We often write about aging-in-place and digital technologies that help older people maintain their independence. Driving is a major factor in senior mobility and self-esteem. However, according to the CDC, older drivers have higher crash death rates than middle-aged drivers, especially males and drivers aged 75 or older. The CDC also cites age-related vision impairment and cognition functioning as likely contributing factors to the rising risk of death while driving.

Researchers from the University of Tsukuba and IBM Research, Tokyo investigated whether cognitive impairment as measured by speech features could predict older drivers’ risk of being in vehicle accidents. The scientists published a report of their findings in the Journal of Medical Internet Research in April 2021.

In the study, the researchers collected cognitive assessment data and speech data from interactions with voice assistants from 71 people aged 61 to 80. Of the original group, 60 agreed to participate in a follow-up survey of their driving experiences about a year and a half later. Analysis of the survey data found that compared to drivers who had no car accidents or near-accidents in the previous year, the drivers who had car accidents or near-accidents had statistically significant differences in their speech. Differences included factors such as reduced speech rate and increased response time. Speech features could predict future accidents or near misses with 81.7% accuracy, compared to 75% accuracy with cognitive testing alone. Used together, speech features and cognitive assessment were 88.3% accurate in accident risk prediction.

This data is intriguing and the hypotheses bears further study. The test participants were all from Japan. They were screened to exclude people with serious disease, disabilities, mental illness, or neurodegenerative diseases. The study was not blind. Participants signed consent forms following full explanations of the study. It would be interesting to know the accident rate correlations with speech and cognitive assessments of those who did not consent to the follow-up survey. It would also be interesting to see studies of older people from various cultures and countries and with participants who didn’t know the purpose of the study ahead of time. Self-reported data is generally loose and stronger arguments could be made about the validity of the study results if the accident and near-accident data were collected by objective means such as accident reports and vehicle recorders. Finally, it will be extremely interesting if in the future IBM, Apple, Amazon, and Google — the Cortana, Siri, Alexa, and Google voice assistant developers — begin to sell car insurance to older drivers who converse with their respective voice assistant.

“I’m sorry, Dave. I can no longer allow you to drive your car…”