Parkinson’s disease (PD) is now recognized as a multisystem disorder manifested by motor, cognitive, and other nonmotor symptoms that affect safe driving ability, and, therefore, mobility and independent community living (Heikkila, Turkka, Korpelainen, Kallanranta, & Summala, 1998; Stolwyk, Charlton, Triggs, Iansek, & Bradshaw, 2006; Wood, Worringham, Kerr, Mallon, & Silburn, 2005). Given the dependence on personal automobiles as the primary source of transportation in Western societies, it has become increasingly important to identify safe or, more important, remediable drivers with PD through driver assessment and rehabilitation programs.

PD is not a static disorder. It is usually slowly progressive with a wide and heterogeneous spectrum of symptoms that affect driving to different degrees among patients, thus making it difficult to determine just when someone crosses over from being a safe to an unsafe driver (Radford, Lincoln, & Lennox, 2004). Although it is assumed to be the responsibility of physicians to determine a patient’s “fitness” to drive, the ability of physicians to determine whether a patient is a safe driver, based on their limited observations in a typical clinical encounter, is likely overestimated (Heikkila et al., 1998; Marshall & Gilbert, 1999; Miller & Morley, 1993). The need for efficient screening batteries with good predictive validity of passing or failing an actual on-road course is critical in promoting safe and independent driving or in making driving cessation decisions.

A body of literature concerning determination of PD patients’ fitness to drive is evolving. Specifically, clinical indicators obtained from neurological or neuropsychological testing batteries on the driving ability of PD patients include motor or mental impairments (Amick, Grace, & Ott, 2007; Devos et al., 2007; Dubinsky et al., 1991; Worringham, Wood, Kerr, & Silburn, 2006), sleepiness (Lachenmayer, 2000; Meindorfner et al., 2005), medication use (Lachenmayer, 2000), severity of disability (Madeley, Hulley, Wildgust, & Mindham, 1990; Singh, Pentland, Hunter, & Provan, 2007; Worringham et al., 2006), and clinical disease markers (Grace et al., 2005; Heikkila et al., 1998; Lings & Dupont, 1992; Radford et al., 2004; Singh et al., 2007; Stolwyk, Triggs, et al., 2006; Wood et al., 2005; Zesiewicz et al., 2002). More recently, studies have shown that decreased cognition, motor performance, and visual processing—specifically reductions in Useful Field of View (UFOV) and contrast sensitivity—are associated with poorer driving performance (Amick et al., 2007; Classen et al., 2009; Devos et al., 2007; Heikkila et al., 1998; Radford et al., 2004; Worringham et al., 2006). From driving simulators or on-road evaluations, we are observing increased driving risk in people with PD (Classen et al., 2009; Dubinsky et al., 1991; Zesiewicz et al., 2002). Compared with neurologically healthy drivers, those with PD show greater impairment on route-finding tasks and landmark and traffic sign identification tasks (mild to moderate PD only; Uc et al., 2006b, 2007). Yet, PD patients are not more prone to cause crashes than the rest of the population (Homann et al, 2003).

Predicting driving ability in patients with PD is essential. Four studies have examined the capabilities of clinical tests to predict driving performance in these patients (Devos et al., 2007; Heikkila et al., 1998; Radford et al., 2004; Worringham et al., 2006). These studies differ from the study presented here in three noteworthy aspects. First, although each of the previous studies included an on-road driving evaluation, none used a certified driving rehabilitation specialist (CDRS) as an evaluator. As an occupational therapy practitioner, the CDRS has undergone healthcare training in addition to training as a driving evaluator, which provides him or her with an understanding of functional limitations caused by disease, disability, developmental challenges, aging, or contextual issues. The CDRS is therefore positioned to suggest rehabilitation strategies to prolong safe and independent community mobility, including driving. In addition, none of the previous studies included in their analysis the standard scoring measure of processing speed (milliseconds) for the divided attention subtest of the UFOV (Subtest 2; Edwards et al., 2005). This particular test has shown significant prediction of on-road driving performance among those with PD in pilot work (Classen et al., 2009). Finally, only the categorical pass–fail outcome was examined in prior studies. In addition to the pass–fail outcome, we examined the potential of clinical tests to predict the Sum of Maneuvers Score (SMS), a numerical outcome based on the error rating by the CDRS of each specific driving maneuver during the on-road evaluation.

Despite a seemingly growing interest among medical practitioners and occupational therapists in the field of driving capabilities of PD patients, no model or method for testing has been instituted in a clinical setting. Thus, the objective of this study was to expand the existing field of research by piloting the predictability for on-road driving performance in PD patients using a standard battery of clinical tests administered by a CDRS and neurologists.

This study was approved by the Institutional Review Board of the University of Florida, Gainesville, and all participants provided written informed consent.

Forty-one participants (age 65–85 yr) diagnosed with idiopathic PD by a movement disorders specialist using defined criteria (Hughes, Daniel, Blankson, & Lees, 1993) were recruited from the University of Florida’s Movement Disorder Center (UFMDC) and referred to a CDRS to determine safe driving ability, regardless of the clinician’s perspective on the participant’s driving competence (Table 1). Each of the PD participants was seen at the UFMDC and was offered participation in this study, regardless of his or her perceived ability to drive, as long as he or she met all criteria for the study.

In addition, 41 healthy community-dwelling older adults (age 65–85 yr) who have previously (2005–2009) participated in the Institute for Mobility, Activity, and Participation’s (I–MAP) on-road studies were selected as control participants. They were age-matched to the group with PD. These control participants were previously (2004–2009) recruited from north central Florida by means of paid advertisements in newspapers; flyers distributed to aging service centers, health clubs, apartment complexes, and community centers; open houses held at the University of Florida’s Gator-Tech Smart House; and word of mouth. Inclusion criteria for both PD and non-PD participants were age ≥65 yr, with a valid driver’s license that met the requirements of the State of Florida statute for licensing, including having satisfactory visual acuity and fields and being seizure-free in the previous 6 mo. No participant had acute confounding medical or psychiatric conditions. Although we age-matched the control participants (mean [M] = 73.0, standard deviation [SD] = 5.2) to PD (M = 73.1, SD = 6.0), no significant differences existed for race or socioeconomic status. However, we had more men in the PD group (χ² = 7.4, N = 31, df = 1, p = .007). We did not document characteristics of those potential participants who refused to participate in the study.

All 41 participants with PD were evaluated at the UFMDC and were referred to the University of Florida’s I–MAP driving assessment and rehabilitation clinic. Participants were evaluated on one of two standardized road courses in either Gainesville, Florida, or Ocala, Florida (Justiss, Mann, Stav, & Velozo, 2006). With the exclusion of neurological tests, control participants underwent similar clinical and on-road testing procedures administered by the same CDRS in the same settings.

We used a prospective design with a convenience sample of people with PD and an existing cohort of people without PD who were previously (2004–2009) tested by I–MAP researchers using the same testing protocol.

Demographics and medical information were obtained for each participant. Clinical evaluation consisted of neurological (for the PD cohort only) and standard cognitive, vision, and motor tests commonly used in a clinical battery preceding on-road driving assessment. An UFMDC neurologist performed the neurological testing. The CDRS administered the remaining exams as part of the comprehensive driving evaluation.

Participants were evaluated with the Unified Parkinson’s Disease Rating Scale Part 3 (motor subscale; UPDRS–Part 3; Fahn & Elton, 1987) in the on medication state (i.e., 1 hr after medication intake) and off medication state (i.e., at least 12 hr since the last PD medication) by UFMDC neurologists before driving evaluation. Participants were also evaluated with the modified Hoehn and Yahr stage, a numerical ranking that indicates the stage of PD as follows: Stage 0 = no signs of disease; Stage 1 = unilateral disease; Stage 1.5 = unilateral plus axial involvement; Stage 2 = bilateral disease, without impairment of balance; Stage 2.5 = mild bilateral disease with recovery on pull test; Stage 3 = mild to moderate bilateral disease, some postural instability, physically independent; Stage 4 = severe disability, still able to walk or stand unassisted; and Stage 5 = wheelchair bound or bedridden unless aided (Goetz et al., 2004).

On a different day, after the visit to the UFMDC, the CDRS administered the comprehensive driving evaluation to participants with PD who were on medication or in the subjective “on” state. Participants were tested only when they considered themselves to be at the peak of their capabilities, in the subjective “on” state. If the participants felt that they were for some reason unable to perform at their optimum capabilities, the CDRS postponed the driving evaluation to a later date.

This comprehensive driving evaluation consists of the cognitive, vision, and motor testing followed by an on-road evaluation (Stav, Justiss, McCarthy, Mann, & Lanford, 2008).

We used the Mini-Mental State Examination (MMSE, total score 30, ≤ 26 indicates cognitive decline; Folstein, Folstein, & McHugh, 1975), because of missing data for other cognitive tests.

Visual sensory testing included far visual acuity, peripheral fields, and contrast sensitivity (Slide A = neurological region; Slide B and C = optic nerve/retina region; Slide D = optic nerve/retina/macular region; Slide E = macular region) using the Optec^® 2500 machine (Stereo Optical, Inc., Chicago). Acuity and peripheral fields were reported as impaired or within functional limits on the basis of the legal criteria to drive in the State of Florida. Contrast sensitivity was reported as impaired or within functional limits for each slide (A–E) and as an overall ranking.

Visual perception and attention were measured with the UFOV (Edwards et al., 2005), which consists of three subtests: 1 = attention; 2 = divided attention; 3 = selective attention, measured in milliseconds, with a cutoff of 500 ms for each one of the subtests. The UFOV Subtest 4 was based on a risk index of overall performance on individual tests and scaled from 1 (low risk) to 5 (high risk; UFOV^® User’s Guide 6.0.6; Visual Awareness, Inc., Birmingham, AL) of motor vehicle crashes. Risk indexes ≥3 were reported as unsafe.

General motor function included lower limb mobility, coordination, and trunk stability while a person walks a 20-ft distance and was measured with the Rapid Pace Walk (RPW; Wang, Kosinski, Schwartzberg, & Shanklin, 2003). The RPW is a measure of postural instability and gait difficulty (PIGD). Taking >7.5 s to complete the walk has been associated, in studies of older adults, with a greater likelihood of at-fault crashes (Staplin, Gish, & Wagner, 2003).

Participants drove a road course of hierarchical complexity in Gainesville, Florida, or Ocala, Florida, in residential, suburban, and highway areas for 1 hr in optimal weather conditions and not during peak hours. We used a Buick Century 2004 test vehicle equipped with an auxiliary brake. This assessment has demonstrated reliability and validity among older drivers and is described in detail elsewhere (Justiss et al., 2006). The two main on-road outcomes, determined by the CDRS, were the global rating score (GRS) of pass–fail and the Sum of Maneuvers Score (SMS), a weighted error score based on the level of assistance needed to safely execute 91 on-road maneuvers (e.g., lane maintenance, speed regulation, gap acceptance, signaling, vehicle positioning, adjustment to stimuli, and yielding). The maximum score of 273 points indicates perfect driving. Because the total number of maneuvers differed between the two courses (Ocala = 84 and Gainesville = 91), both sets of scores were normalized to 91 total maneuvers and a weighted error score was calculated with a common denominator of 273.

All data were entered into a SPSS Version 17.0 (SPSS, Inc., Chicago) file and securely stored in a password-protected server. Quality control was performed to check the data for accuracy, completeness, and reduction. We did not impute missing data, and we excluded the tests of participants with >20% missing data.

Using SPSS, we performed univariate, bivariate correlational (Spearman rank coefficient, or r), χ², and independent sample t-tests analyses. The mean and standard deviation were provided for continuous measures, and frequencies were provided for nominal and ordinal measures. The t statistic or χ² statistic and their corresponding p values were provided to indicate significant group differences. Results for all tests were considered significant if p ≤ .05, two-tailed. Variables from the clinical tests that showed significant correlations with the GRS (pass–fail dichotomous on-road outcome) were analyzed with logistic regression (forward stepwise manner, likelihood ratio test with an entry significance criteria of p = .05–0.10). Model 1 was run with participants with PD and non-PD participants, and Model 2 was run without non-PD control participants because they had no neurological data. Using a similar method, linear regression was used to predict the SMS (continuous outcome) for Parkinson’s disease only.

Table 1 summarizes descriptive statistics, including demographics, medical information, and clinical test and road test results. The average age groups, racial distribution, and educational level were comparable between groups. The only significant difference in demographics between the two groups was gender, with a higher representation of men in the PD cohort. For medical results, the PD cohort had a significantly higher number of medical conditions. Disease severity was measured according to Hoehn and Yahr stage on medication (n = 37, M = 2.4, SD = 0.4) and showed a normal distribution (Stage 2, n = 17; Stage 2.5, n = 9; Stage 3, n = 11); it was also measured for patients off medication (n = 35, M = 2.4, SD = 0.5). The mean UPDRS III for participants on medication was 27.4 (SD = 7); off medication, M = 32.1 (SD = 9.9) points. For the clinical tests, PD patients performed more poorly on UFOV Subtest 1, UFOV Subtest 2, UFOV Subtest 4, and RPW. Significant differences were observed for both outcomes of the road test; participants with PD performed more poorly on the GRS (56.1% failing vs. 12.2% of control participants failing, χ²[1] =17.6, p < .001) and on the SMS (PD patients, M = 209, SD = 51.9, t[80] = 4.1, p < .001).

Table 2 presents correlations of clinical tests with the GRS and SMS for the PD and non-PD cohorts. For the group with PD, of all the clinical tests, UFOV Subtest 2 provided the highest correlation with both the GRS (r = –.607, p < .001) and the SMS (r = –.699, p < .001). Significant correlations were observed for both the PD and non-PD groups with clinical tests in each domain—cognitive, visual, neurological, and motor—although, as a whole, a higher number of correlations was seen for clinical tests associated with the PD group. For almost every clinical test in which a correlation was observed in both the PD and the non-PD groups, the observed correlation was stronger for the PD group, and the PD group performed more poorly than the control participants.

Table 3 presents the predictions of GRS (passing–failing) using logistic regression modeling based on all clinical variables that were significantly correlated with GRS. For Model 1, stepwise entry of these variables generated a model that used the UFOV Subtest 2 and RPW:

where 5.192 = constant, 0.012 = coefficient for UFOV Subtest 2, and 0.532 = coefficient for RPW.

The GRS outcome in the model corresponds to the log ratio of the probability of failing over the probability of passing the road test. Model 1 accurately classified 81% of participants with PD as passing or failing the on-road evaluation and explained 55% of the variance with sensitivity = 83%, specificity = 78%, positive predictive value (PPV) = 83%, and negative predictive value (NPV) = 78%. When applied to non-PD participants, this model accurately classified 88% as passing or failing the on-road assessment and explained 31% of the variance with sensitivity = 50%, specificity = 90%, PPV = 20%, and NPV = 97%. For Model 2, forced entry of additional variables from clinical tests (contrast sensitivity [D] and Hoehn and Yahr stage on meds) resulted in a slight increase in explained variance, but decreased accuracy, sensitivity, specificity, PPV, and NPV. Because Hoehn and Yahr stage was specific to participants with PD, this model was not used for non-PD participants.

Model 3 represents the predictions of SMS using linear regression based on all clinical variables that were significantly correlated with SMS. Stepwise entry of these variables generated an outcome model using UFOV Subtest 2 and Hoehn and Yahr stage on meds that accounted for 55% of the variance in SMS.

where 330.323 = constant, 0.170 = coefficient for UFOV Subtest 2, and 36.318 = coefficient for Hoehn and Yahr on meds.

The objective of this study was to pilot the framework necessary for a predictive model for on-road outcomes by using screening tests administered by a CDRS and PD specialty neurologists. Like other samples, our sample with PD, which was representative pertaining to disease severity, performed more poorly on the UFOV (Amick et al., 2007; Uc et al., 2006a).

Our previous pilot study (n = 19 participants with PD, 104 control participants without PD) indicated that 42% of participants with PD failed the on-road assessment compared with 20% of the non-PD control participants (Classen et al., 2009), a finding that is consistent with the literature (Dubinsky et al., 1991; Heikkila et al., 1998; Wood et al., 2005). What is different in this sample is that the participants with PD (56.1%) failed the road test at an almost 5 times higher rate than the control participants (12.2%) and presented with significantly lower (poorer) SMS (M = 209) than the control participants (M = 246.2). A SMS cutoff score of <230 is predictive of failing our standardized on-road driving evaluation (Shechtman, Awadzi, Classen, Lanford, & Joo, 2010).

Consistent with previous literature, the UFOV (except for Subtest 1) emerged as a superior test to predict on-road outcomes (Amick et al., 2007; Classen et al., 2009; Devos et al., 2007; Heikkila et al., 1998; Klimkeit, Bradshaw, Charlton, Stolwyk, & Georgiou-Karistianis, 2009; Radford et al., 2004; Uc et al., 2009; Worringham et al., 2006). These findings underscore the importance of assessing the visual–perceptual and visual attention functions in PD, specifically the associated underlying, central and peripheral vision, processing speed, divided attention, and selective attention functions.

PD is manifested by decreases in motor, cognitive, and nonmotor function; therefore, our findings that participants with PD performed more poorly than non-PD participants on all significant clinical tests is not surprising. In a battery of tests administered by a DRS and data from neurological tests, we have found that the UFOV Subtest 2 and RPW correctly classified pass–fail outcomes for 81% of the PD participants. Although a step in the right direction, this screening battery is insensitive to 8 of 41 participants, including four false positives and four false negatives, in PD. The predictive power of the model may likely be strengthened by addition of the Hoehn and Yahr on meds but is unsubstantiated by our current findings, because of missing Hoehn and Yahr data.

More likely, predictive power may be strengthened by focusing on key findings from Uc et al. (2005) . These researchers presented an eloquent framework for assessing visual function in PD that extends the measures included in our current battery and, when used, may potentially increase sensitivity and PPV of the battery of tests. Although our battery, designed for community-dwelling older adults, did include basic sensory visual functions (far acuity, peripheral vision, and contrast sensitivity) as well as the UFOV, a visual attention measure of visual perception, we did not include visual–motor perception or visual cognitive tests (i.e., visual construction and visual memory). These functions have been shown to be impaired in people with even mild to moderate PD (Uc et al., 2005). Likewise, executive dysfunction is common in PD and is attributed to the impairment of the frontostriatal circuitry because of disturbed dopaminergic regulation (Cools, Stefanova, Barker, Robbins, & Owen, 2002). Therefore, our battery’s sensitivity may be increased by including a measure of executive function (e.g., Trials B) for set shifting, a very important function for safe driving (Wang et al., 2003). However, selection of a battery of these tests must also weigh into consideration the practicality and feasibility of performing these measures as a screening tool.

Interestingly, the RPW, a measure for PIGD, was a significant predictor with the UFOV of failing the on-road test. Uc et al.’s (2005)  finding, that a significant correlation exists between PIGD and visual attention, provides a mechanism for better understanding our finding that both RPW and UFOV are significant predictors of the on-road outcome. Uc et al. proposed the correlation exists because of the degeneration (severe cortical cholinergic deficit) of the central cholinergic systems regulating attention, accompanied by the significant loss of pedunculopontine nucleus neurons, which play an important part in both maintaining attention and locomotion regulation by means of basal ganglia activity. In essence, visual dysfunction contributes to poorer function in PD and potentially poorer performance on the road through its influence on cognition and locomotion. As such, when tested comprehensively in a large sample, visual attention, visual perception, and visual cognition may crystallize as the core predictors of failing an on-road assessment in PD with mild to moderate disease severity.

Our convenience sample was still relatively small and biased toward educated, mainly White, volunteer drivers. Although the groups were age-matched, we did not control for gender. All participants with PD came from a single movement disorders center, and we did not document the characteristics of those who refused participation, making our study prone to selection bias. Patients were informed that if their driving was deemed unsafe, reporting to the Department of Motor Vehicles might occur. This fact may have led to self-selection for safer drivers, which might mean that a “real-world” sample might identify even more increased risk in PD patients relative to age-matched control participants. Missing data from the UPDRS III motor scores and the Hoehn and Yahr stage scores may have underestimated the impact on the predictive model. We did not control for the effect of medications, sleepiness, or depression in this Parkinson’s disease sample and, as such, findings may be underestimated.

Collectively, these findings will set up the infrastructure as a pilot foundation for further research as follows: In search of a sensitive and efficient screening battery to predict on-road outcomes in PD, we must include neuropsychological measures of visual–motor perception, visual cognition (visual construction and visual memory), and executive function; control for the effects of medications, especially the anticholinergics; and control for depression and daytime sleepiness. Our data will need to be expanded to improve sensitivity and specificity of screening tests for driving in PD.

Clinical implications are that the UFOV and RPW are potentially promising screening tools because they predict PD road test outcomes. Their administration does not require specialty training, and they are fairly inexpensive and not time consuming (15 min) when administered by an occupational therapist or trained office staff in the neurology clinic. Although we cannot generalize our findings to the population of people with PD, they provide empirical support for clinical decision making by occupational therapy generalists and specialists; namely, that poor performance on UFOV 2 should elicit further referral for people with PD as it pertains to on-road safety. These findings also point to an opportunity for multidisciplinary collaboration among the occupational therapy practitioner, neurologists, neuropsychologists, and neuro-ophthalmologists, as it pertains to driving assessments in people with PD. A plausible future directive emerges: To examine and understand the mechanism underlying visual perception, visual attention, and visual cognitive deficits in PD and to develop novel intervention strategies that may prolong safe and independent, on-road driving for people with PD.

The project was funded by a National Parkinson Foundation grant as a Centre of Excellence for Parkinson’s Disease and Movement Disorders at the University of Florida and by the National Institute on Aging (R21) PAR-06-247 (principal investigator, Sherrilene Classen).