How You Drive Can Predict Dementia

Driving is a complex task involving multiple cognitive processes. It's also an essential part of people's lives, so it's surprising how little is known about how changes in the brain that come with aging affect our ability to drive.

But now, using artificial intelligence and machine learning techniques (data analytics), researchers say they can predict which drivers will go on to develop dementia.

Driving behaviors also change with early-stage dementia. These include taking fewer trips, not driving too far from home, driving less on the freeway, and becoming unwilling to drive at night or in rush hour traffic. In recent years, scientists have taken an interest in how shifts within the brain during aging impact driving performance. For example, people with early-stage dementia report getting lost in traffic and becoming less able to navigate their way around their city.

Few scientists are involved in this area of research, but Catherine Roe, Associate Professor of Neurology at Washington University School of Medicine, is one of them. She and her colleagues have conducted a series of studies in recent years and have made some important revelations.

A Very Early Indicator of Dementia

They've discovered that the greater the burden of amyloid plaques and tau tangles - widely regarded as symptoms of Alzheimer's disease - the greater the number of driving errors. These tissue pathologies also predicted the length of time before the participants' driving test was rated as marginal or a failure.

"Alzheimer’s disease affects more than just thinking and memory," she said. "It affects your reaction time, eyesight, strength, gait and mood. All of these factors also are related to driving, so it’s not unreasonable to suppose that you might see some effect on driving even before cognitive symptoms are apparent.

"If we could use driving behaviors to help us figure out who might have underlying Alzheimer’s, that would be an economical, naturalistic way to identify people with mild symptoms or even pre-symptomatic people."

Scientists from Columbia University believe they have achieved just that.


Algorithms Are 88 Percent Accurate

The Columbia team constructed 29 variables using naturalistic driving data. This refers to data captured through in-vehicle recording devices of real people driving cars, not simulations of people driving.

They also developed a statistical technique called “random forest models” for classifying disease status, and in addition trained a series of machine learning models for detecting mild cognitive impairment (MCI) and dementia.

In their study that was recently published in the journal Geriatrics, researchers placed devices in the vehicles of 2,977 active drivers aged 65 to 79 who were cognitively healthy and free from any degenerative medical condition. Researchers also noted specific demographic information such as their race or ethnicity, gender and education level.

Next, researchers tracked participants' driving habits for 45 months, during which time 33 participants were diagnosed with MCI and 31 with dementia. When researchers used demographic information to predict who would develop MCI and dementia, they had a 29 percent accuracy rate. However, when using driving variables, the accuracy rate more than doubled, increasing to 66 percent. Even more exciting, by combining the demographic and driving data, researchers were able to predict MCI and dementia with 88 percent accuracy.

Senior author and professor of epidemiology and anesthesiology, Guohua Li commented, "Our study indicates that naturalistic driving behaviors can be used as comprehensive and reliable markers for MCI and dementia.

“If validated, the algorithms developed in this study could provide a novel, unobtrusive screening tool for early detection and management of MCI and dementia in older drivers."

Researchers hope that these algorithms can eventually be incorporated into a smartphone app so people can assess their own risk for Alzheimer’s or dementia.

Leave A Comment

Please note, comments must be approved before they are published