Abstract
Background: Chronic stress is a highly prevalent condition that may stem from different sources and can substantially impact physiology and behavior, potentially leading to impaired mental and physical health. Multiple physiological and behavioral lifestyle features can now be recorded unobtrusively in daily-life using wearable sensors. The aim of the current study was to identify a distinct set of physiological and behavioral lifestyle features that are associated with elevated levels of chronic stress across different stress sources. Methods: For that, 140 healthy female participants completed the Trier inventory for chronic stress (TICS) before wearing the Fitbit Charge3 sensor for seven consecutive days while maintaining their daily routine. Physiological and lifestyle features that were extracted from sensor data, alongside demographic features, were used to predict high versus low chronic stress with support vector machine classifiers, applying out-of-sample model testing. Results: The model achieved 79% classification accuracy for chronic stress from a social tension source. A mixture of physiological (resting heart-rate, heart-rate circadian characteristics), lifestyle (steps count, sleep onset and sleep regularity) and non-sensor demographic features (smoking status) contributed to this classification. Conclusion: As wearable technologies continue to rapidly evolve, integration of daily-life indicators could improve our understanding of chronic stress and its impact of physiology and behavior.
Original language | English |
---|---|
Journal | Chronic Stress |
Volume | 6 |
DOIs | |
State | Published - 2022 |
Bibliographical note
Publisher Copyright:© The Author(s) 2022.
Keywords
- chronic stress
- heart rate
- machine learning
- sleep
- social tension
- trier inventory for chronic stress
- wearable sensors
ASJC Scopus subject areas
- Clinical Psychology
- Psychiatry and Mental health
- Biological Psychiatry
- Behavioral Neuroscience