Deficits in Attention and Visual Processing but not Global Cognition Predict Simulated Driving Errors in Drivers Diagnosed With Mild Alzheimer’s Disease

Abstract

This study sought to predict driving performance of drivers with Alzheimer’s disease (AD) using measures of attention, visual processing, and global cognition. Simulated driving performance of individuals with mild AD (n = 20) was contrasted with performance of a group of healthy controls (n = 21). Performance on measures of global cognitive function and specific tests of attention and visual processing were examined in relation to simulated driving performance. Strong associations were observed between measures of attention, notably the Test of Everyday Attention (sustained attention; r = −.651, P = .002) and the Useful Field of View (r = .563, P = .010), and driving performance among drivers with mild AD. The Visual Object and Space Perception Test–object was significantly correlated with the occurrence of crashes (r = .652, P = .002). Tests of global cognition did not correlate with simulated driving outcomes. The results suggest that professionals exercise caution when extrapolating driving performance based on global cognitive indicators.

Older drivers are at an increased risk of being involved in motor vehicle collisions (MVCs). ^1,2 This is especially true for individuals diagnosed with Alzheimer’s disease (AD) whose declining cognitive function combined with lack of insight make them particularly at risk. ³ It is estimated that 5 million older Americans have AD, a number that is set to increase given the aging of the North American population. ⁴ Canadian data suggest that the number of older drivers with dementia in the province of Ontario has grown from less than 15 000 in 1986 to approximately 34 000 in 2000 and is expected to reach nearly 100 000 in 2028. ⁵

A number of studies estimate that among drivers with AD and related dementias, the risk of MVCs is at least twice as high as healthy older drivers of a similar age, ^6,7 and other researchers estimate an even higher risk. ^8,9 Driving simulator studies have found that drivers with AD show an increased number of crashes and unsafe behaviors, poorer vehicle control, slower response time, and poorer performance on a composite measure of performance (for a fulsome summary of existing literature, see Uc and Rizzo ¹⁰ ). Another study found among individuals with mild AD, simulated driving resulted in a significant number of errors and that simulated driving performance was so degraded that it was considered unsafe. ¹¹ Cox and colleagues ¹² found that in comparison to the control participants, individuals with AD had cognitive difficulty operating the simulator, drove off the road more often, spent more time driving slower than the posted speed limit, applied less brake pressure in stop zones, spent more time negotiating left turns, and drove worse overall. According to Uc and Rizzo, ¹³ driving simulators offer optimal stimulus and response control in a safe environment allowing for the analysis of patterns in driving behavior of healthy older drivers and those with AD.

Considering the research cited previously, drivers with AD show poorer driving performance in the driving simulator. Still, it is noteworthy that variation in driving performance exists within this group of drivers, and a prime challenge remains discriminating between who is safe and who should cease driving. Indeed, although a diagnosis of AD alone does not necessitate the immediate removal of driving privileges, it may alert health and occupational practitioners of the need for more in-depth assessment into fitness to drive as well as ongoing monitoring. ¹⁴ The safety concerns that accompany an AD diagnosis may discourage on-road testing, and health practitioners are often required to decide whether their patient can safely drive a vehicle. Thus, within the context of clinical care, assessment of fitness to drive among drivers with AD is often accomplished by administering measures of global cognition, referring patients for a more specific and in-depth neuropsychological testing or referring patients to a comprehensive driving assessment. Ultimately, professionals concerned with fitness to drive among patients with AD are in need of specific and efficient tools to make a determination, and research suggests that physicians in particular feel that they lack such tools. ¹³

Tests of general cognition are often used as a screening tool for cognitive impairment and driving safety. Indeed, Odenheimer and colleagues ¹⁵ reported that in a small sample that included some participants with dementia, a test of global cognition (ie, Mini-Mental State Examination [MMSE]) was found to be related to on-road performance (r = .72, P < .01). Similarly, Cox and colleagues ¹² found that a composite score of simulated driving performance correlated significantly with scores on the MMSE (r = −.403, P = .011). Some authors suggest that among individuals with neurological disorders, such as dementia, poor performance on measures of general cognition (ie, MMSE) be used to justify further neuropsychological evaluation ¹⁶ or on-road testing, ¹⁷ whereas others caution against using the MMSE in isolation to predict on-road performance due to its lack of sensitivity in relationship to driving safety. ¹⁸

In their practice guidelines, both the American Medical Association (AMA) and the Canadian Medical Association (CMA) suggest that abnormalities on the MMSE, clock drawing, or Trail Making Test B should lead to more in-depth assessment of driving ability. ^19,20 Although these recommendations are indeed based on research evidence, for example, the study by Roy and Molnar, ²¹ both sets of guidelines indicate that none of these tests have high enough sensitivity and specificity to be relied upon as a sole predictor of driving ability, an opinion that has been echoed by many experts in the field. ^22,23

Driving is a complex and multifactorial behavior that draws upon multiple physical, sensory, and cognitive processes and is impacted by personal and environmental characteristics. Anstey and colleagues ²⁴ sought to develop an evidence-based model to identify the factors that enable driving safety among all older adults, and through an extensive search of the scientific literature, they identified cognitive, visual, and physical factors as contributing to driving safety. Similarly, Reger and colleagues ²⁵ conducted a meta-analysis to determine the relationship between neuropsychological functioning and driving ability in patients with dementia. Their results indicated that a number of measures including those of general cognition, attention, and visuospatial skills were significantly associated with on-road tests and other driving outcomes tests (ie, tests of knowledge, simulator evaluation, etc).

Driving errors among individuals with AD have been attributed to impairments in attention and visual processing. Uc and colleagues ²⁶ administered a neuropsychological test battery assessing visual and cognitive functioning to 33 drivers with mild AD and 137 healthy controls. The participants were also assessed in terms of errors in an on-road driving task. Their results showed that the drivers in the AD group made more errors during the driving task and that performance on the neuropsychological test battery significantly predicted those errors. Rizzo and colleagues ²⁷ examined collision avoidance in 39 licensed drivers (21 with AD and 18 controls) who completed a simulated drive. The results showed that participants with AD crashed more often in comparison to controls and were more likely to experience close calls. The authors also found that visuospatial impairment and performance on the Useful Field of View (UFOV) task to be strongly predictive of crashes in the driving simulator.

The literature in this field indicates significant differences in driving errors between drivers with AD and healthy controls and that these differences are due to underlying cognitive (ie, attention) and visual mechanisms. Current practice approaches involve assessing at-risk drivers with measures of general cognition or through more specific neuropsychological testing. Given the multifactorial nature of the driving task, the fact that AD is associated with deficits in particular domains, and the cautions expressed by other researchers surrounding the use of global measures of cognition to determine driving fitness, there is a need to examine the magnitude of association between measures of attention and visual processing skills used in neuropsychological assessment and driving performance of drivers with AD. Thus, the goals of this study were to examine the simulated driving behavior of drivers diagnosed with mild AD and to examine the roles of attention, visual processing, and general cognition in predicting simulated driving performance. It was predicted that measures of global cognitive functioning, attention, and visual processing would be significantly associated with simulator driving performance, including crash. In line with the existing literature in this area, it was also hypothesized that drivers with mild AD would perform significantly poorer on the driving simulator task in comparison to healthy controls.

Methods

Participants

A group of individuals diagnosed with mild AD (N = 20) and a group of healthy older adult controls (N = 21) participated in the present study. All participants were older than 65 years and English speaking. In addition, all participants held a valid driver’s license. The mean age of the mild AD group was 78.5 years (standard deviation [SD] = 7.2) with a range of 66 to 90 years, the mean years of education was 13.05 (SD = 3.9), and the group was comprised of 9 women and 11 men. The mean age of the control group was 77 years (SD = 5.86) with a range of 68 to 86 years, the mean years of education was 13.14 (SD = 3.18), and the group was comprised of 11 women and 10 men. No statistically significant differences were observed between groups in terms of demographic characteristics.

The participants diagnosed with mild AD were recruited from the Memory Disorders Clinic at the Bruyère Continuing Care Center (ie, a tertiary care facility) in Ottawa, Ontario, Canada. Patients who had a diagnosis of probable mild AD were contacted in order to determine their willingness to participate in the study. The patients were assessed for severity using the Global Deterioration Rating Scale, and only participants in the mild stage of AD were included in this study (ie, Stage 3). ²⁸ Patients taking psychoactive medications, such as acetylcholinesterase inhibitors, were included in the AD group. The exclusion criteria included serious visual or hearing impairments left uncorrected (eg, cataracts), serious health problems aside from dementia (eg, mental illnesses, history of head injury, epilepsy, apoplexy, heart attacks, hypertension, and sleep apnea), any history of substance abuse, and any history of learning disabilities. The MMSE scores of participants with AD ranged from 11 to 30.

The convenience sample of healthy older adults (control group) was recruited through announcements placed in community newspaper. The control participants underwent an initial 20-minute screening call in order to determine whether they qualified to participate in the study. Control participants taking any medications that could alter cognitive abilities were excluded. These exclusion criteria were based on research that indicated certain conditions (eg, psychiatric conditions, history of head injury, epilepsy, apoplexy, heart attacks, hypertension, and sleep apnea) could have an effect on cognitive abilities. ²⁹ None of the control participants had abnormal MMSE scores of less than 25. All participants completed a consent form as approved by the Ethics Committee at the Bruyère Research Institute and at the University of Ottawa.

Measures

Global functioning

Mini-Mental State Examination. ³⁰

The MMSE is one of the most widely used brief screening instruments in dementia. The administration is very quick, at only 5 to 10 minutes per participant. Scores below 24 usually indicate abnormality or probable dementia. ³¹ A cutoff score of 27 is usually used to identify normal healthy adults.

Mattis Dementia Rating Scale (DRS-2 and DRS-2 Alternate). ^32,33

The DRS-2 assesses attention, memory, visuospatial construction, conceptualization, and initiation/perseveration. It was originally designed to assess adults with suspected dementia. It is organized so that the most difficult items in each domain are given first. The purpose of this administration scheme is to save time when testing less impaired participants. As such, high-functioning older adults can complete the battery in less than 10 minutes, but other participants may take as long as 45 minutes. The DRS-2 consists of 36 tasks and 32 stimuli, yielding 5 subscale scores and an assessment of the participant’s overall level of cognitive functioning. The entire test was administered according to the discontinue rules as outlined in the test protocol.

Visuospatial/perceptual abilities

Visual Object and Space Perception Test (VOSP). ³⁴

This is a measure of visuoperceptual and spatial abilities that specifically assesses object and space perception. All 8 subtests of this battery were administered. It should be noted that there was a semantic component to this test, as accurate performance relied on naming. When this was the case, the experimenter assisted participants by asking them to describe the object/animal or to state its use. A VOSP object perception composite score was calculated by adding the first 4 subtests, and a VOSP space perception composite score was calculated by adding the last 4 subtests, and these calculations were extracted from the user manual.

Tests of attention

Test of Everyday Attention. ³⁵

The Test of Everyday Attention (TEA) assesses attention in an ecologically valid manner as the tests are related to everyday tasks. Research has shown that the factor structure of the TEA corresponds with contemporary evidence supporting several independent attentional circuits in the brain. ³⁶ These factors include sustained attention, selective attention, and attentional switching.

The scores from the 8 subtests in the TEA were aggregated so that they aligned with the 3 factors outlined in the study by Robertson and colleagues. ³⁶ To accomplish this, standard scores (ie, z scores) were first computed for each subtest, and subsequently, variables corresponding to each of the 3 factors were summed. The standardized scores for subtests Visual Elevator (time score − switch) and Telephone Search While Counting (dual task decrement) were reversed so that the direction data were consistent with all other subtests in the TEA (ie, higher values indicate better performance).

Useful Field of View. ³⁷

The UFOV task is a computerized measure of attention that examines 3 test variables and is composed of 3 subtests, entitled processing speed, divided attention, and selective attention. Each subtest presents participants with stimuli for increasingly shorter durations, and after each presentation, participants are asked to indicate what they had seen. The shortest presentation times that result in correct responses are recorded. Performance on the UFOV is often cited as being related to performance on functional activities requiring sustained attention, such as driving. ³⁸ Among individuals with AD, research shows that UFOV scores correlate significantly with on-road driving performance. ³⁹

Simulated driving

Participants were asked to complete a simulated driving scenario mimicking an on-road evaluation. The STISIM Drive software (version 2.08.004; Systems Technology, Inc, Hawthorne, California) was implemented on a Dell Precision M6300 laptop computer (Intel Core 2 DUO processor, 2.10 GHz/2.0 GB RAM) with a 17-in display running Windows XP. The laptop was equipped with a brake/throttle and steering wheel (model G25 [Logitech, Newark CA]; Figure 1). The driving simulator was developed as a tool to generate laboratory tasks relevant to psychomotor and cognitive demands of real-world driving. This simulator allows for the inclusion of interactive vehicles on all lanes, buildings, traffic control devices, and pedestrians. The simulated drive was accompanied by realistic audio effects that provide acceleration and deceleration cues. Instructions to the drivers (eg, turn left/right, lane changes, and speed maintenance, etc) were given through laptop speakers. The driving simulator had a 60° horizontal field of view and a 75° vertical field of view. The frame rate of the simulator was also 30 frames per second (30 Hz).

Figure 1.

Portable driving simulator.

Before beginning, participants completed a comprehensive training session including a thorough explanation of the task, an accommodation phase during which they practiced operating the pedals and steering wheel, followed by a training course that took approximately 20 minutes to complete (see Lemieux and colleagues ⁴⁰ for a detailed description). The training course was designed to ensure that participants practiced maintaining their speed, braking, stopping, making left and right turns, and negotiating traffic. Upon completion of the accommodation phase, all participants reported being comfortable using the controls, including steering, braking, and accelerating.

The assessment course used was programmed to mimic a driving assessment by a provincial regulatory body in Canada and is described by Weaver and colleagues. ⁴¹ It was 12.3 km long, based on a real segment of road found in Thunder Bay, Ontario, and included driving in residential, highway, and urban environments. This particular assessment course was selected in our study for several reasons. Notably, given that the course is utilized by a provincial licensing body to differentiate between safe and unsafe drivers, it was found to exhibit high face validity. Similarly, the evaluation course allowed for the collection of 2 aggregate measures of driving performance: the total number of errors recorded by the simulator and a structured rater score. The course presented participants with a wide variety of typical driving situations allowing for generalizability. Most importantly, previous research examining driving behavior among drivers from across the driving lifespan (ages 18-83 years) shows that scores on the simulated assessment course correlate highly with those on the real-world driving course. ⁴²

Measurements of the simulated driving task were generated both from driving performance as recorded by the computer as well as through a demerit point assessment. Driving errors and parameters were recorded (eg, speed exceedances, stops sign violations, traffic light violations, etc) by the simulator. The simulator software automatically records driving violations according to specific in-program parameters, including centerline crossings, road edge excursions, failure to stop at a stop sign or red light, speeding (>5 km/h over the speed limit), illegal turns, off-road crashes, and vehicle collisions. ⁴³ The sum of frequency counts for these measures is reported as the total errors in the simulator. The driving-related variables collected by the driving simulator are presented in Table 1.

Table 1.

Differences in Simulated Driving Performance Between Participants With AD and Controls.

Variable Classification	Variable	AD group		Control group		df	F	P
		Mean	SD	Mean	SD
Intersection behavior	Total number of traffic light tickets	2.15	0.93	0.43	0.60	1, 39	49.97	<.001
Speed	Number of speed exceedances	16.15	4.97	12.24	4.74	1, 39	6.66	.01
	Over the speed limit percent of time	28.89	15.07	12.95	8.04	1, 39	18.11	<.001
Lateral control	Total number of centerline crossings	7.70	6.11	2.62	3.34	1, 39	11.08	<.001
	Total number of road edge excursions	14.35	13.45	7.52	9.76	1, 39	3.49	.07
	Out of lane percent of time	9.00	9.28	3.46	5.40	1, 39	5.53	.02
Crash	Total number of crashes	3.65	3.27	1.05	1.24	1, 39	11.59	<.001
Composite indicators	Rater score	261.00	96.49	107.74	59.64	1, 39	37.84	<.001
	Simulator errors	50.55	24.04	27.86	16.18	1, 39	12.69	<.001

Abbreviations: AD, Alzheimer’s disease; SD, standard deviation.

All simulated drives were recorded and scored by 2 independent raters using a structured demerit point assessment. Both raters independently assessed a video playback of the simulated drive after data collection was completed. They were also blind to the group to which the participants belonged. Inter-rater reliability was found to be high (r > .9), and as such, an average of both scores was taken as the rater score. None of the participants in this study reported any discomfort associated with the driving simulator, a condition often referred to as simulator adaptation syndrome.

Procedure

Potential control participants completed a short demographic questionnaire by telephone, and if they met the inclusion criteria, they were asked to participate in the study and their testing session was scheduled.

All patients at the tertiary care facility with a diagnosis of mild AD were contacted in order to verify whether they were willing to participate in the present study and that they met the inclusion criteria. Once a participant agreed to participate, they were asked to complete a brief demographic questionnaire. All of the participants with AD were diagnosed by the supervising neurologists at the tertiary care memory clinic. All diagnoses of dementia were accomplished using the general guidelines for the assessment of dementia in hospitals, providing a multimodal approach to diagnosis of dementia which greatly reduces diagnostic error. ⁴⁴ Additionally, diagnosis of AD was completed using the diagnostic criteria outlined by the National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA), which has excellent predictive validity. ⁴⁵

Following the completion of the demographic questionnaire, participants were booked for 2 testing sessions with approximately 1 week separating both sessions.

Participants with AD were asked questions pertaining to the Global Deterioration Rating Scale (GDS) in order to verify whether they were in fact in the mild stage of dementia as suggested by the neurologist (stages 3 and 4), and participants whose disease was at other stages were not selected for participation. Additionally, they completed the MMSE. The control participants were also asked to complete the MMSE. Following this screening, testing commenced immediately.

Testing sessions lasted approximately 21/2 hours. All participants underwent a neuropsychological and computerized assessment including a test of general cognitive functioning (DRS-2), visuospatial/perceptual abilities (VOSP), attention (TEA and UFOV), and processing speed (UFOV). Participants also completed a simulated driving assessment and were asked to drive as they normally would, obeying traffic rules and road signs. All neuropsychological and computerized testing was completed according to the protocol specified by each test. The neuropsychological and computerized testing was administered in the presence of the participant and the investigator only. Following data collection, each drive was rated independently by 2 raters using the playback feature of the STISIM Drive.

Results

Driving Performance

The 2 groups of participants were compared using a series of between-group analyses of variance (ANOVAs) with group (AD and control group) as the primary independent variable.

Comparison of the driving variables (Table 1) revealed that drivers with AD consistently performed poorer in comparison to healthy older drivers. In particular, drivers with AD exceeded the posted speed limit significantly more often in comparison to controls (F _1,39 = 6.66, P = .01) and spent an average of 29% of the time driving over the speed limit, a significantly higher proportion than controls who spent an average of 8% of the time over the speed limit (F _1,39 = 18.11, P < .001). In terms of lateral control of the vehicle, drivers with AD crossed the center line (F _1,39 = 11.08, P < .001) more frequently than controls and were found to spend an average of 9% out of their own lane, significantly higher than controls who spent an average of 5% out of their own lane (F _1,39 = 5.53, P = .02). Interestingly, no significant differences in road edge excursions were observed (F _1,39 = 3.49, P = .07).

Comparison of performance between AD and controls at intersections revealed that drivers with AD failed to stop at traffic lights significantly more often than controls (F _1,39 = 49.97, P < .001). Drivers with AD were also found to have difficulty safely interacting with other road users in the driving simulator as evidenced by significantly more crashes compared to controls (F _1,39 = 11.59, P < .001).

The composite measures of driving performance offer a global indication of participants’ driving behavior in the context of the simulated assessment course. The results revealed that the number of errors accrued by the driving simulator was significantly higher among drivers with AD (mean [M] = 50.55) than among controls (M = 27.86; F _1,39 = 12.69, P < .001). Similarly, drivers with AD were assigned over 4 times the number of demerit points (M = 261.00) in comparison to the control group (M = 59.64) during the simulated driving assessment, a difference that reached statistical significance (F _1,39 = 37.84, P < .001).

Cognitive Measures

Prior to calculating the degree of association between the cognitive and neuropsychological measures with indicators of driving performance, the cognitive measures were first compared between experimental groups. To this end, ANOVAs were computed for all cognitive measures with group (AD and control group) as the primary factor.

As anticipated, drivers with AD were found to have poorer performance on the MMSE (F _1,39 = 20.72, P < .001) and the DRS (F _1,39 = 50.95, P < .001) compared to controls. The results also indicated that drivers with AD performed worse on the object perception subscale of the VOSP (F _1,39 = 4.82, P = .03), but no group difference was observed for the space subscale of the VOSP (F _1,39 = 0.59, P = .45). In terms of the TEA, patients with AD showed poorer performance on each of the components of attention assessed by the TEA. Specifically, in comparison to controls, patients with AD showed lower performance in visual selection (F = 8.16, P < .01), sustained attention (F = 17.26, P < .001), and switching attention (F = 37.33, P < .001). Results of UFOV task showed statistically significant differences between drivers with AD and controls on all 3 subtests. The descriptive statistics for each measure and test group are presented in Table 2.

Table 2.

Differences in Cognitive Performance Between Participants With AD and Controls.

Variable Classification	Variable	AD Group		Control Group		df	F	P
		Mean	SD	Mean	SD
Global functioning	MMSE	24.00	4.86	29.00	1.30	1, 39	20.72	<.001
	DRS	115.00	12.96	136.38	4.43	1, 39	50.95	<.001
	DRS—attention	33.35	6.52	35.86	1.12	1, 39	3.01	.090
	DRS—initiation/perseveration	28.40	7.79	35.48	2.80	1, 39	15.27	<.001
	DRS—construction	6.55	3.19	6.00	0.00	1, 39	0.62	.433
	DRS—conceptualization	33.05	3.19	37.24	1.14	1, 39	6.47	.015
	DRS—memory	15.80	6.06	24.10	0.83	1, 39	38.66	<.001
Visual processing	VOSP—object	63.50	7.13	67.95	5.82	1, 39	4.82	.03
	VOSP—space	44.00	7.41	45.48	4.62	1, 39	0.59	.45
Attention	TEA—visual selection	−1.03	2.56	0.97	1.90	1, 39	8.16	<.01
	TEA—sustained attention	−1.21	2.29	1.14	1.16	1, 39	17.26	<.001
	TEA—switching attention	−2.04	2.33	1.94	1.81	1, 39	37.33	<.001
	UFOV—processing speed	143.15	142.79	24.91	16.74	1, 39	14.21	<.001
	UFOV—divided attention	406.10	130.90	152.52	116.45	1, 39	43.05	<.001
	UFOV—selective attention	452.40	90.86	304.76	115.13	1, 39	20.64	<.001

Abbreviations: AD, Alzheimer’s disease; DRS, Dementia Rating Scale; MMSE, Mini-Mental State Examination; SD, standard deviation; TEA, Test of Everyday Attention; UFOV, useful field of view; VOSP, Visual Object and Space Perception Test.

Association Between Cognitive Function and Simulated Driving Among Drivers With AD

One goal of this study was to assess the contribution of neuropsychological tests of global cognition, attention, and space processing in assessing driving outcomes among individuals with AD. To meet this goal, Pearson correlation coefficients were calculated between each of the neuropsychological test scores and key indicators of overall driving performance for the AD group only. Specifically, rater score, number of crashes, and total number of errors were used in this calculation given that previous research has shown the importance of these variables as global measures of performance within the driving simulator environment. ⁴²

Table 3 summarizes the correlations between cognitive tests and driving outcomes among individuals with AD. In terms of global measures of cognitive function, the results indicated nonstatistically significant relationships between the MMSE and the rater score, total errors, and number of crashes. A similar pattern of association emerged between the DRS where the DRS did not significantly correlated with simulated driving outcomes among participants with AD. Indeed, the MMSE and the DRS correlated significantly with one another (r = .856, P < .001).

Table 3.

Pearson Correlation Coefficients Indicating Relationship Between Tests of Cognitive Function and Simulated Driving Outcomes for the AD Group.

Variable Classification	Variable	Rater Score, r (P)	Errors in the Simulator, r (P)	Number of Crashes, r (P)
Global functioning	MMSE	−.399 (.081)	−.126 (.597)	−.193 (.416)
	DRS	−.383 (.096)	.112 (.639)	.096 (.383)
Visual processing	VOSP—object	−.102 (.670)	.257 (.275)	.652 (.002)
	VOSP—space	−.426 (.061)	−.074 (.758)	.252 (.283)
Attention	TEA—visual selection	−.129 (.587)	−.010 (.967)	.352 (.128)
	TEA—sustained attention	−.651 (.002)	−.366 (.113)	−.006 (.980)
	TEA—switching attention	−.379 (.099)	−.118 (.621)	−.098 (.681)
	UFOV—processing speed	.172 (.469)	−.226 (.338)	−.255 (.278)
	UFOV—divided attention	.563 (.010)	.381 (.097)	.159 (.504)
	UFOV—selective attention	.396 (.084)	.271 (.248)	.197 (.406)

Abbreviations: AD, Alzheimer’s disease; DRS, Dementia Rating Scale; MMSE, Mini-Mental State Examination; SD, standard deviation; TEA, Test of Everyday Attention; UFOV, useful field of view; VOSP, Visual Object and Space Perception Test.

In terms of visual processing, a statistically significant correlation was observed between the VOSP–object scores and the number of crashes in the simulator (r = .652, P = .002). The associations between measures of attention and simulated driving indicated that the TEA (sustained attention) was significantly associated with the rater score (r = −.651, P = .002). Similarly, the UFOV (divided attention) was also significantly correlated with the rater score (r = .563, P = .010) such that poorer performance on the UFOV was associated with poorer performance in the simulator as determined by a rater.

Discussion

The present study examined simulated driving performance of older drivers with mild AD in relation to performance on global cognitive measures as well as measures of attention and visual processing. The study yielded a number of interesting findings.

First, the results indicated that among the group of drivers with AD, neither measure of global cognitive functioning (ie, MMSE or DRS) was associated with driving outcomes among the drivers with AD. Certainly, cognitive function is related to driving, and the research shows that some cognitive functions are more salient during driving than others; attention, for example, is often discussed as an important determinant of driving safety. ⁴⁶ Tests of global cognitive function may not be sufficiently sensitive to detect deficits in functions related to driving, such as attention. This perspective aligns with the literature indicating that measures of global cognitive functioning lack specificity and, thus, should not be used in isolation to identify at-risk drivers. ^18,22,47 The results presented here emphasize the importance of tailoring protocols aimed at identifying medically at-risk drivers to assess the deficits associated with a particular disease category as well as those factors that directly relate to driving safety.

Correspondingly, the data did reveal statistically significant associations between more specific measures related to attention and visual processing. The results showed that among drivers with AD, a measure of visual processing (VOSP–object) was significantly correlated with the occurrence of crashes in the driving simulator. Avoidance of crashes requires the rapid perception of visual stimuli, processing of information, and appropriate response maneuvers, usually in complex driving situations. Drivers with AD may experience difficulties retrieving patterns of response due to memory impairments and may have problems implementing responses due to issues with executive functions. Visual processing, however, is a critical initial process in crash avoidance and one that is often impaired in drivers with AD. ⁴⁸ This association illustrates the importance of visual processing in explaining the most severe simulated driving offense (ie, crash) among drivers with mild AD.

The results indicated a significant correlation between measures of attention and performance in the simulator as assessed by a rater. In particular, in the sample of drivers with AD, the TEA, sustained attention, and the UFOV, divided attention, were significantly correlated with the rater score, accounting for 42% and 31% of the variance in the rater score, respectively. The role of attention in safe driving has been well established in the literature, ^10,12,49 however, the majority of research has found strong correlations between driving and computerized tests such as the UFOV or the Attention Network Task. ^41,50 Our data revealed that the magnitude of the association between the TEA (sustained attention) and the rater score was even greater than for the UFOV. These results suggest that the TEA could potentially be used to inform decisions surrounding safe driving among individuals with AD.

Finally, similar to the existing literature on simulated driving and dementia, ¹¹ the data indicated that participants with AD show poorer performance on variables related to longitudinal and lateral control of the vehicle, the number of crashes incurred in the simulator, and composite variables of driving performance. Although the driving-related data presented here derive from a driving simulator, they are consistent with data collected from on-road assessments in that drivers with AD have been shown to commit more errors in comparison to neurologically healthy controls. ⁵¹ This is not overly surprising given that behavior in the driving simulator is highly associated with real-world driving ^52,53 and contributes to the generalizability of the present findings. These findings are also in line with the existing literature and highlight the value of the driving simulator in differentiating between mild drivers with AD and healthy controls. The present research shows that simulated driving performance is impacted by the cognitive decline associated with mild AD; whether this difference is related to on-road crash risk cannot be addressed given our current data.

Limitations

It is curious that drivers with AD show statistically significant decrements on nearly all measures of simulated driving performance assessed by the driving simulator. This finding suggests that participants with AD may have experienced difficulty controlling the simulated vehicle, which contributed to their overall scores. One explanation for this observation may be that in comparison to healthy controls, drivers with AD experience more difficulty adapting to the new environment presented by the driving simulator. Indeed, findings from Lundberg and Hakamies-Blomqvist ⁵⁴ show that driving in an unfamiliar vehicle leads to increases in cognitive demand and consequent decreases in driving performance, especially among older drivers with cognitive decline. It is plausible that for the drivers with AD who participated in this study, the novelty of the driving simulator may have increased the complexity of the driving task over and above what they would normally experience during naturalistic driving in their own vehicles. To address the novelty of the simulated environment and its potential impact on driving performance, researchers could individualize the accommodation phase (also referred to as an orientation scenario) in so far as participants would be required to reach a base level of performance before proceeding to the assessment scenario. Using such a protocol, clinical populations with cognitive decline would complete longer accommodation phases in order to become appropriately familiarized with the driving simulator.

Nevertheless, the capacity to adapt to novel situations, including unexpected events or unfamiliar road configurations, is also an essential component to safe driving. One study including participants with dementia did find a strong correlation (r = −.67) between simulated driving and on-road performance such that fewer errors in the simulator were associated with better on-road driving. ⁵⁵ Although the simulator provides an ideal environment to explore the perceptual and cognitive mechanisms that underlie driving, the use of a driving simulator in this study prevents direct generalization to on-road driving performance.

Summary and Clinical Implications

The present study assessed a sample of patients with dementia having mild AD, administered a number of clinically relevant cognitive tests, and measured objective driving performance through a driving simulator. The driving simulator used in the present study was portable, and the driving assessment course was short in duration (approximately 20 minutes). The simulator component was well tolerated by participants with AD and control participants alike, and we observed that driving performance correlated with performance on a number of cognitive tests. Specifically, we found that among drivers with mild AD, visual object processing significantly explained the occurrence of crash, and measures of attention were associated with a global index of driving performance. These findings underscore the important and fundamental contribution of attention in driving in general. Finally, based on the data we obtained, we question the use of measures of global cognition, such as the MMSE, as a measure of driving safety. These results contribute to a growing body of research, placing a spotlight on visual processing and attention as key underlying mechanisms to safe driving in a variety of at-risk populations. ⁵⁶

In many jurisdictions, physicians are responsible for determining and reporting on medical fitness to drive, yet the research shows that they are often not comfortable assessing driving safety because they lack the appropriate tools and guidelines to do so confidently. ^{57 –59} Although follow-up research is needed, these results suggest that measures of attention and visual processing could be particularly useful for professionals who are often responsible for determining fitness to drive, yet lack the instruments that would allow them to identify specific cognitive and visual deficits related to driving with exactitude. The exploration and identification of such tools will aid professionals to comfortably make this determination. ⁶⁰