Naturalistic Driving Measures

A Growing Population of Older Drivers

Life expectancy in the US has increased such that a person who is 65 can expect to live 19 or more years, and someone who is 85 can expect an additional 6 to 7 years.¹ By 2050, 88 million (20%) Americans will be age 65 or more, and 25% of all drivers will be in this age group.^2,3 In 2016, among the pool of 42 million licensed drivers age over age 65 there were 290,000 injuries and over 7,400 deaths resulting from motor vehicle crashes.⁴ Because adults are driving longer, injuries and mortality from crashes will continue to increase based on the volume of active drivers. An increased crash risk among aging drivers has been attributed to different factors, including frailty, medications, and cognitive decline associated with Alzheimer disease (AD) or AD-related dementias (ADRDs).^5-7 The anticipated tripling of AD prevalence to nearly 14 million in the US by 2050 will also affect the driving and community mobility needs of all drivers.

Driving Decline in AD

Driving is a highly complex and dynamic activity that requires the rapid and simultaneous deployment of sensory (ie, vision, hearing, and touch), motor (ie, grip/release and range-of-motion), and cognitive (ie, divided attention, memory, and executive function) systems. The integration of these subserves the activity of driving. Age-related declines in sensory and motor systems are accelerated by neurologic disease, including cognitive disorders. In AD, driving ability is compromised by impairments in executive function, attention, reaction time, visuospatial skills, and memory. Compared with adults without cognitive impairments, persons with AD have a faster rate of decline in performance,⁸ which is accelerated further when dementia is severe.⁹ AD also confers an increased risk of motor vehicle crashes¹⁰ and a higher likelihood of failing a road test or having a crash during simulator tests of driving performance.^6,11 These findings extend to mild cognitive impairment (MCI) or early symptomatic AD. For many with early symptomatic AD, driving skills remain intact for several years after diagnosis. A recent review and meta-analysis, however, found that persons with AD were more likely to fail a road test compared with people without AD,¹² and the risk of driving-associated morbidity and mortality is ever increasing as the disease progresses.

Driving Performance in AD

Preclinical AD is the asymptomatic period (~15-20 years) during which abnormal levels of biomarkers are present without dementia symptoms. Commonly used biomarkers are obtained from blood or cerebrospinal fluid (CSF) or imaging studies (see Blood Tests for Alzheimer Disease and Neurofilament Light as a Dementia Biomarkerin this issue). Researchers have begun to use AD biomarkers to study the effects of preclinical AD on driving decline among cognitively normal adults over age 65. In a recent study, higher biomarker levels were associated with an increased number of driving errors on a road test, suggesting adults with preclinical AD have worse driving performance.¹⁴ In a follow-up study, adults over age 65 with preclinical AD more quickly received a “marginal” or “fail” rating on the road test after passing the road test at baseline, compared with those without preclinical AD.¹⁵ Most importantly, there were no differences between groups in cognitive functioning on a composite score. It is possible that the accumulation of amyloid and tau pathology may impair the complex processing and communicating in and across multiple brain regions required to sustain safe driving. Thus, driving may serve as an important functional neurobehavioral marker to better track changes during preclinical AD (Figure 1).

Current Challenges in Assessing Fitness to Drive

Assessing driving safety among people over age 65 and in the spectrum of AD is a significant challenge and undertaking for clinicians and researchers.¹⁶ Road tests and driving simulators are the conventional tools used, and both methods are helpful in evaluating poor driving performance or estimating crash risks. Because driving is an overlearned task, however, controlled conditions (eg, road test or driving simulator) may not reflect driving as it occurs on a daily basis or expose errors made by experienced drivers without cognitive impairment outside of these controlled conditions. Although driving simulators preserve the safety of the driver, simulators are typically single-site dedicated measures, participants may develop simulator sickness, and there is a higher cost associated with equipment, maintenance, and programming.¹⁷ Concerns about validity and reliability arise across simulator brands, models, and content specific to the programmed course. Similarly, a road test is only standardized in its local geographic context. There are some national considerations, however, including specific local or regional laws governing driving and behavior or customs of driving (eg, more speeding in rural areas). There are also personal considerations and preferences with respect to what type of vehicle an individual drives or rides on a daily basis. These are not necessarily accounted for within the road test assessment. Limitations of all road tests include the use of an unfamiliar vehicle, driving on unfamiliar roads, performance anxiety (poor performance), the Hawthorne effect (ie, a better performance due to being watched), and reliability of the evaluator and the scoring system.¹⁸

The Widening Gap in Fitness-to-Drive Assessment

Although there is no single recognized benchmark for fitness to drive, behind-the-wheel assessments are commonly used and offered by driving rehabilitation programs, costing an average $400. This cost is typically not covered by third-party reimbursement systems unless it is directly related to workers compensation or vocational rehabilitation.¹⁹ The evaluation is an observation session in a specially instrumented vehicle (ie, not the tested driver’s) on a course or setting unfamiliar to the driver. The evaluator is recommended to have a certified driver rehabilitation specialist (CDRS) certification. The CDRS certification is granted by the Association for Driver Rehabilitation Specialists, which has an extensive prerequisite criteria and time-intensive process to obtain certification. To date, there are only 370 certified and active people with CDRS certification in the US and Canada. A driver rehabilitation specialist (DRS) is a professional without the CDRS certification who may be able to perform similar driving assessments. There are approximately 750 DRS in the US and Canada, but not all CDRS and DRS professionals provide driving evaluation tailored to adults over age 65. In 2017, there were 44 million licensed drivers age 65 or more. The combined effort of 1,100 DRS and CDRS professionals is grossly underpowered to meet the current population’s needs and anticipated growth by 2050. The unmet need for screening and monitoring of fitness to drive among drivers over age 65 is a challenge for which our healthcare system is not prepared. Coupled with the increasing prevalence of ADRD and other neurologic conditions, this unmet need will adversely affect all members of society.

Using Naturalistic Data to Evaluate In Vivo Driving

Over the past decade, telecommunication broadband and telematics have engendered the recording, storage, and analysis of naturalistic driving behaviors using global positioning system (GPS) methods. GPS data loggers have several novel key advantages that exceed contemporary assessments (eg, road test and simulators). GPS data loggers are placed under the dashboard and plugged into an On-Board Diagnostic system (OBD II) port with minimal installation effort. The concealed placement has no effect on the behaviors of drivers or passengers. All of the raw data are extracted in real-time from the vehicle’s computer and transmitted via mobile communication towers to secured servers in real-time. This data extraction and transmission method has a significant benefit—if a vehicle is driven in areas where the signal is lost, data continues to be collected and is then transmitted when a strong signal is reestablished. As a result, the data logger yields real-time data gathered on a continuous basis from the actual environments in which people drive with minimal missing data. The data loggers are commercially available, require no modification to the vehicle, and collect large amounts of data on driving performance with minimal effort for the researchers or demand characteristics for the driver.

Over the past 4 years, we developed a naturalistic driving data collection system tailored to the needs of conducting driving research with adults over age 65. This system combines a modified, commercial, off-the-shelf data logger (Azuga G2 Tracking Device) used in fleet management, with custom codes (using R and ArcGIS) to capture, manage, curate, and analyze naturalistic driving behavior. This process is termed the Driving Real-World In-Vehicle Evaluation System (DRIVES).²⁰ The DRIVES data logger plugs into the OBD-II port of a vehicle and extracts the signal from the vehicle speed sensor, the reference speed that a vehicle’s subsystems rely upon to function. The data logger is equipped with jamming detection, Bluetooth, a GPS, and a triaxial (X, Y, Z) accelerometer configured to detect hard cornering and impacts if the resultant force crosses a configured threshold.

DRIVES captures driving behavior, including latitude, longitude, speed, date, time, and event type at 30-second data collection intervals. Additional fields provide data regarding peak speed and average speed for a speeding event and initial and final speeds of braking or acceleration for events characterized by a rapid change in velocity. A summative activity report provides the date and start/end time, starting/ending locations, trip duration in seconds and distance, average and maximum vehicle speed, and the number/duration of specific driving events (eg, sudden acceleration, hard braking, over-speeding, hard cornering, and impacts) for all trips over a 24-hour period. Data collection is standardized across all 50 states. These data provide a critical window into in vivo driving behavior delineated by nonspatial frequency data (eg, counts of events and length of trip/duration) and spatial data (eg, driving area and unique destinations visited). When combined, both data types can inform a driver’s baseline behavior and how that behavior changes over time. By leveraging this technology, researchers can deeply phenotype a driver’s behavior with ecologically valid data extracted from that driver’s own vehicle as they drive around daily (Figure 2).

DRIVES is being used to examine daily driving behavior in a longitudinal study of over 300 adults over age 65 with and without preclinical AD (NIH/NIA R0AG056466, R01AG068183). Pilot studies using DRIVES demonstrated adults over age 65 with preclinical AD have decreased driving space, lower driving exposure (eg, miles, trips, or places visited), fewer trips with aggressive behavior, and a greater rate of decline over 2.5 years compared with those without preclinical AD.²¹

In symptomatic AD, hyperactivity or excess activation of the memory network (eg, hippocampus and medial temporal lobe) is thought to serve as internal compensation for degrading neuronal structures in order to retain normal performance on tasks.^22,23 When hippocampal hyperactivity is no longer supportive, a person may compensate by writing down notes, relying on a calendar, or depending on others to help with recall of planned events. Previous driving studies have established that a period of self-regulation emerges before eventual driving cessation in adults with cognitive impairments.²⁴ Drivers with mild AD self-report a tendency to drive fewer miles, make fewer trips, avoid driving at night or during inclement weather, chain trips together, avoid high speed or highway driving, and seldom speed, hard brake, or hard corner.¹⁸ The DRIVES data suggests that the self-regulation process may already begin in preclinical AD. The DRIVES has also been used in patients recovering with antibody-mediated encephalitis to monitor and model return to driving over a 1-year period.²⁵

Potential Clinical Utility of DRIVES

Our aging population and increasing prevalence of ADRD require a novel methodologic tool able to assess fitness to drive in a standardized manner irrespective of location. Naturalistic data can be a solution to conduct risk modeling for the likelihood of a crash, the decline in driving behavior, and support conversations between neurologists and patients about a nondriving future. We believe data loggers and programs like those used in DRIVES can be widely adopted and used for people who have concerns about their driving ability. Vehicle metrics pulled from driving behavior can be used to calculate a fitness-to-drive score integrating data points (eg, speed, hard braking, acceleration, and impacts) and spatial-temporal factors like road types, landmark density, street network entropy, time of day/night, and weather (Figure 2). Raw data can be captured from the driver’s vehicle each time they drive, and this data can be filtered through several software programs and analyzed using weights to provide the severity of driving behavior decline and the likelihood of a crash. The accuracy of the risk rating can likely be further improved by demographics, comorbid conditions (eg, stroke, epilepsy), self-reported driving habits, and performance on neuropsychologic tests. More importantly, data can be captured over any epoch so that change over time can be measured and potentially identify cognitive and or driving ability decline. This data-based approach will alleviate the onus from the neurologist to respond when there is a high crash risk, thus easing discussions with the patient by focusing on planning for a nondriving future. Finally, the ongoing COVID-19 pandemic has adversely impacted our healthcare system and diminished the volume of patients seen for neurology consultation even with the aid of telehealth platforms. The use of data loggers affords a contactless protocol because the device can be shipped and plugged in by the patient, and remote collection and offsite processing do not require any in-person interactions.