The findings from the current study indicate that traditional neuropsychological measures are the best predictor of driving safety. Specifically, we found that performance on trails B and copy of the ROCF differed between marginal and safe drivers. Longer completion times for trails B and inaccurate drawing of the ROCF were associated with more errors on the road test. A composite measure derived from the neuropsychological tasks was the only predictor of driving errors and the composite neuropsychological score correctly classified safe and marginal drivers for more than 70% of cases.
Consistent with these results, previous studies have also shown that neuropsychological measures that are visually mediated and multifactorial are predictive of on-road driving safety. Heikkilä et al. (1998)
found in a sample of 20 male PD patients that time to perform a visual reasoning task, errors on a visual reasoning task, and time to recall visual material accounted for 62% of the variance of on-road test scores. Using a larger sample size (n
= 51) Radford et al. (2004)
found that performance on a visual cancellation test was an important predictor of road test performance. One limitation of the two previous studies is the fact that the neuropsychological measures selected are not commonly used clinically. By contrast, Worringham et al. (2006)
, like the current study, employed widely used neuropsychological tasks. In a sample of 25 PD patients the authors reported that performance on the Oral Symbol Digit Modalities (speed of information processing), the Purdue Pegboard (motor speed), and the Pelli Robson (contrast sensitivity) correctly classified 68% of PD drivers. In a large sample of PD drivers (n
= 71) a measure of executive functioning with the motor demands controlled for (reaction time for trails B—reaction time for trails A) was found to be the only predictor of errors during a live road test performed during a distracting task (Uc et al., 2006a).
Taken all together, these data suggest that neuropsychological measures that require rapid responding, visual spatial cognition, and executive functioning are most useful for distinguishing safe from marginal drivers. At this time, given the small sample sizes of previous studies, cut scores have not yet been determined for specific neuropsychological tasks to distinguish safe from unsafe driving. The findings from this study suggest that neuropsychological assessment is helpful for discriminating safe drivers, but almost 30% of our participants were misclassified. Currently, performance on certain neuropsychological measures may not clearly indicate which drivers are unsafe. Rather, neuropsychological assessment may be most helpful to inform clinical decisions as to which patients should undergo formal driving evaluations. Neuropsychological screening to determine which patients need the further procedure of undergoing a road test is helpful as it may limit the number of individuals who have to undergo this stressful and expensive examination.
Given that previous studies in PD patients have found that visually mediated neuropsychological measures were related to driving safety and visual impairments are prevalent in PD it was hypothesized that measures of basic visual perception (CS) would be related to driving skills. It was observed that scores on the FACT significantly differed between safe and marginal drivers. Additionally, compared to safe drivers, marginal drivers tended to require greater contrast levels to perform the backward visual masking task.
CS deficits are common in PD, can affect performance on higher-level tasks (Amick et al., 2003
), and several studies including this report, have found a relation between CS and driving performance (Uc et al., 2006a,b). Our primary hypothesis that CS performance would predict driving performance was not supported. Despite the difference in the CS profiles of safe and marginal drivers, a composite measure of this visual ability did not predict driving safety. This is consistent with two previous reports, which found that CS performance, while correlated with measures of driving performance, was not predictive of driving errors (Uc et al., 2006a,b). There has been one report that the Pelli Robson, one of three measures, is a significant predictor of driving skills (Worringham et al., 2006
). Yet the clinical utility of the Pelli Robson seems limited because only 68% of the PD patients were correctly classified in the discriminant function and the predictors were from highly diverse domains. A large study of AD patients and healthy older adults did observe that CS predicted errors on a road test (Uc et al., 2004
) but differences in driving performance measures (their task required memorization of the route) could explain the discrepant findings. A larger study employing sensitive CS measures, such as those used in this study is necessary to properly evaluate the relative contribution of CS to driving safety in this population.
Compared to safe drivers, marginal drivers performed worse on the UFOV (part III) and poorer performance on part III of the UFOV was associated with more errors on the road test. A composite measure of UFOV scores, however, was not predictive of road test performance. A previous study using a newer version on the UFOV reported that safe and unsafe PD drivers did not perform differently on this task (Worringham et al., 2006
). Uc et al. (2006b) have found that the performance on the UFOV is predictive of landmark and traffic sign identification but not at-fault safety errors. While the UFOV has been found to be predictive of driving safety in healthy elderly (Owsley, 1994
), in PD patients this measure may not be as useful for determining driving safety. Our studies as well as others have observed that measures of neuropsychological function and to a lesser degree contrast sensitivity are associated with driving performance. Across studies the UFOV is less consistently associated with driving performance and may lack clinical utility in the PD population.
In the current study marginal drivers, unlike safe drivers, committed more frequent tactical errors such as clumsy lane changes (55%), turning into incorrect lanes (55%), not checking traffic when backing out (45%), and not signaling (45%) and these errors are committed by less than 20% of healthy elderly drivers (Grace et al., 2005
). Marginal drivers also committed more frequent operational and strategic errors including decreased awareness of how their driving was affecting others (73%); they hesitated longer before making a turn (55%), did not accelerate to the proper speed (45%), and had lapses of concentration (45%). Previously it was observed that these types of errors occurred in less than 10% of elderly control participants (Grace et al.).
While other studies of live road test performance by PD patients have not use this classification system, it is clear that unsafe PD drivers have previously been found to commit both tactical and operational errors. Previously reported tactical errors include difficulty with visual scanning (Radford et al., 2004
), turning left (Heikkilä et al., 1998
), changing lanes, staying in the lane, monitoring blindspots, reversing car parking, and traffic light controlled intersections (Wood et al., 2005
). While less frequently noted, other studies report that PD patients committed operational errors such as poor positioning of the car (Radford et al., 2004
) and difficulty driving in traffic flow (Heikilla et al.). To summarize, PD patients commonly violate basic rules of the road and have difficulty monitoring the driving environment. Importantly these mistakes might be amendable to correction through interventions such as driving education or the use of adaptive equipment (to aid scanning and monitoring of blind spots).
Stolwyk et al. (2006)
analyzed types of errors committed on a simulated driving task, using the same classification system as this study. They observed that PD patients more frequently committed tactical and operational compared to strategic errors and performance on neuropsychological measures was correlated with the tactical and operational errors (Stolywk et al.). The authors report that performance on measures of executive abilities and working memory are associated with tactical driving errors. By contrast, measures of basic visuo-perception and information processing speed were observed to relate to operational driving skills. We did not observe the same distinctions in our correlational analyses. In our study, CS, executive functioning, visual attention, and visuospatial abilities were associated with tactical and operational elements of driving. It should be recognized that Stolywk et al. used a driving simulator task and precisely measured participants’ responses to variations in the difficulty of these tactical and operational maneuvers. The degree of experimental control in the study by Stolywk et al. may have isolated particular cognitive contributions to specific aspects of driving. In the current study, which used an on-road driving test, the method for assessing driving safety was less experimentally controlled but is believed to better resemble real life driving situations that PD patients will encounter.
At present time it is clear that the skills necessary to safely operate a motor vehicle are multifactorial. Driving safety could be determined with approximately 70% accuracy in PD patients based on neuropsychological performance. This suggests that screening visuospatial skills, executive functions, and possibly visual perception may help identify which patients should undergo a road test. Larger studies including longitudinal investigations are necessary to confirm the predictive utility of neuropsychological measures for determining PD driving safety.