AI can spot depression through driving habits, study finds

New research suggests that the way older adults drive could offer subtle clues about their mental health, potentially paving the way for innovative screening tools for depression. Two related studies, the latest published in the journal npj Digital Medicine, investigated the connection between driving patterns and depression in older individuals. The findings indicate that changes in driving behavior, detectable through everyday technology, could be analyzed by artificial intelligence to identify older adults who may be experiencing depression, a condition that often goes unnoticed in this age group.

“Older adults are driving longer, yet we know little about how mental health conditions like major depressive disorder (MDD) impact real-world driving behavior,” said study author Ganesh M. Babulal, an associate professor at Washington University in St. Louis and director of the DRIVES Project.

“Given that depression is associated with cognitive and motor impairments, we wanted to examine whether older drivers with MDD exhibit riskier driving patterns that could affect safety and independence. Understanding these patterns is crucial for developing interventions that help older adults maintain mobility while minimizing risk.”

The researchers first conducted a study to understand how depression might affect the actual driving behavior of older adults in their everyday lives. This initial study involved 395 participants, some diagnosed with major depressive disorder and others without. All participants were at least 65 years old and were part of a larger, ongoing project examining aging, driving, and depression at Washington University in St. Louis.

The group with depression included 85 individuals who had been diagnosed with major depressive disorder by a clinician, or who scored high on a depression questionnaire called the Patient Health Questionnaire-9 (PHQ-9). The control group consisted of 310 individuals without a diagnosis of depression.

Participants underwent clinical assessments, including the Clinical Dementia Rating (CDR) to evaluate cognitive function, and completed neuropsychological tests to measure different aspects of thinking and memory. Crucially, all participants had a small device placed in their personal vehicles. This device, connected to the car’s computer system, used GPS technology to continuously record their driving behavior over time, capturing data such as speed, location, and events like sudden braking or sharp turns. The researchers collected this driving data for a period of time and then analyzed it to see if there were noticeable differences in the driving patterns of those with and without depression.

The first study revealed that older adults with depression exhibited distinct driving patterns compared to those without depression. Even though there were no significant differences in their cognitive test scores at the start of the study, those with depression demonstrated riskier driving behaviors over time. Specifically, they had more instances of hard braking and hard cornering during their trips. They also showed less predictable driving routes and traveled to a wider range of destinations.

Interestingly, at the beginning of the study, the group with depression showed a tendency for speeding and spent more time driving overall. These findings suggested that depression is indeed linked to changes in driving behavior that could potentially increase the risk of accidents. Importantly, these differences in driving behavior were observed even after considering the use of medications that could impact driving, indicating that depression itself was a significant factor.

“We expected some impact of MDD on driving, but we were surprised by the extent of risky behaviors, including increased driving distances and unpredictability in destinations,” Babulal told PsyPost. “Rather than self-regulating their driving, as many older adults do when experiencing cognitive or physical decline, those with MDD continued riskier driving patterns over time. This suggests that depression may impair awareness of functional changes, which has important safety implications.”

Building on these findings, the researchers then conducted a second study to investigate whether artificial intelligence could be used to automatically detect depression in older adults based on their driving data. This follow-up study used machine learning, a type of artificial intelligence that allows computers to learn from data without explicit programming. For this study, they analyzed two years of driving data from 157 older adults, including 81 diagnosed with major depressive disorder and 76 without. The participants were recruited from the same ongoing research project and met similar criteria as in the first study. The driving data was collected using the same GPS devices in their cars.

In addition to driving data, the researchers also collected information on demographics, such as age, gender, and education level, and medication use, including whether participants were taking antidepressants and the total number of different medications they used. They then trained machine learning models, specifically using a method called Extreme Gradient Boosting (XGBoost) and also logistic regression, to see if these models could learn to distinguish between depressed and non-depressed older adults based on their driving data and other information. They tested various models, some using only driving features, some adding demographic information, and others incorporating medication data.

The machine learning models were able to identify depression with a good degree of accuracy based on driving data alone. The best performing model, which combined driving features with information on the total number of medications participants were taking, achieved a high level of accuracy in distinguishing between depressed and non-depressed individuals. This top model correctly identified individuals with depression in 90% of cases and accurately identified those without depression in 82% of cases.

The driving features that were most important for identifying depression included the rate of hard cornering and hard braking, as well as the number of trips of different lengths. Surprisingly, adding demographic information like age, gender, and education did not improve the performance of the models and, in some cases, even made them slightly less accurate. This suggests that driving behavior and medication use are stronger indicators of depression in older adults than demographic factors alone.

“Depression in older adults is not just about mood—it can also affect complex daily behaviors like driving,” Babulal said. “This highlights the importance of routine depression screening and targeted interventions to promote safer driving habits while supporting older adults’ independence.”

Looking ahead, the researchers suggest that future studies should include larger groups of participants and follow them for longer periods. They also recommend incorporating more detailed health information, such as electronic health records, and exploring more advanced artificial intelligence techniques to further refine the accuracy of depression detection. Future research could also investigate how to better ensure that driving data truly reflects the behavior of the intended participant.

“While our study provides valuable insights, it does not establish causality—MDD is associated with changes in driving behavior, but we cannot conclude that depression directly causes these changes,” Babulal noted. “Additionally, our sample was predominantly non-Hispanic White, so additional research is needed to explore these effects in more diverse, representative populations. We also did not track changes in depression symptoms over time, which could influence driving behaviors.”

“We aim to refine strategies for identifying older adults at risk for unsafe driving due to mental health conditions like MDD. Future research will explore whether interventions—such as cognitive training, medication management, or driving modifications—can help mitigate these risks. Ultimately, we hope to develop clinical and policy recommendations that balance safety with the need for mobility and independence.”

“Our findings emphasize the need for a comprehensive approach to older driver safety, including mental health assessment,” Babulal added. “Depression is treatable, and addressing it proactively could improve both driving safety and overall well-being. We encourage older adults and their families to discuss driving concerns with healthcare providers and explore resources for safer mobility.”

The study, “Identifying major depressive disorder in older adults through naturalistic driving behaviors and machine learning,” was authored by Chen Chen, David C. Brown, Noor Al-Hammadi, Sayeh Bayat, Anne Dickerson, Brenda Vrkljan, Matthew Blake, Yiqi Zhu, Jean-Francois Trani, Eric J. Lenze, David B. Carr, and Ganesh M. Babulal.