Voice recordings from mobile memory tests allowed investigators to predict subjective sleepiness in older adults with strong accuracy, suggesting a practical screening tool outside the clinic.
Key takeaways:
- Researchers found that verbal reaction time correlates with self-reported sleepiness in adults aged 55 and older.
- A computer model predicted sleepiness based on voice recordings with an F1-score of 0.80.
- The method offers a nonintrusive, scalable way to monitor alertness in at-risk populations, such as those taking benzodiazepine receptor agonists.
- Future applications could include integration into smartphones and telehealth platforms to monitor medication effects.
A new study led by investigators at the University of California, Los Angeles (UCLA) indicates that verbal reaction time—the amount of time it takes a person to respond verbally—can serve as a marker of sleepiness in older adults.
Published in Sleep Science and Practice, the research demonstrates how measuring voice data through standardized cognitive assessments can passively detect excessive sleepiness, specifically among older individuals using sedative medications.
Sleepiness risks are particularly acute in older adults using sedative medications like benzodiazepine receptor agonists (BZRAs). While current methods to assess sleepiness can be intrusive or impractical for real-world application, this study suggests a scalable alternative to identify at-risk individuals before health declines or accidents occur.
The researchers studied adults aged 55 and older with a history of insomnia and BZRA use, recruiting participants from a deprescribing clinical trial. Participants completed memory tests via a mobile app that recorded their verbal responses. The research team measured verbal reaction time, defined as the delay between the start of the recording and the first spoken word, and compared it to participants’ self-reported sleepiness.
According to the findings, the computer model successfully predicted participants’ self-reported sleepiness based on their voice recordings. Individuals who took longer to start speaking after a prompt also reported feeling sleepier. The model identified different levels of sleepiness with an F1-score of 0.80 ± 0.08. Furthermore, the voice analysis method detected when someone was speaking versus silent with 92.5% accuracy.
“This study shows that something as simple as how quickly someone starts speaking can tell us a lot about their level of alertness,” says Dr Tue T. Te, lead author of the study and a researcher at the David Geffen School of Medicine at UCLA and the VA Greater Los Angeles Healthcare System, in a release. “It opens the door to using voice as a passive, scalable tool for monitoring sleepiness during everyday activities.”
The results suggest that voice timing could provide a low-effort method to monitor sleepiness outside of clinical settings. The research team plans to expand this approach to larger, more diverse populations and explore integration into technologies such as smartphones and telehealth platforms. Future research may also investigate how voice-based markers can monitor medication effects or detect early signs of cognitive decline.
