye-gaze patterns. When responding to added cognitive demands, they positioned their vertical gaze higher and toward distant objects with more visual diversion. This can reduce the detection of hazards on the peripheral visual field of the individuals. The only quasi-experimental study included in the review used a driving simulator to training drivers on the spectrum [21]. The study showed that drivers on the autism spectrum had poorer baseline executive function and underperformed in tactical skills during simulated scenarios of unanticipated traffic demands [21]. Participants were trained in stages using a driving simulator program during one-hour training sessions across 10-12 weeks, focusing on alternated training of executive functions and tactical driving skills in driving. Outcome measures from the study showed that the training program significantly improved both executive function and driving skills.

Performance in Virtual Reality
Driving. These studies all used some form of a desktop computer and screen as the basis of the virtual reality driving experience with drivingstyled consoles (e.g., Logitech™ steering wheel) attached, to enable the participant to mimic driving. Six of the seven studies were led by various members from the same US Electrical Engineering and Computer Science research team. Many were proof of concept studies to determine whether the various data collection methods and tools can be augmented with a virtual reality program to detect between subject differences in areas, for example, such as eye gaze, EEG data, and physiological responses. The seventh study was from a team of UK researchers and reported that participants on the autism spectrum did not detect the time to arrival of two moving cars on a straight road as accurately as the comparison group [26]. Although this study used a virtual reality driving program, none of the participants were actual drivers and so the ability to draw real-life conclusions is limited.

Barriers to
Obtaining a Licence and Training of Drivers on the Spectrum. Both drivers with ASD (learners and with a licence) experienced greater problems with driving and require more lessons and more tests than neurotypical drivers. Learning to drive was sometimes more difficult for the person on the autism spectrum as the process of reading, understanding, and converting driving theory into driving practice was demanding [3]. Unexpected changes to the usual routine or driving norms-where other drivers did not obey the traffic rules-were problematic for people on the autism spectrum [3]. Other reported problems included a lack of confidence, overconfidence, sensory overload, anxiety, poorer concentration, and greater distractibility.
Although all the research studies in this category were either observational or literature reviews that focussed on barriers, there were a few suggestions of potential ways to overcome these barriers. These included strategies to assist driving instructors, such as using direct communication and concrete instructions, providing shorter driving lessons, to commence each lesson in a quiet neighbourhood, and only giving the minimum amount of required information at a time [9]. There was a broad suggestion to use strategies to address anxiety; however, these would need to be individualised to the person on the autism spectrum. In Tyler's [17] four case studies with young adult males on the autism spectrum, some of these individualised strategies included using parent/learner/instructor communication logbooks to enhance consistency, structured language, tasks analysis, and breakdown, using a range of "what if" scenarios to increase insight into unpredictable driving and traffic situations and using visual markers to aid with judging distance.  Figure 1: Flowchart of included studies. Table   1 Authors, Use of a sample that was nonrepresentative; data collection is limited to internet survey. (ii) 29% who are eligible to drive currently drive.
Respondent bias (use of caregivers other than the person on the spectrum) limits validity. Selection bias leading to (iii) The interactive exercises and the process used worked well to Small sample size and underpowered Females are excluded in the ASD group.
Unclear of any real-life translation: interactive exercises do not include any road scenes that help to eliminate the sensory overload, anxiety, and stress that participants might experience in naturalistic driving environments  (ii) Youth with HF-ASD displayed a nominally higher and unvaried heart rate compared to controls.
(iii) Youth with HF-ASD showed a gaze pattern suggestive of a diversion of visual attention away from high-stimulus areas of the roadway when responding to increased cognitive demands (cell phone task; continuous performance task).
(iv) Youth with HF-ASD tended to position their vertical gaze higher and away from nearer objects and more toward the distance than controls. Sample of convenience-Caucasian comparison groups are not representative of the population of interest-limits generalisability of results.
Small sample size: underpowered study Without screening on age and cognitive functions, a convenient sample from the community as the control group Assessment of intellectual functioning or ASD was not obtained. Degree to which outcome measures relate to other covariates such as driving frequency and exposure is not reported. (ii) Participants with ASD were slower to respond than those without ASD.
Participants of the study are likely to have strategies prior to the experience to identify and respond to both types of hazard.
Limited to only male participants Results cannot be generalised to ASD drivers with a licence or of different age ranges or females with ASD. (ii) There were no differences between participants with ASD and the comparison group when the simulation was along a curved road.

Sheppard
Participants were all nondrivers so generalisability to drivers is not possible.
Unclear of any real-life translation Study included the collection of both driving and public transport data, with sufficient data about driving to be included in this review.

Discussion
There is inconsistent evidence on the driving ability of people on the autism spectrum. For example, some studies reported that there is no difference between drivers on the autism spectrum and neurotypical counterparts in the maintenance of lane position and speed [8], response time in braking, and overall driving ability [2]. These studies were conducted either in simulated driving or in virtual reality environments generated by computer desktops. Conversely, when the ability of drivers was assessed in naturalistic on-road driving environments, drivers on the autism spectrum experience difficulty in complex traffic scenarios, especially driving tasks that required nonverbal communications with other drivers [23]. In addition, when compared with neurotypical controls, drivers on the autism spectrum had difficulty in multitasking and performing complex driving tasks and consistently maintaining a safe speed. They were also observed to have decreased manoeuvring ability and increased in response time to traffic hazards [23]. Unexpected changes to the usual driving routine or driving norms from other drivers distracted the attention of drivers on the autism spectrum and induced performance anxiety. When responding to added cognitive demands, drivers on the autism spectrum had poorer baseline executive function and performed worse on general and tactical driving [23]. Similarly, discrepancies were found by Cox et al. [21] who reported a greater proportion of traffic offences in drivers on the autism spectrum, whereas Chee et al. [23] found that drivers on the autism spectrum performed better than neurotypical drivers in rule-following aspects of driving. A possible explanation for these discrepancies is that the Cox et al. [21] study observed drivers on the autism spectrum using a driving simulator assessment, which is an artificial context. Consequentially, the real-life behaviour changes of drivers on the autism spectrum in responding to high-demand traffic scenarios were not known. This notion is supported in a validity study using driving simulators to assess driving behaviours of older drivers. Lee et al. [27] reported that only 67% of driving behaviours observed in a driving simulator was transferable to an on-road driving environment. Simulated driving environments might not be sensitive enough to detect changes of the driving performance. On-road driving is the gold standard to assess and observe ability of drivers. Drivers on the autism spectrum underperform in some aspects of driving or drive differently in naturalistic environments, but there is insufficient evidence to support the notion that drivers on the spectrum are more at risk of being involved in an accident than their neurotypical counterparts [28].
The observational studies [8,26] using cross-sectional designs reported drivers on the autism spectrum to have difficulty in interpreting the driving actions of other road users. The eye-gaze patterns of drivers on the autism spectrum demonstrated an increased response time in detecting hazards, especially those associated with ineffective interpretation of the intentions of other drivers by means of body gesture and social clues. Insufficient attention to the traffic events detected on the peripheral visual fields can be hazardous and unsafe on road [26].
Individuals on the autism spectrum appear to have decreased ability to multitask and manoeuvre through complex traffic scenarios. To accomplish these tasks, drivers on the autism spectrum require training on the tactical level skills of the Michon's model of driving [29]. Tactical behaviour in driving involves striking a balance between the demands of driving and the driver's ability to drive safely in accordance to the road rules [30]. Successful acquisition of the tactical skills may allow drivers on the autism spectrum to reserve their cognitive capacity to cope with tasks with high number of executive functions and decisionmaking demands.
In identifying the training of the individuals on the spectrum, there is no experimental study aimed at testing novel ways to train individuals on the autism spectrum to obtain their driver's licence. In the current review, most observational studies described the barrier individuals on the autism spectrum experience in getting a driver's licence. None of them focussed on investigating the best strategies to help individuals on the spectrum to obtain their driver's licence.

Future
Research. The current literature reviewed provides a preliminary driving profile of individuals on the autism spectrum in simulated contexts or during on-road assessments with predetermined routes. Given that individuals on the autism spectrum are susceptible to stress and anxiety, their driving experience is very likely to deteriorate in a test environment. For this particular cohort, research into the individuals in naturalistic driving environments will generate a more comprehensive driver profile. Driving is essential for individuals on the spectrum, but they showed a relatively low take-up rate of formal driving licence. There is limited research on how to effectively train individuals on the autism spectrum to drive independently and safely. Future research should focus on identifying strategies and the best practice of training to support individuals on the autism spectrum to get their licences.

Limitations.
Most of the studies included in this review used to establish a profile of drivers on the spectrum were based on observations in driving simulator and/or virtual reality settings. This is a major limitation, given the low transferability of observations from a driving simulator environment to real-life on-road driving [27]. Findings from this scoping review should be interpreted with caution, as we only included studies published in English; grey literature, books, and theses were outside the scope of our review. Further, there are eight studies that employed self-report methodology to collect information from individuals on the autism spectrum on their ability to drive. Although there is no incentive for the participants to have falsely reported their driving ability, their reporting may have inflated their ability to drive and may have been influenced by recall bias.

Conclusion
Synthesis of the evidence in this review showed that individuals on the autism spectrum drive differently from their neurotypical counterparts. There are shortcomings in tactical skills of the drivers on the autism spectrum, but the extent to which this affects their own safety or the safety of other road users is not yet clear. Tactical skills can be improved through targeted training programs, specifically designed to accommodate the driving characteristics of the autism population. There are few autism spectrum-specific learner training programs available. Development of an effective training program will benefit the individuals on the spectrum to learn to drive, be independent, and be safe on the road.

Conflicts of Interest
The authors declare that they have no conflicts of interest.