|Home | About | Journals | Submit | Contact Us | Français|
The growing literature on driving in Parkinson disease (PD) has shown that driving is impaired in PD compared to healthy comparison drivers. PD is a complex neurodegenerative disorder leading to motor, cognitive, and visual impairments, all of which can affect fitness to drive. In this review, we examined studies of driving performance (on-road tests and simulators) in PD for outcome measures and their predictors. We searched through various databases and found 25 (of 99) primary studies, all published in English. Using the American Academy of Neurology criteria, a study class of evidence was assigned (I–IV, I indicating the highest level of evidence) and recommendations were made (Level A: predictive or not; B: probably predictive or not; C: possibly predictive or not; U: no recommendations). From available Class II and III studies, we identified various cognitive, visual, and motor measures that met different levels of evidence (usually Level B or C) with respect to predicting on-road and simulated driving performance. Class I studies reporting Level A recommendations for definitive predictors of driving performance in drivers with PD are needed by policy makers and clinicians to develop evidence-based guidelines.
In addition to the typical motor symptoms of Parkinson disease (PD), persons with PD may develop cognitive impairment/dementia, emotional impairments (e.g., apathy and disinhibition), and visual-perceptual deficits that often do not respond to dopaminergic medications.1,2 Together with the variability in response to the timing of dosage (e.g., on-off phenomenon) and possible side effects (e.g., daytime sleepiness) of PD medications, these conditions can impair driving ability and potentially lead to elevated crash risk.3,4
The ability to drive safely may be impaired even in the early stages of PD.5-17 Prior studies show that associations exist between impaired driving performance and deficits of contrast sensitivity,13,16 visual processing,5,10,16,17 set-shifting,8,10,11,13 and psychomotor speed.5,11-13,16 Epidemiologic data, however, are not well-established concerning crash rates in PD.18-20 A retrospective survey study found that patients with Hoehn & Yahr (H&Y) stages 2 and 3 had a significantly higher crash risk compared to healthy controls. However, there was no evidence of increased crash risk among patients in H&Y stage 1.20 Another survey study found that 82% of patients with PD held a driving license and 60% of them were still driving. Of the patients holding a driving license, 15% reported being involved in an accident, with 11% being at fault in the past 5 years.21 However, whether crash rates are higher in PD is unclear as no age- and gender-matched controls were included for comparison purposes.21 Recently, a prospective cohort study compared drivers with PD to drivers without neurologic conditions and found no clear link between PD and occurrence of real-life crashes.22
To date, there are no evidence-based practice parameters to guide physicians in determining driving fitness in PD. The American Medical Association recommends physicians base their decisions surrounding driving on both motor and cognitive impairments, response to dopaminergic medications, and side effects of medications.23 Guidelines developed by the National Highway Traffic Safety Administration18 and Federal Motor Carrier Safety Administration24 suggest a case by case, multidisciplinary evaluation of the patient due to the highly individualized nature of the disease and variable progression. In the absence of clear guidelines, clinicians can only make subjective decisions on fitness to drive.25,26 This may be problematic as physicians/neurologists often overestimate the driving ability of their patients with PD.25-27 Additionally, drivers with PD may not reveal medical information when renewing their license or adhere to physician's advice to quit driving.27
Thus, it is critical to have guidelines for clinical decision making on driving ability in persons with PD. However, determining the level of evidence in prior PD and driving studies has proven difficult, primarily from the inconsistent results and various methodologic approaches used. In a recent review article, reasons for discrepant findings are related to 1) differences in sample sizes; 2) use of various rating scales; 3) heterogeneity of symptoms; 4) not stratifying by disease severity; and 5) varying driving performance measures.26 The purpose of this current review is to discern the levels of evidence for reported predictors of driving fitness, using on-road and simulator studies, and to provide recommendations for driving performance.
We searched and analyzed the results of primary studies addressing PD and driving. Primary studies included only empirical and original peer-reviewed published manuscripts.28 Specifically, we focused on the evidence used to determine driving ability and performance in persons with PD. We conducted the review using the following search criteria: a literature search, determining the inclusion/exclusion criteria, and ratings of the evidence and recommendations. We searched databases representing medicine, health science, psychological, and social science (e.g., PubMed, CINAHL, and Web of Science and Google Scholar) for key words, search terms, and MeSH headings including Parkinson's disease, driving, automobile, driving performance, driving ability, simulator, simulated, and road test. We also identified articles via footnote chasing (secondary sourcing). We excluded studies if they were 1) published prior to 1995 (due to the scant number of studies published before then); 2) duplicates; 3) not primary studies; 4) qualitative or descriptive; 5) of psychometric designs (e.g., test-retest, rater-reliability, or validity of measures); 6) not including driving as a primary outcome variable; 7) samples of a mixed composition (e.g., PD and Alzheimer disease); and 8) based on survey design. We included primary peer-reviewed articles published between 1995 and 2011 that reported empirical findings on driving performance and PD. The search yielded 57 citations with abstracts. After review, we excluded 32 as 9 were not primary studies; 11 did not use driving as an outcome measure; 3 were based on expert opinion; 3 were conducted prior to 1995; 2 were not in English; 3 were descriptive; and 1 was based on psychometric design. The remaining 25 studies met all criteria for inclusion and were critically appraised, classified, and synthesized.
We used the classification criteria of the American Academy of Neurology to assign the level of evidence for studies examining drivers with PD.29 These same guidelines were used to provide recommendations after studies were appropriately classified. As shown in table 1, we used the following parameters for rating an article by class (I–IV, with Class I being the highest level of evidence) and for making recommendations: Levels A–C (A being predictive or not of the outcome, B being probably predictive or not of the outcome, C being possibly predictive or not of the outcome) and Level U if inadequate data or conflicting findings existed. Classifications of studies and recommendations made were consistent among the 3 reviewers, with 92% agreement. We used team consensus to classify ratings for any studies that initially differed between raters.
The 25 studies were published between 1998 and 2011. Thirteen studies reported federal or foundational funding. There were 16 on-road studies with PD sample sizes ranging from 19 to 154. Study design for on-road studies was primarily prospective (n = 15), with 1 retrospective study. Of the 9 simulator studies, 8 were experimental and 1 quasi-experimental. PD sample sizes ranged from 6 to 67. Disease severity of PD samples was considered to be mild to moderate across all studies. All studies included participants having a confirmed diagnosis of PD by neurologists or movement disorders specialists, although only 3 studies reported using the UK Brain Bank as a criterion for PD diagnosis.5,16,29 Additionally, most studies evaluated participants for cognitive function. However, only 5 studies excluded participants based on cognitive screening. Four studies excluded those who scored ≤24 on the Mini-Mental State Examination (MMSE),8,9,15,16 one study excluded those who scored <267 on the MMSE, and one other study excluded those who scored ≤1 on the Clinical Dementia Rating Scale (CDR).30
A summary of the 25 primary studies included in this review, containing the title, authors, year and funding, primary objective, sample characteristics, independent and outcome variables, design, key findings, level of evidence, and conclusions, is shown in table e-1 on the Neurology® Web site at www.neurology.org. Based on the criteria outlined in table 1, results, conclusions, and recommendations concerning predictors of driving performance (on-road and simulator) are discussed next.
The review yielded 12 Class II studies5,7-13,15-17,25,30 and 4 Class III studies.31-34 Only 2/16 on-road studies utilized a certified driving rehabilitation specialist (CDRS) to conduct the on-road tests.
From the 12 Class II studies, 11 studies found that drivers with PD had significantly worse driving performance (p < 0.05) than healthy controls.5,7,8,10-13,15-17,25 No definitive conclusions could be made for one study as a control group was not included for on-road driving.30 Although most drivers with PD were considered safe to drive, deficits that may affect driving were apparent even in the early stages of PD. There were 6 studies that found a clinical battery predicted driving performance5,10,12,13,17,30 while one study did not.9 Variables that predicted driving performance also differed across studies, as shown in table e-1. As findings were mixed concerning what clinical tests predicted driving performance, we examined and derived conclusions for predicting driving performance based on individual tests employed in studies (e.g., Useful Field of View [UFOV], MMSE).
We concluded the Unified Parkinson’s Disease Rating Scale (UPDRS) motor scores obtained while patients are on medication to be probably not predictive of driving performance based on 10 Class II5,7,10-13,15-17 and 3 Class III studies.31,32,34 However, based on 2 Class II studies,5,17 UPDRS motor scores during the practically defined “off” period were concluded to be probably predictive of driving performance. We also concluded the H&Y stages to be probably not predictive of driving performance based on 4 Class II studies5,15,16,30 and 2 Class III studies.33,34 We concluded no recommendations could be made for the use of the Webster's rating scale. We considered disease duration to be probably not predictive of driving performance based on 6 Class II studies12,13,15-17,28 and 1 Class III study.31
Six Class II studies5,11-13,16,17 and 1 Class III study31 assessed contrast sensitivity. Five studies used the Pelli-Robson chart (4 Class II11-13,16 and 1 Class III31) and 2 Class II studies used the Optec 25005,17 to assess contrast sensitivity (scored as either acceptable or impaired). We concluded that contrast sensitivity was probably predictive of driving performance. Visual acuity was assessed in 6 Class II studies5,10-13,16 and 1 Class III study.31 While most studies found that visual acuity was not predictive of driving performance, 1 Class II study found that far visual acuity was predictive of at-fault errors on the road.13
Seven studies included the UFOV (6 Class II and 1 Class III). From the 6 Class II studies,5,10,12,13,16,17 we concluded the UFOV was probably predictive of impaired driving performance. Two Class II studies found the UFOV subtest 2 was probably predictive,5,17 and 1 Class II17 and 1 Class III study32 found that the UFOV subtest 3 was possibly predictive of driving performance. Two Class II studies5,10 found that the UFOV Global Risk Index of 3 (moderate crash risk) was probably predictive of differentiating between safe and unsafe drivers with PD. From 4 Class II studies,10-13 we concluded that cumulative scores based on the 4 UFOV subtests were probably predictive of driving performance. Only one study had determined UFOV cutpoints,17 as shown in table e-1.
We concluded the Rey-Osterrieth Complex Figure Test (ROCF) to be probably predictive of driving performance based on 6 Class II studies8,10-13,30 and 1 Class III study.31 The Trail Making Test Part B (Trails B) was used in 2 Class II8,17 and 2 Class III studies.31-33 We concluded that the Trails B is probably predictive of driving performance. We concluded the Trails A, used in 1 Class II study8 and 2 Class III studies,31-33 to be possibly predictive of driving performance. Four Class II studies10-13 used Trails B–A (defined as time on Trails A subtracted from time on Trails B) and were found to be strongly associated with at-fault driving errors. We concluded that Trails B–A is probably predictive of driving performance. The MMSE, used in 6 Class II studies5,7,12,13,17,25 and 2 Class III studies,33,34 is probably not a useful predictor of driving performance in drivers with PD. We considered the CDR, Hopkins Verbal Learning Test (HVLT), Judgment of Line Orientation (JOLO), Benton Visual Retention Task (BVRT), and the Auditory Verbal Learning Test (AVLT) possibly predictive based on the results of 1 Class II study or 2 Class III studies.
We concluded that functional reach tests are probably predictive of driving performance based on 2 Class II studies.11,13 Finger tapping, Pegboard test, Rapid Paced Walk test, and Timed Get Up and Go test are possibly predictive of driving performance determined by 1 Class II study and 2 Class III studies.
Level B: The UFOV subtest 2, Risk Index, cumulative UFOV scores (based on 4 subtests), contrast sensitivity, Trails B and B–A, ROCF, functional reach, and UPDRS “off” motor scores are probably predictive of driving performance. Conversely, UPDRS motor scores, H&Y, disease duration, Geriatric Depression Scale, and MMSE scores are probably not predictive of driving performance. Level C: UPDRS-activities of daily living (ADL) scores, Trails A, CDR, HVLT, JOLO, Wechsler Intelligence Test, BVRT, AVLT, Finger tapping, Rapid Paced Walk test, Timed Get up and Go test, and Pegboard test are possibly predictive of driving performance. Epworth Sleepiness Scale scores are possibly not predictive of driving performance. Level U: Age,13,30,34 reaction time,25,33 and Webster's Rating Scale32 cannot be interpreted due to inadequate (tests used in 1 study32) or conflicting data.
From the 9 Class III studies, we concluded that cognitive abilities may deteriorate even in mild to moderate stages of PD. Declining cognitive function was related to impaired simulated driving performance,6,9,37,38,40 particularly in low visibility conditions (assessed using the Pelli-Robson chart).14 Disease severity (H&Y) was associated with worse simulated driving performance.35 The use of external cues may help improve skills related to driving.39
Level C: MMSE scores and contrast sensitivity (assessed via the Pelli-Robson Chart) are possibly predictive of simulated driving performance in the mild to moderate stages of PD. Level U: We could not make any recommendations with respect to age, UPDRS motor and ADL scores, H&Y, Schwab-ADL scores, Trails A and B, UFOV, ROCF, JOLO, Symbol Digit Modalities Test (SDMT), Finger tapping, or reaction time due to inadequate data (tests only used in 1 study).
Based on the findings of this review, certain risk factors may be more heavily weighted than others when determining fitness to drive. Modified from a prior diagram,41 the figure outlines a hierarchy of the primary findings of this review for both on-road and simulator studies. Increasing disease severity, in addition to risk factors (as concluded by this review), may provide a framework to help clinicians determine when drivers with PD are at risk. However, in the absence of a meta-analysis with pooled effect sizes to make clinical inferences, this information should be considered as supplemental to clinician's judgment.
While the figure provides a general baseline of risk factors concerning fitness to drive, when a person may reach a high level of risk is unclear. Besides UFOV risk index score, which needs to be replicated in larger studies, we cannot suggest any cutoffs for the remaining risk factors that would indicate a high-risk driver or a patient with PD who should discontinue driving. Future studies are needed to determine cutpoints of risk factors with on-road driving performance (e.g., sensitivity, specificity, negative predictive value, and positive predictive value) and prospective crash risk. In the absence of definitive cutpoints to help clinicians determine fitness to drive, we recommend patients in question be referred for a multidisciplinary evaluation (e.g., a team consisting of neurologist, neuropsychologist, CDRS), which includes a comprehensive driving evaluation. If the patient is unwilling or unable to take the evaluation by a CDRS (e.g., no insurance coverage for CDRS evaluation), a referral can be made to the state Department of Motor Vehicles for a road driving test.42
We found no standard clinical battery to predict driving performance in drivers with PD. Clinical tests and in some instances neuropsychological tests differ from study to study and are administered on small PD samples. Moreover, studies have differed in their methods and evaluation of driving performance (e.g., CDRS, driving evaluator, driving instructor, research assistant). Some studies used instrumented vehicles, while others did not. Many studies also did not provide pass/fail outcomes on the on-road tests leading to difficulties interpreting what is truly considered “impaired driving performance.”
From the on-road studies, we concluded that Level B evidence exists for the UFOV (subtest 2, risk index and cumulative across the 4 subtests), contrast sensitivity, Trails B and B–A, ROCF, Functional Reach, and UPDRS “off” motor scores for probably predicting driving performance. However, UPDRS “off” motor scores may not be a clinically useful predictor of driving performance as driving is usually assessed during the “on” state. Still, the predictive ability of these “off” period scores, as a measure of general motor severity of PD, rather than a snapshot of motor function in a compensated state, indicates the important influence of severity of parkinsonism on driving. A longitudinal study showing that UPDRS-ADL scores and the daily antiparkinsonian medication predicted future driving cessation supported this view.22 However, given that studies have shown mixed results concerning impaired driving performance in drivers with PD, we recommend the need for Class I studies with Level A recommendations.
As visual attention, spatial, and executive skills are critical abilities for driving in general, it is not surprising that UPDRS “on” motor scores, H&Y, and disease duration are probably not predictive of driving performance. Disease severity indices mainly capture motor symptoms and do not capture visual or cognitive deficits. Additionally, deficits of visual and cognitive abilities may occur independently of motor symptoms.43 Moreover, performance on basic visual sensory tests may be associated with performance on more complex visual cognitive tests.2 Therefore, multivariate modeling of predictors (rather than just bivariate correlations) is critical to assess independent predictive contributions of different interrelated measures.
Of the 9 simulator studies examined in this review (all Class III), we found that cognitive deficits may be present in the early stages of PD, consistent with reported findings that suggest 20% have mild cognitive impairment at time of PD diagnosis.44 We found simulated driving performance in drivers with PD to be worse in low contrast conditions, which may possibly be predicted by deficits of contrast sensitivity. We also determined the MMSE to be possibly predictive of simulated driving performance. However, caution is warranted as we could only make Level C recommendations. Additionally, the MMSE has been shown to be a poor predictor of on-road performance in drivers with PD.5,15,25,45
Differences between various simulator studies in technical characteristics (e.g., desktop simulator with 60-degree field of view35 vs medium-high fidelity simulator with 180-degree field of view9,37,38) and the primary outcomes, as well as a lack of validity due to the various types of simulators used in prior studies, are also a concern. However, simulators can be used to determine predictors of driving performance that cannot be tested on the road test due to ethical, safety, and practical reasons (e.g., night, high volume traffic, poor weather conditions, or hazardous experiments where rapid reactions are needed). For example, in a simulated collision avoidance experiment, motor measures (UPDRS motor scores, tapping speed) were associated with time to first reaction to the hazard in a simulated collision avoidance.16 Predictors identified on the simulator can be further tested on the road although follow-up studies are needed to discern the concurrent validity between on-road tests and simulators. We could not make a recommendation for age, UPDRS motor and ADL scores, H&Y, Schwab ADL scores, Trails A and B, UFOV, ROCF, JOLO, SDMT, Finger tapping, and reaction time.
In addition to psychometric and motor measures, factors such as driving records, exposure, and habits (e.g., use of compensation strategies), and input from caregivers7 can be useful in determining driving outcomes in PD, especially driving cessation.22 However, these measures are largely based on self-report and subject to associated methodologic problems. As an alternative, longitudinal “naturalistic” driving studies have the potential to improve prediction of driving outcomes by providing continuous, objective, and quantitative measures of driving in the patient's own vehicle and environment over a long time period while performing real-life tasks.46
The study limitations include making recommendations despite the small, heterogeneous, and male- dominated samples among the primary studies. We only searched and included studies published in English and those within the last 13 years. Although we searched for secondary sources, we did not pursue government publications, unpublished manuscripts, or disseratations.28 However, to our knowledge, this is the first evidence-based review on driving on-road and simulated performance in PD. The strengths of this study include a team consensus process for study selection and expertise in evidence-based approaches, driving research, or PD.
Class I studies are needed to develop more effective screening tools for clinicians to identify at-risk drivers with PD. Class I studies with large and representative samples, standardized evaluation batteries, and driving protocols are needed to provide Level A recommendations to ultimately help policy makers and clinicians develop evidence-based decision guidelines.
Supplemental data at www.neurology.org
Dr. Crizzle was responsible for the conceptualization and study design, data collection and analysis, interpretation and drafting the original manuscript, and subsequent revisions to the manuscript. Dr. Classen provided contributions to the study design, data analysis, interpretation of study findings, and editing first and all subsequent drafts of the manuscript. Dr. Uc provided expertise on data analysis and interpretation, and critically reviewed, edited, and revised all drafts of the manuscript.
A. Crizzle reports no disclosures. S. Classen receives support from the National Parkinson's Foundation [NPF 00085107 (Project PI)]. E. Uc is receiving research support from Rehabilitation Research and Development Branch, Department of Veterans Affairs as the PI of the Merit Review awards on aerobic exercise in PD (1 I01 RX000170) and driving rehabilitation in PD (B6918). He has received research support from the NIH as the PI of the project “Predicting Driver Safety in PD” (R01 NS044930). Go to Neurology.org for full disclosures.