Utility of wearable technology in predicting panic attacks: A scoping review

Area of Science:

• Wearable technology
• Machine learning
• Psychophysiology

Background:

• Panic attacks (PAs) significantly impact global health, causing unpredictable symptoms like rapid heartbeat and trembling.
• Wearable technology and machine learning (ML) show potential for managing health conditions, but PA prediction remains underexplored.

Purpose of the Study:

• To review methodologies, features, and ML models used in wearable-based panic attack (PA) research.
• To assess the current state and limitations of PA prediction using wearable devices.

Main Methods:

• A systematic scoping review of seven studies identified through major academic databases (PubMed, PsycINFO, Embase, Google Scholar).
• Analysis of diverse ML models including deep learning (LSTM, RNN), random forests, and regression models.
• Examination of physiological metrics such as heart rate variability and activity levels.

Main Results:

• Studies utilized various ML models and analyzed physiological data, achieving predictive accuracies ranging from 67.4% to 94.8%.
• Key predictors identified include resting heart rate, heart rate variability, and sleep metrics.
• Integration of psychological, physiological, and environmental data enhances prediction accuracy.

Conclusions:

• Wearable sensors and ML show utility in identifying panic attack predictors.
• Current prediction time frames are often impractical, with limited success in near-real-time PA onset prediction.
• Further research is essential to develop effective real-time panic attack prediction systems.