Frailty Identification Using Heart Rate Dynamics: A Deep Learning Approach.

Area of Science:

• Gerontology
• Cardiovascular Physiology
• Computational Medicine

Background:

• Frailty impacts autonomic nervous system function and heart rate response to physical activity in older adults.
• Previous studies often analyzed resting heart rate or lacked control for physical exertion.
• Understanding heart rate dynamics during activity is crucial for assessing frailty and maintaining homeostasis.

Purpose of the Study:

• To classify pre-frail/frail versus non-frail older adults using heart rate dynamics during physical activity.
• To investigate the efficacy of a long short-term memory model for frailty classification based on heart rate response.
• To establish an objective marker for frailty screening using physiological data.

Main Methods:

• Eighty-eight older adults (≥65 years) were divided into non-frail and pre-frail/frail groups based on the Fried frailty phenotype.
• Participants underwent normal speed walking while heart rate was continuously monitored using a wearable electrocardiogram recorder.
• A long short-term memory model was employed to analyze heart rate time series for frailty classification.

Main Results:

• The long short-term memory model achieved high classification performance in a 5-fold cross-validation.
• Key metrics included 83.0% sensitivity, 80.0% specificity, 87.0% F1-score, and 82.0% accuracy for distinguishing frailty classes.
• The model demonstrated effective classification without requiring manual feature engineering.

Conclusions:

• Heart rate dynamics during physical activity can serve as an accurate and objective indicator for frailty screening in older adults.
• Long short-term memory models offer a promising approach for analyzing complex physiological data for frailty assessment.
• This method may facilitate early detection and intervention for frailty, improving health outcomes in the elderly population.