Predicting Heart Rate Variability from Heart Rate and Step Count for University Student Weekdays.

Area of Science:

• Wearable technology
• Biomedical engineering
• Machine learning applications

Background:

• University students experience high stress and anxiety.
• In-the-moment interventions can help manage stress.
• Wearable devices measuring heart rate variability (HRV) show promise for stress monitoring.

Purpose of the Study:

• To investigate the feasibility of estimating HRV from readily available data (heart rate and step count).
• To develop and evaluate machine learning models for HRV estimation using proxy variables.
• To assess the potential for low-cost wearable devices in stress management interventions.

Main Methods:

• Collected 201 hours of data from 14 university students during normal weekday activities.
• Utilized machine learning models trained on heart rate (HR) and step count to estimate HRV.
• Evaluated model performance in predicting highest-stress times (HRV below 10th percentile).

Main Results:

• Machine learning models achieved moderate accuracy in predicting highest-stress periods.
• Model performance indicated an Area Under the Curve (AUC) of 0.746.
• Achieved 60% sensitivity and 80% specificity in identifying high-stress instances.

Conclusions:

• Estimating HRV from HR and step count is a viable approach for stress monitoring.
• This method can enable large-scale, in-the-moment stress and anxiety interventions.
• Findings support the use of accessible wearable technology for mental well-being support in specific populations.