Real-world driving digital data reliably index MCI over time (S15.008)

Abstract

<h3>Objective:</h3> Our objective is to detect age-related cognitive decline from driver behavior. The overarching goal is to develop real-world, digital biomarkers of early dementia, including Alzheimer’s disease (AD), to inform clinical care and intervention. <h3>Background:</h3> Cognitive decline affects driving abilities. Driver behavior patterns, in turn, index cognitive abilities. Roadway environments present varying challenges, revealing driver strategies accommodating challenges. Strategies map to cognitive decline, even in transitional dementia stages like mild cognitive impairment (MCI). <h3>Design/Methods:</h3> Real-world driving data (249,104 miles) were collected for 2, 3-month periods across 2 years (baseline/longitudinal) using sensors installed in participants’ vehicles. 74 participants (mean age = 75.1, 44 females) self-reported demographics and completed neuropsychological assessments relevant to aging, driving, and AD. Neuropsychological data were used to classify MCI at baseline (MCI: N = 14; Peterson, 2004). Mixed-effect linear regressions assessed changes in speed limit compliance (difference between vehicle speed and posted speed limit) across MCI class and roadway environment (residential/commercial/interstate roads [20–30/35–45/≥ 55 mph]). <h3>Results:</h3> Drivers with MCI drove further below the posted speed limit compared to controls (<i>b</i> = −0.235, <i>p</i> &lt; 0.001). Individual driver speed patterns correlated at baseline and longitudinal (r = 0.63). Commercial (<i>b</i> = −0.541, <i>p</i> &lt; 0.001) and residential (<i>b</i> = −0.647, <i>p</i> &lt; 0.001) roadways showed ability to detect differences in speed limit compliance based on driver MCI class. High-speed driving (interstate roadways) revealed greater differences in speed limit compliance for MCI drivers compared to those without MCI. <h3>Conclusions:</h3> Results demonstrate feasibility to detect early dementia stages, like MCI, from real-world, digital driving sensor data. Findings reveal key environments that show discriminative MCI utility. Data show longitudinal reliability across individuals suggesting utility of data to predict decline progression for AD intervention. Sensor-based digital health biomarkers hold promise to deliver clinicians timely, objective data on patient health and risk. <b>Disclosure:</b> Dr. Chang has nothing to disclose. Mr. Wu has nothing to disclose. The institution of Prof. Zhang has received research support from National Institute of Health. Prof. Zhang has received personal compensation in the range of $5,000-$9,999 for serving as a Consultant with Nationwide Children Hospital. The institution of Dr. Rizzo has received research support from NIH. Dr. Rizzo has a non-compensated relationship as a Chair with ABC that is relevant to AAN interests or activities. Jennifer Merickel has nothing to disclose.