Detecting Prolonged Stress in Real Life Using Wearable Biosensors and Ecological Momentary Assessments: Naturalistic...

Area of Science:

• Psychophysiology
• Digital Health
• Mental Health Monitoring

Background:

• Growing need for unobtrusive monitoring of stress-related mental disorders.
• Wearable devices show promise for real-life stress detection, but systematic investigation is lacking.

Purpose of the Study:

• To determine the utility of ecological momentary assessments (EMA) and physiological arousal from wearable devices for detecting stress states.
• To compare the effectiveness of EMA, physiological data, and their combination in stress detection.

Main Methods:

• Ecological momentary assessments (EMA) and ecological physiological assessments (EPA) using wearable biosensors were employed during an examination week and a control week.
• Generalized linear mixed-effects models and machine learning were used to analyze subjective stress, mood, and physiological arousal.
• Individualized and group-based machine learning models were tested for classifying stress periods.

Main Results:

• Participants reported increased negative affect and decreased positive affect during examination weeks, with decreased average physiological arousal.
• Individualized EMA showed better classification than EPA alone, but the combination of EMA and EPA yielded optimal classification accuracy.
• Individualized models significantly outperformed group-based models for all data inputs.

Conclusions:

• Wearable biosensors have potential for stress monitoring, but psychological context is crucial for interpreting physiological arousal.
• A personalized approach, referencing momentary stress against an individual's own data, is optimal for detection.
• Combining EMA with wearable biosensor data enhances the accuracy of real-life stress detection.