Abstract
INTRODUCTION
Cognitive impairment occurs along a wide spectrum, ranging from normal aging, to the mild deficits associated with mild cognitive impairment (MCI), to the more moderate/severe impairments associated with Alzheimer’s disease (AD) and related dementias. Approximately 5.3 million Americans have AD or related dementia [1]. By 2025, this figure is expected to rise to over 7.1 million Americans [1].
Driving is a major source of independence for older adults with and without cognitive impairment, and one in five licensed drivers will be 65 years or older by 2030 [2]. However, driving is a complex task that requires the integration of multiple cognitive, perceptual, and motor abilities [3]. Cognitive impairment can affect a variety of domains that are crucial for safe driving, including diminished navigational ability, decision making, and the ability to interpret surroundings [4]. In spite of the prevalence of cognitive impairment and its potential impact on driving, there are no published guidelines as to when patients with MCI or mild AD should be considered unsafe to drive; however, most guidelines deem that individuals with moderate to severe AD are unsafe to drive [5–7]. Although impairments in driving have been demonstrated in patients with AD and MCI, diagnosis alone does not imply definitive driving impairment as this varies with the type and severity of the cognitive impairment [8]. A driving assessment with high reliability, validity, and specificity is required in order to help physicians accurately determine whether individuals from these populations are able to drive safely, or whether their licenses should be revoked or restricted.
Our objective was to conduct a systematic review and meta-analysis of the three primary driving assessment methods (i.e., cognitive tests, on-road evaluations, and driving simulation) within the MCI and AD population. Specifically, we performed a meta-analysis to investigate: (1) the predictive utility of individual cognitive and sensory tests, (2) the predictive utility of impairment in specific cognitive and sensory domains (i.e., attention, executive function, memory, psychomotor function, vision, visuospatial function, global cognition), and (3) on-road pass/marginal/fail classifications in patients with MCI and AD. Furthermore, we systematically reviewed the results of on-road and simulator-based studies to determine areas and degree of driving impairment in patients with MCI and AD. Finally, limitations as well as future directions in the AD and MCI driving literature are discussed.
MATERIALS AND METHODS
Study selection criteria
MEDLINE, EMBASE, and Psycinfo were searched for articles written in the English language that compared cognitive functioning with driving performance in AD and healthy populations using the keywords “automobile driving,” “car driving,” “driver,” “driving behavior,” “driving ability,” “automobile driving examination,” “assessment,” “Alzheimer’s disease,” “mild cognitive impairment,” “dementia,” “neuropsychological tests,” and “cognitive assessment” with the appropriate MESH terms until August 14, 2014. The reference list of relevant reviews and included studies were also searched for additional relevant articles. Case studies, reviews, and abstracts were excluded. Articles were restricted to studies of human participants who had AD or MCI, established via commonly used and validated measures (e.g., Clinical Dementia Rating (CDR) score [9], the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria [10]). Only studies that administered commonly used cognitive tests and either on-road tests, driving simulations, caregiver reports of driving independence and level of impairment, or self/caregiver-reported driving status were included. Given the paucity of research investigating the driving performance of patients with AD and MCI, no minimum sample size was required in order to meet inclusion criteria. Further, no restrictions were placed on the type of driving outcomes reported, for the same reason.
A separate search was conducted on driving simulator and on-road studies in AD and MCI populations for the purpose of a systematic review. MEDLINE and Psycinfo were searched for articles using the terms “driving,” “assessment,” “driving simulator,” “on-road,” “Alzheimer’s disease,” “mild cognitive impairment,” and “dementia” with the appropriate MESH terms until May 25, 2013. The reference list of relevant reviews and included studies were also searched for additional relevant articles. Case studies, reviews, and abstracts were excluded. No minimum sample size was required in order to meet inclusion criteria. Articles were restricted to studies of human participants who had AD or MCI, established via commonly used and validated measures (e.g., CDR score [9], NINCDS-ADRDA [10] or similar criteria, Mini-Mental Status Examination (MMSE) [11]). Only studies that utilized driving simulations or on-road assessments were included.
Data synthesis and analysis
Effect sizes (ES) of cognitive/sensory test and domain predictors of driving outcomes were derived from descriptive or inferential statistics (e.g., correlation coefficients between test and driving performance, means and standard deviations of safe and unsafe drivers, odds ratios, t-values) and analyzed in a random effects model. The ES from the different cognitive/sensory tests were analyzed as single tests (Maze test, Trail Making Test Part A (TMT-A) and Part B (TMT-B), Verbal Fluency, Useful Field of View task (UFOV), Benton Visual Retention Test (BVRT), Hopkins Verbal Learning Test (HVLT), Auditory Verbal Learning Test (AVLT), Finger Tapping, Rey-Osterrieth Complex Figure (ROCF), Contrast Sensitivity, Structure from Motion, and MMSE) and categorized into cognitive domains (executive functions, attention, visual memory, verbal memory, visuospatial function, vision, psychomotor, and global cognition). The coding sheet utilized included the following information: Study author and year, test/measure, driving outcome, sample size, dementia rating, age, gender, driving experience, and descriptive/inferential statistics (e.g., correlation coefficients, means and standard deviations, odds ratios, t-values). From this information, the following was calculated: ES, mean weighted random effects ES, I2, and 95% confidence interval. Due to the limited number of studies in the literature, when calculating cognitive individual and domain ES, all driving outcomes were pooled together, including on-road test scores, on-road test pass/fail classifications, on-road test errors, caregiver reports (e.g., impaired versus not impaired, drives with difficulty versus unable to drive), real world crash involvement, as well as driving simulator crash involvement and risky avoidance behavior. Effect sizes within the cognitive domains analyses were correlated with the mean age of patients in each study to determine whether the predictive utility of the cognitive domains varied as a function of age (i.e., whether the ES changed with age). Lastly, ES from the different cognitive and sensory tests were analyzed according to the driving outcome utilized (i.e., on-road test pass/fail classification, on-road test scores, on-road test errors, and driving simulator errors). This analysis was conducted to determine whether cognitive tests as a whole were able to predict particular driving outcomes better than others. All cognitive and sensory tests were pooled together for this analysis (i.e., including tests of executive function, attention, visuospatial function, motor function, visual and verbal memory, vision, and global cognitive function), rather than separating by domain, due to an insufficient sample size. Statistical heterogeneity was assessed using the I2 test, which describes the percentage of variation between study results. I2 values of 0.25, 0.50, and 0.75 indicate approximate classifications of low, medium, and high heterogeneity, respectively. Publication bias was assessed using funnel plots, Egger’s test [12], and the trim-and-fill method. The methods were conducted, and results reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15].
RESULTS
Search results
The search yielded 106 abstracts, from which three duplicates were removed. From the 103 abstracts, 53 were excluded for failing to meet inclusion criteria. The remaining 50 studies were examined in detail, from which 17 were excluded because they pooled the cognitive test data between patient participants [13, 15–30] and one because it was a dissertation [31]. The final sample included 32 studies (Fig. 1), 18 of which incorporated cognitive tests in conjunction with driving outcomes and were included in the meta-analysis, 19 involved on-road evaluations, and 9 involved driving simulation. Of these articles, 29 assessed the driving performance of patients with AD and 4 assessed the driving performance of patients with MCI. See Table 1 for the demographic characteristics of the study sample.
Cognitive assessment
Cognitive and sensory tests
Several tests assessing a variety of cognitive domains showed some utility in predicting driving outcomes (i.e., on-road safe/unsafe/borderline, on-road score, on-road errors, and simulator errors, caregiver ratings of driving impairment or driving independence (e.g., drives alone versus not alone), driving status, real-world collisions). While there did not appear to be much specificity by cognitive/sensory domain, in general, executive function tests had the largest ES and were the most represented in the literature (Fig. 2; ES [95% CI]). The Maze test (domain: Executive function; 0.88 [0.60, 1.15]), TMT-B (domain: Executive function; 0.61 [0.28, 0.94]), verbal fluency (domain: Executive function; 0.44 [0.03, 0.86]), TMT-A (domain: Attention; 0.65 [0.08, 1.21]), UFOV (domain: Attention; 0.34, [0.05, 0.63]), Structure from Motion (domain: Vision; 0.40 [0.01, 0.79]), Contrast Sensitivity (domain: Vision; 0.37 [0.04, 0.70]), Finger Tapping (domain: Psychomotor speed; 0.33 [0.12, 0.53]), BVRT (domain: Visual memory; 0.39 [0.10, 0.68]), ROCF-copy (domain: Visuospatial function; 0.37 [0.17, 0.57]), and MMSE (domain: Global cognition; 0.46 [0.24, 0.67]) were predictive of driving performance. Conversely, neither of the verbal memory tests, AVLT and HVLT, significantly predicted driving outcome (0.17 [–0.19, 0.53] and 0.07 [–0.50, 0.64], respectively).
Cognitive and sensory domains
Mirroring the individual test results, most cognitive and sensory domains showed some utility in predicting driving outcomes with minimal specificity (Fig. 3). Executive function (0.61 [0.41, 0.81]), attention (0.55 [0.33, 0.77]), visuospatial function (0.50 [0.34, 0.65]), global cognition (0.61 [0.39, 0.83]), visual memory (0.35 [0.13, 0.58]), and vision (0.26 [0.12, 0.40]) emerged as significant predictors of driving outcome. Psychomotor speed (0.51 [–0.14, 1.16]) and verbal memory (0.14 [–0.23, 0.51]) were not significantly predictive. Note that the discrepant ES between the psychomotor domain and finger tapping test above is due to the inclusion of additional psychomotor tests in the domain analysis; analyses on these individual tests could not be performed due to insufficient sample sizes.
Driving outcomes
ES from the different cognitive and sensory tests (i.e., all tests pooled together) were analyzed according to the driving outcome utilized. Results revealed significant ES for all driving outcome measures (Fig. 4): On-road score (0.59 [0.45, 0.73]), on-road safe/unsafe (0.59 [0.36, 0.83]), on-road errors (0.43 [0.20, 0.66]), and driving simulator errors (0.30 [0.20, 0.40]).
Age correlations
Visuospatial function ES correlated negatively with age (r = –0.62, p < 0.01) while psychomotor measures correlated positively with age (τ= 0.59, p < 0.05). The other correlations (attention: r = 0.72; executive functions: r = –0.94; vision: –0.02; visual memory: 0.15; verbal memory: τ= 0.42; global cognition: r = –0.35) were not statistically significant.
Study heterogeneity and publication bias
Statistical heterogeneity varied between analyses (Table 2). I2 ranged from 28% (vision) to 89% (psychomotor) in the cognitive/sensory domain analyses; and 0% (Finger Tapping) to 88% (TMT-A) in the cognitive test analyses. Additionally, publication bias was present in most of the cognitive domain analyses, as indicated by the trim-and-fill method and Egger’s test (Table 2).
On-road assessment
The on-road test is the most widely used method of driving assessment across all patient populations [32], including patients with AD [33–50]. However, the results of the current meta-analysis suggest that the outcomes of on-road assessments tend to be highly variable. Based on performance on an on-road test, patients receive an overall rating of driving performance based on the discretion of a formal driving instructor, with three possible outcomes: pass/safe, marginal/borderline, or fail/unsafe. Seven of the nine studies assessed on-road test outcomes using the Washington University Road Test (WURT) [40, 44] or an adapted version of the WURT with the same maneuvers and identical scoring procedures [36, 47]. A rating of “safe” is assigned when a driver’s behavior is not likely to result in a collision [44]. A rating of “marginal” is assigned when a driver has a small to moderate increased risk of collision or “traffic conflict” (e.g., driving too slowly, needs some assistance when performing routine driving maneuvers) [44]. Finally, a rating of “unsafe” is assigned when a driver poses a substantial risk of collision or traffic conflict (e.g., stopping unnecessarily, failing to stop at a stop sign or traffic light, unsafely changing lanes) [44]. The remaining two studies [43, 45] used different on-road driving tests. Both of these studies assessed driving maneuvers similar to the WURT (e.g., traffic light and stop sign behavior, lane changing, traffic merging, highway driving, etc.), were conducted by a formal driving instructor, and categorized patients into fail/unsafe, marginal/borderline, or pass/safe. Studies that administered driving tests were approximately the same duration (45–60 min) [37, 45–47] and under similar conditions (i.e., time of day, [36, 47] and good road conditions without precipitation [36, 47]). Given that pass, marginal, and fail classifications were assigned using similar criteria across studies, the number of patients and controls receiving each time of classification (i.e., pass/safe, marginal/borderline, fail/unsafe) were pooled and a chi-square analysis was run.
A few studies reported over 25% of AD patients failed the assessment [40, 45], whereas others reported a patient failure rate of 10% or less [37, 42]. Of the 460 patients with AD assessed using an on-road test, 214 (46.5%) patients definitively passed, 87 (18.9%) definitively failed, and 159 (34.6%) received a marginal rating (Table 3). The chi-square analysis revealed significant differences in on-road outcome (i.e., pass, marginal, fail) across CDR groups (χ2 = 116.634, p < 0.001). All studies [36, 47] used the CDR scale [9] to classify patients with very mild AD (CDR score of 0.5) or mild AD (CDR score of 1). All patients were diagnosed with AD using the NINCDS-ADRDA [10, 47] or similar [43, 44] criteria. Drivers with very mild AD (CDR = 0.5, n = 154) were less likely to receive a pass rating (56.5% versus 79.4%, p < 0.05), and more likely to receive a marginal (29.9% versus 19.0%, p < 0.05) or fail rating (13.6% versus 1.6%, p < 0.05) than healthy older adults (CDR = 0, n = 364). Drivers with mild AD (CDR = 1, n = 120) were less likely to receive a pass rating than both healthy drivers (41.7% versus 79.4%, p < 0.05) and drivers with very mild AD (41.7% versus 56.5%, p < 0.05), and were more likely to receive a fail rating than healthy drivers (33.3% versus 1.6%, p < 0.05) and drivers with very mild AD (33.3% versus 13.6%, p < 0.05).
One study investigated the driving performance of individuals with MCI using an on-road assessment [51]. The results suggested that those with MCI are more likely to receive a poorer rating than healthy drivers on lane control and overall performance; however, MCI patients may simply have less than optimal performance rather than definitive driving impairment [51].
Driving simulation
Results from driving simulator studies have replicated the findings of on-road assessment, demonstrating that some individuals with AD or MCI retain their driving ability [52–54] and that performance deterioration may be related to disease progression [55, 56]. Specific driving impairments have been identified in individuals with AD, including driving slower [55–57], applying less brake force when attempting to stop [57] and during collision avoidance [58], taking longer to complete left-hand turns [57], making judgment errors at traffic lights [56], unsafe outcomes in rear-end collision avoidance [54], shorter mean time to collision [55], greater risk of collisions [52, 56], and driving more poorly in general [55, 57]. However, other studies report a non-significant difference between drivers with AD and healthy controls in terms of vehicle control [53], centerline crossings [57], and collision involvement [53, 54].
A few driving simulator studies have investigated the driving performance of persons with MCI [55, 59, 60, 55, 59, 60]. The results of these studies suggest that individuals with MCI tend to have minor difficulties with driving rather than definitive driving impairments [55, 59]. Specifically, patients with MCI may be less likely to stop at intersections with stop signs as well as at light changing intersections [59] (although results did not reach statistical significance), and may perform worse on a car-following task [60] than healthy drivers without cognitive impairment.
DISCUSSION
Physicians and other healthcare professionals often do not feel confident assessing the driving ability of patients with cognitive impairment [61] due to the absence of a reliable and valid in-office driving assessment that is able to accurately differentiate between individuals who are safe to drive from those who are not. The results of the current meta-analysis suggest that measures of executive function, attention, visuospatial function, and global cognition may be predictive of driving performance (effect sizes≥0.5). Furthermore, TMT-B, TMT-A, and Maze tests (both errors and time) emerged as the best single test predictors of driving performance. The results of neuroimaging studies suggest that an extensive brain network is recruited during driving [62–67], supporting the notion that integration of multiple cognitive functions, including attention, executive function, and visuospatial function, is required to drive safely. Specifically, parietal [62–66], occipital [62–66], motor [62–67], cerebellar [62, 64–67], and frontal [58, 63–67] regions have been shown to be recruited during various driving tasks. Given that driving requires the coordination of multiple functions and recruitment of multiple brain regions, all of which can be impacted by AD and MCI, it is not surprising that TMT-A and TMT-B (measures of attention, processing speed, and mental flexibility) [68] as well as Maze tests (i.e., a measure of planning, speed, and visuospatial function) emerged as the best predictors of driving. Despite this, variability and inconsistencies remain, with some studies reporting that TMT-A, TMT-B, and Maze tests are predictive of driving [42, 69], whereas others demonstrate little predictive utility [38, 49, 54]. Several factors may contribute to these inconsistencies, including the absence of evidence-based cut-off scores [70], small sample sizes, variability in experimental procedures (e.g., outcome variables, administration parameters, pooling AD patients with controls for analyses [71], etc.), the complex and multi-faceted nature of driving [65], and the heterogeneous cognitive presentation of patients with AD and MCI. Specifically, due to the low number of studies included in the current meta-analysis, all driving outcomes (e.g., on-road test classification, on-road test score, on-road test errors, driving simulator crashes and risky outcomes, real-world crash involvement, caregiver reports on driving competency) were pooled together when calculating individual cognitive test and cognitive domain ES. In addition, there are currently no validated cut-off scores [70] that allow cognitive tests to be implemented by clinicians on an individual level to establish safety to drive. Consequently, too few studies have supported the predictive utility of any specific cognitive measure in the AD and MCI population to translate into clinical recommendations.
Effect sizes were the largest when the outcome investigated was on-road test safe/unsafe classification or on-road test score; the ES was smallest for driving simulator errors. This result suggests that on-road evaluations (i.e., test score and safe/unsafe classification) of driving ability appear to be the outcome measures most related to cognitive test performance. Thus, cognitive functioning may be a good predictor of on-road driving ability. This is practical, as the results of in-office cognitive assessments could assist clinicians in determining which patients are at increased risk of driving impairment and thus should be referred for a more comprehensive driving evaluation, administered by a specialist. However, given the relatively large variability associated with on-road outcomes and scores (i.e., confidence intervals), it would be important to determine whether there are specific, individual cognitive tests (i.e., with high specificity and evidence-based cut-off scores) that are best able to predict on-road test scores and safe/unsafe classifications. Given the insufficient study sample size, this was not possible in the current analysis. Consequently, the current results do not support cognitive measures being used to predict on-road performance on an individual basis. In addition, driving errors, as measured both via on-road and simulator evaluations, had the smallest ES. A possible explanation for this finding is that overt and critical driving errors (e.g., collisions, failure to stop at a traffic light or stop sign) occur relatively infrequently, resulting in low sensitivity.
Interestingly, the age of patients correlated negatively with visuospatial tests and positively with psychomotor tests. Thus, measures of visuospatial function may show greater predictive utility in studies that assessed younger patient and control cohorts, whereas the opposite pattern emerged for psychomotor tests. These discordant results are difficult to interpret as both domains are key basic features of driving; though psychomotor deficits are often present in older adults and these likely interfere with driving performance, the same can be said for visuospatial functions. A possible explanation is variability in the data, and other confounding factors (e.g. differing degrees of cognitive impairment across and within studies, driving outcomes utilized, etc.).
On-road and simulator assessments have yielded inconsistent results in terms of the safety to drive of persons with AD and MCI; however, the safety to drive appears to be associated with severity of cognitive impairment [40, 56]. Specifically, on-road results suggest that patients with very mild AD (CDR = 0.5; 13.6%) and mild AD (CDR = 1; 33.3%) are more likely to fail an on-road evaluation than healthy drivers (CDR = 0; 1.6%); patients with mild AD are more likely to receive a fail rating than patients with very mild AD. Patients with very mild AD (29.9%) are more likely to receive a marginal rating than healthy drivers (19.0%). Despite the increased tendency for patients with AD (CDR = 0.5 or 1) to receive a fail rating compared to healthy control drivers (18.9% versus 1.6%), it is important to note that patients with AD (CDR = 0.5 or 1) most commonly received a pass rating (46.5%) rather than a marginal (34.6%) or fail (18.9%) rating. Furthermore, the results of both on-road and simulator studies suggest that patients with MCI demonstrate minor, rather than definitive driving impairments [51, 59]. Despite this general trend, variability between studies is apparent due in part to the heterogeneous presentation of cognitive deficits and disease progression in patients with AD and MCI. Methodological inconsistencies between studies are also prevalent, including test administration and instruction (e.g., WURT versus adapted WURT versus unspecified road test), number and type of challenging events, test scoring, variations in driving simulator hardware (i.e., portable [55, 59] versus fully immersive [52–54, 60]) and software, complexity of simulator scenarios, driving simulator outcomes of interest (e.g., vehicle control [57, 60], intersection behavior [55, 59], time to collision [55], collisions [52–54, 56], accident avoidance [52, 58]), on-road test outcomes of interest (e.g., errors, score, pass/marginal/fail classification), level of traffic on-road (i.e., variability both within and between studies), etc.
Though the results have important implications, this study is not without limitations. The quality of a meta-analysis depends heavily on the quality of the original studies on which it is based; publication bias is one such factor, which influences meta-analytic outcomes. Most of the studies included had positive effect sizes (also indicated by the multiple significant Egger’s tests), and consequently the mean effect sizes were all positive (though not all were significant). Thus, results revealed that publication bias was present in the majority of the cognitive predictor analyses. Publication of all outcomes, including both statistically significant and non-significant descriptive data and effect sizes (i.e., instead of only reporting p-values), would allow for more accurate summaries of the literature. Furthermore, it is possible that variability in methodological quality across studies may be a factor contributing to the inconsistent, and often contradictory, results. For example, 22% of studies (n = 7) reported statistically significant age differences between patient and control groups, and only 34% of the studies included (n = 11) explicitly stated that patients and controls were age-matched.
Another issue is the relative paucity of published studies, and those that are published often have small sample sizes, which results in underpowered analyses. Only a few studies have been published on cognitive and driving abilities in AD, and even fewer in MCI, which prevented the inclusion of the latter patient population in the meta-analysis. Small sample sizes are often utilized due to the inherently expensive and challenging nature of driving research. As a result, the generalizability of study findings is difficult and often not possible. Of the 32 studies included in the current systematic review and meta-analysis, 41% (n = 13) utilized a control or patient sample group of less than 20 individuals [34, 72]. Exacerbating this issue is the trend for the included studies to pool participant groups for analyses (i.e., AD and controls analyzed together) [33, 57]. While this approach provides some information on the predictive power of cognitive tests for driving ability, it does not inform group differences and, in particular fails to address potential driving impairments associated with AD and MCI. Indeed, Reger and colleagues [71] demonstrated that after excluding studies that pooled participants, the mean ES were considerably smaller and even non-significant in some cases. Furthermore, given that there are multiple studies conducted by the same research groups, there is the potential concern of overlap of populations, and thus the double counting of participants, across studies.
Lastly, variability is especially prominent in this field on multiple levels, including the cognitive presentation of patients, diagnostic criteria and procedures utilized, and the areas and severity of driving impairment observed across patients. Consequently, it is sometimes difficult to find general patterns of neurocognitive impairment that correlate with driving ability. Variable neurocognitive functioning is inherently a part of AD and MCI populations and it remains unclear whether there is in fact a distinction between a diagnosis of MCI and a diagnosis of very mild AD [73]. A CDR score of 0.5 is often used to categorize both patients with MCI and patients with very mild AD [73, 74]. Furthermore, due in part to logistical barriers and potentially the use of variable diagnostic criteria, study recruitment captures patients at different stages of cognitive impairment, thus producing a heterogeneous sample. Given the heterogeneous presentation of MCI and AD (i.e., domains and severity of cognitive impairment, co-morbid psychiatric conditions, medications, etc.) coupled with the cognitively complex and multi-faceted nature of driving, patients can present with varying degrees and areas of driving impairment. Future large-scale studies should investigate the driving performance and associated brain activation patterns of various subgroups of AD (i.e., very mild, mild, and moderate) and MCI (i.e., single domain amnestic, single domain non-amnestic, and multi-domain) as cognitive and functional presentation is highly variable within these populations. This would serve as an important step in determining if certain driving behaviors and activation patterns are associated with different severities and cognitive presentations of patients with AD and MCI. Ultimately, isolating the brain regions that are predictive of specific driving impairments may lead to the development of more accurate cognitive measures of driving. Furthermore, given that hazard perception assessment tools have shown high predictive utility in terms of driving outcome, including real-world collision involvement [75, 76] and on-road test performance [77, 78], future research should investigate the ability of these measures to predict driving performance in the various subgroups of AD and MCI. If accurate cognitive-based measures are developed that can be administered in-office, physicians would be able to perform a quick, reliable, and valid assessment of patients and refer borderline drivers for more in-depth on-road assessments [61].
Footnotes
ACKNOWLEDGMENTS
This work was supported by an Alzheimer’s Society Research Program Research Grant from the Alzheimer’s Society of Canada as well as an Early Researcher Award from the Ontario Ministry of Research and Innovation awarded to Dr. Tom Schweizer, a CIHR Frederick Banting and Charles Best Canada Graduate Scholarship— Master’s awarded to Megan Hird, an Ontario Graduate Scholarship awarded to Peter Egeto, and the University of Toronto, George, Margaret and Gary Hunt Family Chair in Geriatric Medicine awarded to Dr. Gary Naglie.
