Indoor Mobility Patterns Measured by PIR Sensors for Classifying Health-Related Quality of Life in Older Adults Using...

Area of Science:

• Gerontology
• Biomedical Engineering
• Health Informatics

Background:

• Older adults living alone require continuous monitoring for health management.
• Passive infrared (PIR) sensors offer a privacy-preserving method for activity tracking.
• Health-related quality of life (HRQoL) is a key indicator of well-being in aging populations.

Purpose of the Study:

• To investigate if indoor mobility features from PIR sensors can differentiate HRQoL levels in community-dwelling older adults.
• To evaluate the effectiveness of machine learning models in predicting HRQoL based on PIR sensor data.

Main Methods:

• Extracted mobility variables from three months of PIR sensor data for 40 older adults.
• Classified participants into high- and low-HRQoL groups using the EQ-5D index.
• Employed a nested stratified five-fold cross-validation with RandomOverSampler for class imbalance.
• Evaluated Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbors (KNN) classifiers.

Main Results:

• SVM achieved the highest performance: accuracy 0.825, precision 0.860, recall 0.900, F1-score 0.876, and AUC 0.937.
• Random Forest demonstrated comparable performance with an AUC of 0.933.
• KNN showed lower performance and higher variability.

Conclusions:

• PIR-derived mobility features combined with machine learning models show feasibility for assessing HRQoL in older adults.
• This approach offers a low-cost, privacy-preserving method for health monitoring.
• Further validation in larger, diverse cohorts is warranted.