Evaluating Driving Performance of Cognitively Impaired and Healthy Older Adults: A Pilot Study Comparing On‐Road Testing and Driving Simulation

To the Editor: The responsibility for determining the driving fitness of older adults is increasingly falling upon the medical profession. Medical professionals and licensing agencies need to be equipped with methods to evaluate fitness to drive in order to achieve a balance between individual independence and societal protection. Although road testing has been considered the criterion standard by which to evaluate driving competence, it is costly and can be dangerous when the driver is very incompetent. Furthermore, with few exceptions, road test courses and evaluator training are not standardized. Driving simulation appears to be a sensitive method to objectively evaluate driving performance,1–6 but the recurrent criticism is that it has not been compared with on-road testing. Driving simulators are increasingly employed in research because they greatly reduce the risk associated with impaired drivers. Because simulators provide a safe, cost-effective environment for mimicking driving tasks, and contain features that allow for simulating real-world driving, they are ideal for most types of driving research, but their ability to reveal or predict actual on-road performance in older adults, particularly those with cognitive impairments, has not been reported. We compared on-road and simulated driving performance of nine older adults (five men and four women) aged 67 to 78. Of these participants, four were classified as cognitively impaired (Folstein Mini-Mental State Examination (MMSE) <27, Trails B <50th percentile, Clock Drawing Test abnormal) and five were control subjects. Overall, the mean MMSE score was 25.6, the median was 27.5, and the range was 18 to 30. The subjects completed a 30-minute driving simulation through an urban course, programmed to require execution of maneuvers that allowed assessment of the ability to drive and perform executive function tasks in a multitask environment and a 30-minute on-road test. Conditions that reveal discriminating errors were emphasized. The driving simulator used was a commercially available product, the STISIM Drive, developed by Systems Technology, Inc., Hawthorne, California. A dual-brake–equipped automobile was used for all on-road evaluations, and an occupational therapist dually certified as a driving rehabilitation specialist and commercial driving instructor who was blind to the results of the driving simulation evaluated performance. Performance measures for both driving tasks included hazardous or potentially catastrophic errors, traffic violations, and rule violations. The average scores for simulated and on-road performances were significantly correlated at −0.670 (n = 9, P = .048, Pearson correlation, 2 tailed). The results show a strong association between the two driving tests, and they are negatively correlated: the lower the score on the simulator (performance with few errors), the higher the score on the road test (performance with competency ratings>90%). There was a strong association between hazardous and lethal errors committed on the simulator and failing the road test (r = −.830, P = .010 (2-tailed) for hazardous errors; r = −.816, P = .011 (2-tailed) for lethal errors). Subjects who failed the on-road test committed an average of 5.4 hazardous errors and an average of 4 lethal errors on the driving simulator compared with no errors in those categories by subjects who passed the on-road test. Despite the small sample size, there was a surprisingly strong correlation between simulated driving performance and on-road driving performance by cognitively impaired and healthy older adults. Our results suggest that high fidelity driving simulation is a sensitive method to objectively evaluate driving performance and may be a valid alternative to on-road testing. In fact, driving simulation allows for evaluation of risky situations that would be unsafe to duplicate in on-road testing. Furthermore, some situations cannot be evaluated on-road because of the lack of control over numerous variables. Although the number and sequence of left and right turns and of intersections controlled by traffic lights and stop signs matched those on the simulator, it was not possible to control lights changing or remaining green or to simulate pedestrians or other vehicles placing themselves in the path of the driver during the on-road test. In addition, certain errors that might occur on the simulator (crashes, crossing the midline, etc.) could not be permitted during on-road testing for obvious safety reasons and would have been influenced by evaluator prompting. Nonetheless, errors committed during the driving simulation and on the road fell within the identified error categories, allowing a score to be calculated. These scores showed significant correlation despite the small sample and lend support to the utility of driving simulation as a method to objectively evaluate driving performance of cognitively impaired and healthy older adults.