Burnout Risk Prediction through Wearable Devices: An Initial Assessment.

Area of Science:

• Digital Health
• Machine Learning in Healthcare
• Occupational Health

Background:

• Burnout syndrome poses significant health risks, often diagnosed late due to reliance on self-reported questionnaires.
• Wearable devices offer continuous, unobtrusive data collection for early detection of mental health conditions like burnout.
• Machine learning can potentially analyze physiological data for early burnout risk assessment.

Purpose of the Study:

• To investigate the machine learning-based prediction of baseline burnout risk using physiological data from wearable devices.
• To assess prediction performance across cognitive, emotional, and physical burnout dimensions.
• To identify key physiological features for burnout risk detection.

Main Methods:

• Utilized data from 239 participants over the initial 30 days of a 9-month longitudinal study.
• Employed aggregated mean and standard deviation of physiological features (sleep, cardiac, stress) from smartwatches.
• Applied machine learning models to predict baseline burnout risk across cognitive, emotional, and physical dimensions.

Main Results:

• Models achieved balanced accuracies of 0.66 for cognitive weariness and 0.68 for physical fatigue risk.
• Prediction performance for emotional exhaustion risk was lower (0.55 balanced accuracy).
• Sleep, cardiac, and stress-related physiological features were key predictors.

Conclusions:

• Wearable device data and machine learning show potential for early detection of cognitive and physical burnout symptoms.
• Emotional exhaustion prediction requires further improvement, possibly through integration of additional data sources.
• Future work will focus on feature engineering and longitudinal data analysis to enhance prediction accuracy.