Integrating Heart Rate Variability Improves Machine Learning-based Prediction of Panic Disorder Symptom Severity

Area of Science:

• Cardiology
• Psychiatry
• Machine Learning
• Autonomic Nervous System Research

Background:

• Panic disorder (PD) is often linked to autonomic nervous system imbalances, with heart rate variability (HRV) being a key indicator.
• Previous research has explored the association between PD and HRV, but predicting PD severity remains challenging.

Purpose of the Study:

• To evaluate the predictive capability of HRV, in conjunction with psychometric scales, for determining panic disorder severity using machine learning.
• To compare the efficacy of different feature sets (HRV alone, scales alone, and combined) in predicting PD severity.

Main Methods:

• 507 patients diagnosed with PD were recruited, providing data on psychometric scales and HRV components.
• Three machine learning experiments were conducted using different input feature combinations: scales and HRV (ExSH), scales only (ExS), and HRV only.
• Nine machine learning models were applied in each experiment, with performance metrics analyzed and SHapley Additive exPlanation (SHAP) used for feature importance assessment.

Main Results:

• The Random Forest model utilizing both psychometric scales and HRV (ExSH) achieved the highest performance, with an f1-score of 76.50% and sensitivity of 75.35%.
• ExSH demonstrated significantly superior sensitivity and f1-score compared to the scales-only approach (ExS).
• For the Random Forest model in ExSH, the Hamilton Rating Scale for Anxiety, Hamilton Depression Rating Scale, and low-frequency (LF) power of HRV were identified as the most important predictive features.

Conclusions:

• Combining HRV measures with traditional psychometric scales enhances the accuracy of machine learning models in predicting panic disorder severity.
• Low-frequency (LF) power of HRV is identified as a significant and promising biomarker for predicting PD severity.