Sleep-phasic heart rate variability predicts stress severity: Building a machine learning-based stress prediction...

Area of Science:

• Biomedical Engineering
• Sleep Science
• Cardiovascular Physiology

Background:

• Stress significantly impacts health and well-being.
• Accurate, accessible stress monitoring is crucial for early intervention.
• Current stress assessment methods can be invasive or lack continuous monitoring.

Purpose of the Study:

• To develop and validate a novel method for predicting stress severity using smart device-measured sleep phasic heart rate variability (HRV).
• To assess the feasibility of using smart devices for large-scale stress self-screening.
• To compare HRV indices during different sleep stages (CAP, NCAP, REM) for stress prediction.

Main Methods:

• Dual 24-h recordings using Holter ECG and a Huawei smart device in 159 medical workers.
• Sleep stage classification (CAP, NCAP, REM, wakefulness) based on smart device PPG and accelerometer signals using cardiopulmonary coupling (CPC) algorithms.
• Calculation of HRV indices from both Holter ECG and smart device PPG signals during specific sleep stages.
• Development of a machine learning model to predict stress severity using only smart device data.

Main Results:

• Sleep phasic HRV indices effectively predict individual stress severity, with higher predictive performance during Cyclic Alternating Pattern (CAP) or Rapid Eye Movement (REM) sleep compared to Non-Cyclic Alternating Pattern (NCAP) sleep.
• The optimal machine learning model, using only smart device data, achieved 80.3% accuracy, 87.2% sensitivity, and 63.9% specificity for stress prediction.
• Smart device-derived HRV measurements are comparable to Holter ECG for stress prediction.

Conclusions:

• Sleep phasic heart rate variability, measurable via smart devices, is a viable indicator for stress prediction.
• Smart devices offer a promising, non-invasive tool for continuous stress monitoring and self-screening in the general population.
• This approach can facilitate early detection and management of stress-related conditions.