IJGG Journal

Discussion

In this study we aimed to investigate whether specific motor and cognitive tests performed in clinic can predict behind-the-wheel driving behavior in Parkinson’s disease, through a multidisciplinary driving simulator experiment.

First, our individual feature group comparisons between PD and HC groups indicated that drivers with PD performed worse in both metrics of tactical car control and operational safety, which are also in keeping with prior literature [9-14]. Our results add to our knowledge that people with PD maintain lower average speed and greater headway distance from the vehicles ahead. These behaviors may burden urban networks through traffic jams, further complicating the use of roads by drivers and pedestrians. Reduced speed may also lead to dangerous overtaking by following vehicles, increasing the rate of accidents [15]. Additionally, the increased reaction time observed in people with PD is a critical operational safety metric that has been associated with increased risk of accidents [11,16]. The worse performance observed in executive and visuospatial abilities in PD is in line with previous studies [17-20] and is likely to interfere negatively with dynamic driving demands, both tactical and operational. This can be expressed as difficulties in adjusting driving behaviors according to changing traffic burden, when approaching intersections or traffic lights, changing lanes, or car maneuvering [9,12,21-23].

A novelty in our work involves the detailed inclusion of in-clinic motor tasks (RPW, FTT, TWT, TWT-RC, and HRT), in addition to the UPDRS-III and H&Y, for predicting driving behavior in PD. In contrast to UPDRS-III and H&Y, these tasks can be performed by non-experts. As anticipated, participants with PD performed worse than HC on RPW and TWT-RC, but not on the other metrics (Table 3). This can be explained by RPW being affected by cardinal motor features of PD, namely rigidity and bradykinesia, whereas TWTRC represents both motor and cognitive features of speed, balance, and executive functioning. The lack of significant differences between groups for the other metrics, despite a trend in most, is more likely explained by mild to moderate overall motor symptom severity in our participants (median H&Y = 2), and less so by not capturing relevant clinical signs, which are more pronounced later in the disease course. Motor scales, especially RPW, TWT-RC proved reliable to lesser extent before Bonferroni correction.

Dimensionally reducing our original dataset (Tables 4-6) allowed for interpretation of driving, motor, and cognitive performance into latent conceptual domains. These latent variables provide a degree of abstraction that permits us to describe relations among a class of events or variables that share something in common, rather than making highly concrete statements restricted to the relation between more specific, seemingly idiosyncratic variables. In other words, latent variables permit us to generalize relationships [41]. Some caution is required when interpreting loading contributions to latent factors depending on their sign. Specifically, positive or negative loading signs in Tables 4, 5, and 6 guide interpretation of a factor’s scores as to whether a high factor score indicates better or worse performance. For example, a high score in factor Speed and Dexterity indicates worse performance (Table 5), since its loadings stem from large values in tests regarding speed (i.e., slow performance). Our latent driving simulator factors showed similar loading contributions in high and low traffic loads except for accidents related to tactical driving and reaction time (Table 4). The differential loading of driving simulation metrics observed between low (i.e., wheel positioning, headway distance, average speed) and high (i.e., speed limit violation) traffic loads to tactical driving related accidents can be explained through driving adaptation to dynamic road conditions that have different driving demands. For example, in low traffic loads there is more time-toclosure (τ), defined as the ratio of the current distance-to-target over the current speed towards the target, [42] by maintaining longer headway distances, whereas in high traffic loads, where headway distance has little variability, speed limit violation contributes more to accident outcomes. Similarly, although reaction time contributes similarly to both high and low traffic loads, its contribution in high traffic loads is co-mediated by wheel position manipulation, possibly indicative of more spatial restrictions in high traffic load environments.

In western societies maintaining driving independence is considered an instrumental activity of daily living that enables individuals to engage in activities that are identified as crucial to maintain their quality of life [24]. In order to do so, studies indicate that drivers with PD have developed both strategic and operational compensatory driving behaviors. On a strategic level, interviewbased feedback with patients revealed that their awareness of PD-related symptoms (rigidity, fatigue, reduced concentration) interfering negatively with their driving performance made them avoid driving long distances, plan a priori optimal routes, and plan rest stops [25].  Additionally, they avoid driving in the dark, snow or heavy rain, or crowded urban environments [42, 43]. Behind the wheel studies also indicate that driving behavior outcomes are defined by tactical parameters, which include all voluntary actions made by the driver to maintain safety and avoid potentially dangerous driving situations [18, 26].

Despite that both non-motor and motor symptoms affect driving behavior, it is not a sine qua non that people with PD should be automatically precluded from driving, especially when symptoms are mild [10, 11, 27]. For this reason, we focused on identifying early features and their combinations that interfere significantly with driving behavior, and we thus recruited our cohort of active drivers with PD with mild motor and cognitive signs. Even more, although not part of our exclusion criteria, none of the participants had a history of traffic accidents provoked by them. To that extent, we pursued multivariate regression with backward selection and identified which latent cognitive and motor factors best predicted individual driving behavior domains (Table 7). Of the five driving domains, the performance in three (i.e., Tactical Driving Praxis, Car Spatial Road Position, Reaction Time Related Accidents) could be significantly explained by cognitive and motor symptoms. Tactical driving praxis, a critical parameter of driving behavior, [16] was best predicted by cognitive-motor domains (planning and processing speed, speed and dexterity, and axial movement) whose performance is mediated by frontal subcortical systems [28]. Instead, spatial elements of driving behavior, as represented in Car Spatial Road Position, were mediated predominantly by spatial cognitive abilities that typically involve parieto-frontal dorsal executive networks, as represented in visuospatial attention and planning, and less so by frontal and frontal-subcortical executive function networks of planning and processing speed. Additionally, executive-predominant features contributed to reaction time related accidents, but were co-mediated by learning and memory features that involve temporo-parietal networks as well. Overall, the above suggest that tactical driving praxis and car spatial positioning are voluntary agencies associated with planning and linked to frontal cortical and subcortical networks. Instead, reaction time related accidents, an operational safety feature further mediated by immediate “reflex” re-action, further require an ability to learn and retain information, possibly learning how to avoid environments providing little time to react, and perhaps “remembering” how to react in accident-prone environments.

Since multivariate regression approaches do not account for the combined relationship between the five derived factors of driving behavior on one hand and the five derived clinical motor and cognitive factors on the other while accounting for their interrelations, we pursued CCA. Our CCA on objective data on simulated driving behavior add to our understanding of the aforementioned observations and prior studies by revealing the main dimensions that associate a set of driving behavior features to a set of cognitive and motor performance features (Table 8). Specifically, the predominant association was between driving canonical variate of tactical driving praxis and reaction time related accidents to multidomain cognitive features, especially executive functions of planning and processing speed (~52% of explained variance), as well as axial motor function (~57% of explained variance), which typically localize to frontal and frontal-subcortical networks and are well known to be affected in PD [28]. Nonetheless, significant contributions were also observed in visual attention and planning (~34% of explained variance), likely reflective of parieto-frontal dorsal executive network deficits in PD, verbal memory (~34% of explained variance), which may indicate temporo-parietal network deficits through frequent co-morbid Alzheimer’s disease pathology, and motor speed and dexterity (~29% of explained variance), another characteristic motor feature of PD localizing in frontal and frontal-subcortical networks[28]. The above confirm hypotheses raised by previous studies that both motor and nonmotor symptoms of PD influence tactical and accident related driving behavior [2, 12, 13, 20, 29-33]. Our CCA (Table 8) also revealed that spatial elements of driving (i.e., Car Spatial Road Positioning) together with reaction time accidents were associated to cognitive domains of visuospatial attention and planning and motor domain of speed and dexterity. This finding suggests that there is a less prominent, but still significant, contribution by the dorsal executive parieto-frontal network on unique aspects of driving performance whose hubs are also in posterior brain areas [28].

The data obtained, combining the specific driving simulator, motor, and neuropsychological variables into factors, also provide the opportunity to examine whether any of their combinations can help diagnose Parkinson’s disease, while using expert diagnosis following clinical criteria as the gold standard. Specifically, using LDA, and after cross-validation, best results were achieved through driving simulator factors, and similar accuracy was observed when combining motor and cognitive factors, where approximately three out of four people were correctly classified (Table 9). Instead, weaker prediction was observed when combining all factors, probably a result of overfitting as the number of variables in the model increases disproportionally to the number of observations. These results can prove useful for planning public health and safety policies. First, at face value, older people who participate in simulator driving assessments, such as in re-certification driving examinations, could be given a likelihood of having PD and whether they would benefit from being evaluated by an expert. Similarly, in well-visits that include motor and cognitive performance, non-expert health professionals can pursue motor and cognitive tasks that specifically target the aforementioned domains towards identifying people at risk for having PD. Even more, as the next frontier, driving-related technological advancements, such as built-in car sensors, on-road traffic cameras, and GPS systems, can provide information on strategic and tactical driving parameters, which in turn could serve in identifying people who are at risk for PD and guide them for formal assessments.

Inversely, if a person has PD, it is useful to identify who is at higher risk of being in an accident using objective motor or cognitive metrics obtained in clinical settings [27, 34]. Our LDA analyses of combining motor and cognitive factors (Table 10), revealed that the commonly used UPDRS and H&Y PD motor scales provided fair, but not as useful, predictive accuracy, especially when compared to cognitive testing. This is in keeping with previous studies on the utility of motor scales such as UPDRS [35-38] and cognitive testing [12, 14, 18, 22, 39-41]. Otherwise, the fasterto-extract motor factors from our analyses yielded mildly worse specificity levels to the UPDRS and H&Y scales, pointing to a potentially more time-efficient motor task combination that can be used in non-expert settings, although its utility is to be established given its comparatively low sensitivity. Finally, the combination of motor and cognitive factors led to weaker predictive accuracy, possibly relating to overfitting when the number of variables is disproportionate to the number of participants, as well as motor factors alone not adding significantly useful information to cognitive metrics.

Combining the above information, our findings indicate that quantification of cognitive abilities, especially those mediated by frontal, frontal-subcortical, and parieto-frontal executive networks, as well as driving simulator metrics, especially of tactical driving praxis and car spatial road positioning, can be useful in early diagnosis of PD as well as in predicting accidents in people with PD. Additionally, motor factors of speed and dexterity, as well as axial movement, can also prove useful on their own, especially in predicting tactical driving parameters and, indirectly, accident probability. With the advent of self-administered cognitive tests, the cost of performing these assessments in quantifying a person’s cognitive performance can be further decreased, and all the above assessments can be performed in non-expert settings at a large scale.

Such an infrastructure can also help in planning future studies in lager samples sizes that can help validate their utility in real-world settings and policy planning. Our current study is, thus, limited by its relatively smaller sample size, despite the deep phenotyping of our participants. Nonetheless, our driving and clinical inclusion criteria were very strict, in order to resemble participants who are current drivers and at the early stages of their disease. The above indicate that driving behaviors in our cohort were mostly affected by non-motor symptoms of the disease, which do not show large variations, depending on the sample size, compared to the variations that would be present with motor symptoms. Another, parameter that needs to be recognized is that the driving measures were obtained from a simulator in rural road and not in urban or in highway setting. Although a simulator permits very accurate and reproducible data, examination is performed in a virtual environment and not in real conditions as an on-road driving evaluation. However, driving in a simulator setting is still considered a valid method for examining driving behavior and provides the opportunity to evaluate participants under the exact same conditions, as well as to measure critical driving indices, which is not feasible under on-road driving conditions [42, 43]. Nonetheless, future studies could increase our insight and strengthen our findings using larger sample sizes under on-road driving conditions or analyzing real word data from the safety systems that nowadays modern vehicles have. Τhis approach could also be applied to other degenerative diseases that have motor and cognitive symptoms such as Alzheimer’s Disease, Multiple Sclerosis, and Stroke.

In conclusion, the findings of this study indicate that driving ability in PD is related to an interrelation between cognition, especially executive and visuospatial abilities, and motor performance. Driving simulator factors alone or in combination with factors of cognition and motor performance can be used for distinguishing PD form healthy individuals. From our results it is evident that cognitive testing alone, or in combination with easy to administer motor tasks, can be used by physicians as an objective tool for predicting driving behavior, including accident probability, in order to suggest who may need to stop driving in this specific population.

Acknowledgments

This paper is based on the research project DriverBrain – Performance of drivers with cerebral diseases at unexpected incidents, implemented within the framework of the Action «ARISTEIA» of the Operational Program ‘‘Education and Lifelong Learning” of the Greek General Secretariat for Research and Technology, and is co-financed by the European Social Fund (ESF) and the Greek State. This work was also supported in part by the MES-CoBraD Project H2020-EU.3.1.5 (ID: 965422) grant to the Neurological Institute of Athens.

References

1. Martin AJ, Marottoli R, O’neill D (2013) Driving assessment for maintaining mobility and safety in drivers with dementia. Cochrane database of systematic reviews 2013: CD006222.
2. Uc EY, Rizzo M, Johnson AM, Emerson JL, Liu D, et al. (2011) Reallife driving outcomes in Parkinson disease. Neurology 76:1894-1902.
3. Hopkins RW, Kilik L, Day DJ, Rows C, Tseng H (2004) Driving and dementia in Ontario: a quantitative assessment of the problem. The Canadian Journal of Psychiatry. 49:434-438.
4. Crizzle AM, Classen S, Uc EY (2012) Parkinson disease and driving: an evidence-based review. Neurology 79:2067-2074.
5. De Raedt R, Ponjaert-Kristoffersen I (2001) Predicting at-fault car accidents of older drivers. Accident Analysis & Prevention 33:809-819.
6. Klimkeit EI, Bradshaw JL, Charlton J, Stolwyk R, Georgiou-Karistianis N (2009) Driving ability in Parkinson’s disease: Current status of research. Neuroscience & Biobehavioral Reviews. 33:223-231.
7. Yannis G, Golias J, Antoniou C, Papadimitriou E, Vardaki S, et al. (2013) Design of a large driving simulator experiment on performance of drivers with cerebral diseases.
8. Hair Jr J, Anderson R, Tatham R, Black W (1998) Multivariate Data Analysis., 5th edn.(Prentice Hall: Upper Saddle River, NJ.). 6: 12.
9. Cordell R, Lee HC, Granger A, Vieira B, Lee AH (2008) Driving assessment in Parkinson’s disease--a novel predictor of performance? Movement disorders: official journal of the Movement Disorder Society 23:1217-1222.
10. Heikkila VM, Turkka J, Korpelainen J, Kallanranta T, Summala H (1998) Decreased driving ability in people with Parkinson’s disease. Journal of neurology, neurosurgery, and psychiatry. 64:325-330.
11. Madeley P, Hulley J, Wildgust H, Mindham R (1990) Parkinson’s disease and driving ability. Journal of Neurology, Neurosurgery & Psychiatry. 53:580-582.
12. Stolwyk RJ, Charlton JL, Triggs TJ, Iansek R, Bradshaw JL (2006) Neuropsychological function and driving ability in people with Parkinson’s disease. Journal of clinical and experimental neuropsychology 28:898-913.
13. Uc EY, Rizzo M, Johnson AM, Dastrup E, Anderson SW, et al. (2009) Road safety in drivers with Parkinson disease. Neurology 73:21122119.
14. Wood JM, Worringham C, Kerr G, Mallon K, Silburn P (2005) Quantitative assessment of driving performance in Parkinson’s disease. Journal of Neurology, Neurosurgery & Psychiatry. 76:176-180.
15. Horberry T, Hartley L, Gobetti K, Walker F, Johnson B, et al. (2004) Speed choice by drivers: The issue of driving too slowly. Ergonomics. 47:1561-1570.
16. Grace J, Amick MM, D’Abreu A, Festa EK, Heindel WC, et al. (2005) Neuropsychological deficits associated with driving performance in Parkinson’s and Alzheimer’s disease. Journal of the International Neuropsychological Society. 11:766-775.
17. Beratis IN, Andronas N, Fragkiadaki S, Kontaxopoulou D, Pavlou D, et al. (2018) Exploring the association of the Comprehensive Trail Making Test with driving indexes in patients with Parkinson’s disease. Transportation Research Part F: Traffic Psychology and Behaviour, 59:535-544.
18. Ranchet M, Paire-Ficout L, Uc EY, Bonnard A, Sornette D, et al. (2013) Impact of specific executive functions on driving performance in people with Parkinson’s disease. Movement disorders: official journal of the Movement Disorder Society, 28:1941-1948.
19. Uc EY, Rizzo M, Anderson SW, Shi Q, Dawson JD (2005) Driver landmark and traffic sign identification in early Alzheimer’s disease. Journal of neurology, neurosurgery, and psychiatry, 76:764-768.
20. Uc EY, Rizzo M, Anderson SW, Sparks JD, Rodnitzky RL, et al. (2006) Driving with distraction in Parkinson disease. Neurology, 67:17741780.
21. Adler G, Rottunda S, Bauer M, Kuskowski M (2000) The older driver with Parkinson’s disease. Journal of Gerontological Social Work, 34:39-49.
22. Lee H, Falkmer T, Chee D (2011) Performance of drivers with parkinson’s disease under the effect of cognitive overloading: insinuation for assessment and training. Accid Anal Prevent.
23. Uc EY, Rizzo M, Anderson SW, Sparks JD, Rodnitzky RL, et al. (2007) Impaired navigation in drivers with Parkinson’s disease. Brain, 130:2433-2440.
24. Ranchet M, Paire-Ficout L, Marin-Lamellet C, Laurent B, Broussolle E (2011) Impaired updating ability in drivers with Parkinson’s disease. Journal of neurology, neurosurgery, and psychiatry, 82:218-223.
25. Holmes JD, Alvarez L, Johnson AM, Robinson AE, Gilhuly K, et al. (2019) Driving with Parkinson’s Disease: Exploring lived experience. Parkinson’s Disease.
26. Rizzo M, Uc EY, Dawson J, Anderson S, Rodnitzky R (2010) Driving difficulties in Parkinson’s disease. Movement disorders : official journal of the Movement Disorder Society, 25:S136-140.
27. Dubinsky RM, Gray C, Husted D, Busenbark K, Vetere-Overfield B, et al. (1991) Driving in Parkinson’s disease. Neurology, 41:517-520.
28. Zgaljardic DJ, Borod JC, Foldi NS, Mattis P (2003) A review of the cognitive and behavioral sequelae of Parkinson’s disease: relationship to frontostriatal circuitry. Cognitive and behavioral neurology,16:193-210.
29. Stolwyk RJ, Scally KA, Charlton JL, Bradshaw JL, Iansek R, et al. (2015) Self-Regulation of Driving Behavior in People with Parkinson Disease. Cogn Behav Neurol, 28:80-91.
30. Hong Z, Shi M, Chung KA, Quinn JF, Peskind ER, et al. (2010) DJ-1 and alpha-synuclein in human cerebrospinal fluid as biomarkers of Parkinson’s disease. Brain, 133:713-726.
31. Uc E, Rizzo M, Anderson S, Qian S, Rodnitzky R, et al. (2005) Visual dysfunction in Parkinson disease without dementia. Neurology, 65:1907-1913.
32. Uc EY, Rizzo M, Anderson SW, Sparks J, Rodnitzky RL, et al. (2006) Impaired visual search in drivers with Parkinson’s disease. Annals of neurology, 60:407-413.
33. Wood JM, Owsley C (2014) Useful field of view test. Gerontology, 60:315-318.
34. McLay P (1989) The Parkinsonian and driving. International disability studies, 11:50-51.
35. Devos H, Vandenberghe W, Nieuwboer A, Tant M, Baten G, et al. (2007) Predictors of fitness to drive in people with Parkinson disease. Neurology, 69:1434-1441.
36. Radford K, Lincoln N, Lennox G (2004) The effects of cognitive abilities on driving in people with Parkinson’s disease. Disability and rehabilitation, 26:65-70.
37. Singh R, Pentland B, Hunter J, Provan F (2007) Parkinson’s disease and driving ability. Journal of Neurology, Neurosurgery & Psychiatry, 78:363-366.
38. Zesiewicz TA, Cimino CR, Malek AR, Gardner N, Leaverton PL, et al. (2002) Driving safety in Parkinson’s disease. Neurology, 59:17871788.
39. Amick MM, Grace J, Ott BR (2007) Visual and cognitive predictors of driving safety in Parkinson’s disease patients. Arch Clin Neuropsychol, 22:957-967.
40. Stolwyk RJ, Triggs TJ, Charlton JL, Moss S, Iansek R, et al. (2006) Effect of a concurrent task on driving performance in people with Parkinson’s disease. Movement disorders: official journal of the Movement Disorder Society, 21:2096-2100.
41. Worringham CJ, Wood JM, Kerr GK, Silburn PA (2006) Predictors of driving assessment outcome in Parkinson’s disease. Movement disorders : official journal of the Movement Disorder Society, 21:230-235.
42. Fisher DL, Caird JK, Rizzo M, Lee JD (2011) Handbook of driving simulation for engineering, medicine and psychology: an overview. Handbook of driving simulation for engineering, medicine, and psychology.
43. Yannis G, Papantoniou P, Nikas M (2015) Comparative analysis of young drivers behavior in normal and simulation conditions at a rural road.  Proceedings of the 5th International Conference on Road Safety and Simulation, Universities of Central Florida and Tennessee, Orlando, Florida.