Mild Cognitive Impairment and Everyday Function: An Investigation of Driving Performance

Abstract

Mild Cognitive Impairment (MCI) involves subtle functional losses that may include decrements in driving skills. We compared 46 participants with MCI to 59 cognitively normal controls on a driving evaluation conducted by a driving rehabilitation specialist who was blinded to participants’ MCI classification. Participants with MCI demonstrated significantly lower performance than controls on ratings of global and discrete driving maneuvers, but these differences were not at the level of frank impairments. Rather, performance was simply less than optimal, which to a lesser degree was also characteristic of a subset of the cognitively normal control group. The finding of significantly lower global driving ratings, coupled with the increased incidence of dementia among people with MCI and the known impact of dementia on driving safety, suggests the need for increased vigilance among clinicians, family members, and individuals with MCI for initially benign changes in driving that may become increasingly problematic over time.

INTRODUCTION

For many older adults, the ability to drive safely is essential for maintenance of mobility, independence, health, and an active lifestyle^1–3. Cross-sectional and longitudinal studies indicate that older adults who have given up or reduced their driving report more negative health and psychosocial outcomes than those who have not, such as increased depressive symptoms^4,5, social isolation and diminished participation in out-of-home activities³, difficulty running their households and reduced access to essential services^5,6, loss of independence and personal mobility⁷, and greater likelihood of placement in long-term care². In general, driving restriction is associated with reduction in overall quality of life which, in turn, is associated with increased mortality⁸. Accordingly, research on driving behavior and safety among older adults is becoming increasingly important.

Due to rapid expansion in the number of licensed drivers over age 65⁹ and evidence that risk for dementia and other cognitive disorders increases with age^10,11, the effect of cognitive impairment on driving performance and safety among the elderly is also receiving increased attention in the literature^12,13. In one of the earliest studies of fitness to drive in Alzheimer’s disease (AD), Friedland and colleagues¹⁴ found that about 50% of AD patients had been involved in a crash in the five to six years prior to the study compared to only 10% of control participants. In another retrospective study, examination of driving records revealed that patients with dementia were 2.5 times more likely to have been involved in a motor vehicle collision than demographically-matched controls¹⁵. Studies that have utilized direct assessment approaches have similarly found impaired driving performance in AD. For example, a longitudinal study of on-road driving performance found that patients with mild AD exhibited a steeper decline in driving safety across assessments than did control participants¹⁶. The study also included 21 participants with very mild AD (i.e., Clinical Dementia Rating [CDR¹⁷] of 0.5—a rating acknowledged by the study’s authors to be consistent with mild cognitive impairment in the view of some); these participants also declined in driving skills over time, but not as steeply as the mild AD group. In a driving simulator study, Uc and colleagues¹⁸ found that drivers with AD were more likely than controls to experience an at-fault, rear-end collision and to exhibit other unsafe behaviors such as slowing down abruptly or prematurely, or swerving out of the traffic lane. Finally, an evidence-based review concluded that drivers with AD at a severity of CDR 1.0 pose significant safety risks as evidenced by both driving performance and crash measures¹⁹.

Mild cognitive impairment (MCI) often represents an intermediate stage between normal aging and AD²⁰. Persons with MCI exhibit cognitive and functional impairments similar to those that characterize AD, albeit of a milder and usually more focal nature^21–23. MCI is considered a strategic intervention point in the clinical management of AD and its functional sequelae ²⁴.

Based on our review of the literature, there have been four studies of driving performance, assessed either on the road or via crash data, among persons classified as CDR 0.5. Among these studies, two have dubbed this group “very mild DAT [Dementia of the Alzheimer Type]” ^16,25, one has termed this group “questionable dementia” ²⁶ and the final study used the term “probable AD at a severity of CDR 0.5” ¹⁹. Although a subset of MCI cases may be assigned a CDR rating of 0.5, this rating does not capture all cases of MCI, many of whom receive CDR ratings of 0.0. Moreover, because persons with AD also may receive CDR 0.5 ratings, a CDR of 0.5 may not be assumed to signal MCI. The MCI designation generally is reserved for patients subjected to additional, multifaceted criteria for case identification²⁷. To our knowledge, there have been no empirical studies of driving performance in MCI as defined by Petersen/Mayo criteria²⁷. This is a surprising knowledge gap given the heavy reliance of older adults on personal transportation for meeting various needs and the aforementioned associations between dementia status and negative driving outcomes.

In this study, we objectively examined the driving performance of well-characterized samples of cognitively normal older adults and persons with MCI using an on-road driving assessment. We hypothesized that persons with MCI would demonstrate decrements in driving skills and performance relative to cognitively normal controls.

METHOD

Participants

Potential participants were recruited to the Alzheimer’s Disease Research Center (ADRC) at the University of Alabama at Birmingham (UAB) through clinical cases presenting to UAB’s Memory Disorders Clinic and through community talks and health fairs. Irrespective of recruitment source, all ADRC enrollees underwent neurological examination and neuropsychological testing for subsequent consensus determinations of normal, MCI, and other conditions. Medical history and family reports were also provided. Diagnoses were determined in ADRC consensus conferences by neurologists, neuropsychologists, and nursing staff using Petersen/Mayo criteria for MCI²⁷. A diagnosis of MCI was not given to any ADRC participant with medical or psychiatric conditions that might account for their cognitive difficulties. Individuals with neurological illness or events such as prior stroke, traumatic brain injury, or brain tumor were excluded. Participants diagnosed with MCI or designated as normal controls via the consensus process were subsequently invited to enroll in the ADRC’s Measuring Independent Living in the Elderly Study. The present analyses include baseline data of current drivers from this ongoing longitudinal study.

The MCI group consisted of 46 current drivers (19 women and 27 men; 4 African Americans and 42 European Americans) with a mean age (± SD) of 71.30 (± 7.79); mean education of 14.83 (± 3.06) years; mean far visual acuity score, assessed with corrective lenses if typically worn, of 20/21 Snellen (range 20/36 to 20/13); and mean Dementia Rating Scale, 2^nd edition (DRS-2) ²⁸ total score of 132.60 (± 8.49). Three additional study participants with MCI had not driven during the past year (2 women, 1 man), and two additional women with MCI had never driven a car; none of these participants are included in the present analyses. Within the drivers with MCI, 43 were diagnosed with amnestic and 3 with non-amnestic MCI. The amnestic group primarily included persons with memory deficits only; a few had multiple-domain deficits including memory. However, the distinction between single and multi-domain impairment was not a focus of the consensus process. Fifty-nine neurocognitively normal control drivers (36 women and 23 men; 9 African Americans and 50 European Americans) had a mean age of 67.07 (± 6.72), mean education of 15.08 (± 2.60) years, mean far visual acuity score of 20/20 Snellen (range 20/30 to 20/13), and mean DRS-2 total score of 137.48 (± 6.26).

There were no group differences on race, education, or visual acuity (ps > .10). The MCI participant group was significantly older than the control group (p < .01), received, as expected, significantly lower DRS-2 total scores (p < .01), and had fewer females (p < .045). Relevant driving history variables were assessed with self report questionnaires. Twelve MCI drivers and 7 control drivers reported being involved as a driver in an automobile crash during the two years prior to study enrollment. Police were reportedly called to the scene in 7 of the MCI participants’ crashes and 6 of the controls’ crashes. Eight MCI drivers and 11 control drivers reported being pulled over by the police at least once in the past two years, resulting in traffic citations for 1 MCI participant and 2 controls. MCI drivers reported driving an average of 134.75 miles per week (s.d. 108.27), and control drivers 155.84 (s.d. 160.39), a difference that was not statistically significant (p > .10).

Written informed consent was obtained from all participants. The research was conducted in accordance with the Helsinki Declaration, and the UAB Institutional Review Board approved all procedures.

On-Road Driving Assessment

The driving skills of each licensed and currently driving participant were assessed in collaboration with the UAB Driving Assessment Clinic, using a standardized route and rating tool developed for research studies²⁹. This assessment was performed during the study visit under clear weather conditions. Under rainy or inclement weather conditions the evaluation was postponed and conducted in fair weather within two weeks of the study visit. Each drive lasted about 45 minutes and occurred between the hours of 12:30 and 4:00 p.m. Central Standard Time. Each participant drove the clinic’s vehicle (1998 Chevrolet Lumina) with dual controls under the supervision and evaluation of a Certified Driving Rehabilitation Specialist (CDRS) who was also a licensed occupational therapist (OTR/L). The CDRS was blind to participants’ group status (i.e., MCI or normal control). Although the present study used only one driving evaluator, previous research using the same route and evaluation tool has reported acceptable inter-rater reliability for global ratings (weighted kappa coefficient = .72) using two raters²⁹, one of whom conducted all the evaluations in this study. The drive was performed along a predetermined route that included (a) two- and four-lane roads with and without median strips and barriers, (b) highways and interstates, (c) intersections with and without traffic lights, and (d) stop and yield signs. Roads were also of varying traffic density. At multiple pre-established points during the drive, the CDRS coded each participant’s performance on specific driving skills such as lane control, gap judgment, turning, maintaining proper speed, stopping distance, signaling, obeying traffic signs, pre and post turn position, spacing, steer steadiness, pre and post crossing position, and proper scanning of driving space. These behaviors were sampled within varying contexts, including crossing intersections, merging, turning at intersections, exiting the interstate, changing lanes, driving on straight stretches, and taking curves.

Coding was done on a 5-point Likert scale: 1 = evaluator took control of car; 2 = unsafe; 3 = unsatisfactory; 4 = not optimal; and 5 = optimal. At the end of the drive, the CDRS also rated the participant’s overall driving skills on the same 5-point scale. Examples of behaviors associated with each rating level appear in Table 1.

Table 1.

Sample behaviors associated with Below Optimal ratings

Rating	Sample Behaviors
4—Not Optimal	Driving too fast or too slow (5 mph over or under limit)
	Driving too close to center of two-lane road
3—Unsatisfactory	Driving too fast around curve or turn (5 mph over limit)
	Driving on the center line of two-lane road
2—Unsafe	Driving too fast around curve or turn (10 mph over limit)
	Driving across the center line of two-lane road
1—Evaluator took control of car	Speed, lane position, or gap judgment behaviors requiring evaluator to engage dual brake or steering controls

Statistical Analysis

Because a variety of driving skills were coded, and because coding was performed at multiple time points during the driving evaluation, there were numerous indices of driving behavior that were not necessarily independent observations and therefore did not warrant individual examination. We developed a 2-step data reduction strategy. The first part of this strategy involved a priori selection of specific driving skills that were judged to be critical for proper and safe operation of a motor vehicle. These skills were turning (right and left turns were examined separately), lane control, gap judgment, steering steadiness, and maintaining proper speed, as well as the global rating that was assigned at the end of the assessment. The second part of this data reduction strategy involved creating composite variables by averaging each participant’s ratings on each of the selected driving skills across the three to five occasions that each skill was assessed. The number of occasions for measurement of each skill did not vary among participants.

We used a threshold of < 0.05 for statistical significance. We did not correct for multiple comparisons because 1) we reduced the data to essential skills in order limit the total number of comparisons, and 2) we were interested a priori in discrete driving behaviors in addition to the global driving impression of the CDRS.

An examination of group means on the driving variables of interest (Table 2) revealed a restriction in range, because many participants received ratings at or near ceiling. This non-normal distribution of the variables prohibited the use of parametric tests in our analyses. Therefore, we recoded each participant’s average scores on the driving variables, as well as each participant’s global rating, as dichotomous variables with 0 indicating “less than optimal driving” (mean score less than 5) and 1 indicating “optimal driving” (mean score of 5). Chi-square analyses were then performed to examine the proportion of MCI versus control participants who were designated as “less than optimal” on each driving variable. Next, logistic regression was used to estimate unadjusted and adjusted (for age and gender) odds ratios (ORs) with 95% confidence intervals (CIs) for less than optimal driving only for those variables on which the χ2 tests revealed significant group differences. Group, age, and gender were entered sequentially so that potential effects of age and gender on the association between diagnostic classification and driving outcome could be assessed. All analyses were performed using SPSS 12.0 (SPSS Inc., Chicago, IL).

Table 2.

Group ratings on driving skills

Variable	Range	Controls, n = 59	MCI, n = 46
Right turn	1–5	4.94 (0.12)	4.75 (0.75)
Left turn	1–5	4.82 (0.28)	4.72 (0.34)
Lane control	1–5	4.94 (0.17)	4.86 (0.22)
Gap judgment	1–5	4.97 (0.05)	4.93 (0.12)
Steer steadiness	1–5	4.92 (0.16)	4.80 (0.30)
Maintaining speed	1–5	4.93 (0.12)	4.85 (0.22)
Global rating	1–5	4.83 (0.38)	4.63 (0.49)

For Range, 1 = evaluator took control of car; 2 = unsafe; 3 = unsatisfactory; 4 = not optimal; and 5 = optimal. Note: No ratings of 1 were given in this study.

Values are mean (SD).

RESULTS

Table 3 displays the results of the χ² analyses that were performed to determine the proportion of participants in each group who received less than optimal ratings on the driving variables examined. MCI participants were significantly more likely than controls to receive less than optimal ratings on left-hand turns, lane control, and the global rating. MCI participants also tended to receive less than optimal ratings on maintaining proper speed and gap judgment. There were no group differences for right-hand turns or steering steadiness.

Table 3.

Group comparisons on proportion of participants receiving less than optimal ratings

Variable	Controls, n = 59	MCI, n = 46	χ2	p value
Right turn	12 (20.3)	10 (21.7)	0.03	.861
Left turn	22 (37.3)	27 (58.7)	4.76	.029
Lane control	8 (13.6)	18 (39.1)	9.07	.003
Gap judgment	14 (23.7)	18 (39.1)	2.89	.089
Steer steadiness	24 (40.7)	24 (52.2)	1.38	.241
Maintaining speed	21 (35.6)	24 (52.2)	2.90	.088
Global rating	11 (18.6)	20 (43.5)	7.67	.006

Values are n (%) and represent participants who received less than optimal ratings

At the lower end of the ratings distributions, mean ratings between 3.0 and 4.0 indicating unsatisfactory performance occurred for 4 MCI participants and 2 control participants in the domains of right and left turns and for 1 control participant in the domain of lane positioning. Mean ratings of 2.0, indicating unsafe performance, occurred for 3 MCI participants and no control participant in the domain of right turns. No instance occurred in this study in which the evaluator was required to take control of the car.

We had an insufficient number of cases with nonamnestic MCI to conduct comparisons of their driving performance to those with amnestic deficits. One of the three nonamestic MCI participants received a global rating of 4; the other two received a rating of 5.

Results of logistic regression models estimating ORs for less than optimal ratings on left turns, lane control, and the global rating appear in Table 4. For left turns, the unadjusted OR indicated that MCI patients were 2.39 times more likely than controls to receive a less than optimal rating. After adjusting for age and gender this OR decreased to 1.93 and was no longer significant (p > .10). For lane control, the unadjusted OR indicated that MCI patients were 4.10 times more likely than controls to receive a less than optimal rating. This OR was attenuated to 3.69 but remained significant after adjusting for age and gender. Finally, for global ratings, the unadjusted OR indicated that MCI patients were 3.36 times more likely than controls to receive a less than optimal rating. Adjustment for age and gender resulted in an OR of 4.23, which remained significant. Overall, the association between MCI status and less than optimal driving skills was accounted for by age and gender in one (left-hand turn) but not other (lane control and global rating) driving outcomes.

Table 4.

Odds Ratios and 95% CIs for Less than Optimal Driving Performance

	Left turn		Lane control		Global rating
	Unadjusted, OR (CI)	Adjusted,† OR (CI)	Unadjusted, OR (CI)	Adjusted,† OR (CI)	Unadjusted, OR (CI)	Adjusted,† OR (CI)
MCI	2.39 (1.09, 5.26)	1.93 (0.82, 4.54)	4.10 (1.58, 10.62)	3.69 (1.30, 10.46)	3.36 (1.40, 8.07)	4.23 (1.47, 12.15)

Controls are the comparison group.

^†Adjusted for age and gender

DISCUSSION

The primary contribution of this study is the finding of subtle functional decrements in discrete and overall driving skills and behaviors in persons with MCI. Conceptualized as an instrumental activity of daily living (IADL)—even as “the ultimate IADL” ³⁰—driving may be added to the list of IADL domains such as handling finances³¹ in which the functional performance or efficiency³² of individuals with MCI differs significantly from that of cognitively normal controls.

It is equally important to note that the performance decrements found in this study did not rise to the level of frank driving impairments. Such impairments would have resulted in mean ratings associated with descriptors of “unsatisfactory,” “unsafe,” or “evaluator took control of car.” Instead, mean differences in group ratings were small in magnitude (see Table 2), with both MCI participants and control participants receiving high mean ratings representing performance levels between “optimal” and “not optimal.” As a group, however, it appears that individuals with MCI are less likely than cognitively normal peers to seamlessly perform certain routine driving maneuvers, and they may more often evoke a global impression of less-than-optimal driving performance.

As is apparent in Table 3, there was individual variability in both the MCI and control groups, such that from 14% to 41% of control participants received less than optimal ratings on each of the six discrete driving behaviors, as did 22% to 59% of MCI participants. In both groups, a large proportion of suboptimal ratings occurred in conducting left hand turns and steadiness of steering. Thirty seven and forty one percent of the controls, along with 59% and 52% of the MCI participants, committed minor infractions during left turn and steering maneuvers, respectively. The greatest disparities in proportion of suboptimal ratings between groups occurred in maintaining lane control and negotiating left hand turns; these maneuvers may represent greater demands on cognitive abilities beginning to be affected in MCI, particularly in executive functions needed to coordinate responses under challenging conditions. Although speculative, this interpretation is consistent with recent research demonstrating the importance of executive function to maintenance of IADL function³³.

Relative to the ratings of discrete driving behaviors, the global driving ratings in this study, as in most on-road driving studies, were more subjective, in that they were based on the evaluator’s overall impression of the drive. Despite the fact that these ratings were not derived quantitatively from the ratings of discrete behaviors, the evaluator’s global impressions were no doubt influenced by her immediately preceding observations, and corresponding ratings, of each driving behavior. It is therefore probable that these global ratings represent a robust index of participants’ overall driving performance.

Limitations of this study include the cross-sectional design, which does not allow us to conclude definitively that subtle differences in driving behavior among participants with MCI are due to progressive underlying changes in the brain, although this explanation is plausible. It will be of interest to continue this research over time in the final cohort of this longitudinal study. Another potential limitation is that a restricted range and modest variability were obtained in the driving ratings in this study, emphasizing that the performance of controls and MCI participants was similar and grossly intact. We dichotomized our outcome variables and have interpreted our findings with due caution.

We controlled for the influences of age and gender on driving, and our groups were equivalent on indices of visual acuity. Although our sample did not include individuals with physical disabilities or pain syndromes that would preclude driving, it is possible that unmeasured physical limitations or pain may have affected driving performance. In addition, this study primarily recruited individuals with MCI who presented to a memory disorders clinic and therefore may not be representative of a population-based sample of MCI cases. A related issue is that the large majority of persons with MCI in this analysis were classified with amnestic MCI. At least one report has suggested that amnestic deficits may be no more common than nonamnestic deficits constituting alternate forms of MCI³⁴. The driving performance of our MCI sample therefore may not represent all MCI cases, especially those who present with primary nonamnestic deficits. To our knowledge, driving performance among individuals diagnosed with nonamnestic MCI remains to be explored and compared to that of individuals with amnestic MCI; unfortunately, we had insufficient nonamnestic cases to conduct such comparisons.

Strengths of this study include the use of well-characterized samples of MCI and cognitively normal control participants who were consensus-diagnosed using multiple sources of information and following standardized and widely accepted criteria for MCI²⁷. In addition, we used a comprehensive and objective evaluation of driving performance rather than self or informant report of driving competence. Although clinically meaningful, the latter indices are subject to biases associated with recall, diminished awareness, denial of deficits, affective states, and social desirability by cognitively impaired individuals and their proxies^35,36. Finally, the driving evaluations in this study were conducted by a licensed occupational therapist (Ms. Elgin) who is a certified driving rehabilitation specialist involved in both research and clinical driving evaluations. This rater was, and remains, blind to MCI classification of study participants in this ongoing research.

Clinicians should recognize that a substantial proportion of individuals with MCI might be experiencing subtle changes in driving skills that do not rise to the level of impairments, and in most cases do not warrant driving restriction or cessation, but certainly do warrant monitoring. At the same time, results of this study suggest that a lesser but still fair proportion of cognitively normal older adults also evidence less than optimal driving behaviors, particularly in the performance of left hand turns and steering steadiness. Although individuals with MCI progress to dementia at a much higher rate per year than cognitively intact peers³⁷, neither MCI nor “normal” cognitive aging are static conditions. Even among cognitively normal older adults who may not progress to either MCI or dementia, subtle driving changes could progressively worsen. Because driver interventions may be more effective when problems are identified early, monitoring for changes in driving among all older adults may be beneficial.

The present finding of significant differences between MCI and controls on global driving ratings extends recent evidence³⁸ that some but not all drivers with early or very early Alzheimer disease, and relatively fewer normal elders, decline in driving abilities over time to a degree that is potentially hazardous³⁸. These findings argue for increased vigilance among clinicians, family members, and individuals with MCI for initially benign changes in driving that may become increasingly problematic over time. Clinicians should become aware of and utilize available resources for on-road driving evaluations in their vicinities, because such referrals can provide a valuable source of objective information regarding driving performance. Finally, given the relationship between complex attention, processing speed, and driving outcomes such as crash risk³⁹, interventions that improve function in these cognitive domains may hold promise for maintaining driving skills and safety in MCI.