Development and Validation of Machine Learning Models for Predicting Early Cognitive Decline Using Home...

Area of Science:

• Gerontology and Digital Health
• Biomarker Discovery
• Predictive Analytics in Healthcare

Background:

• Increasing global population aging leads to higher prevalence of dementia and frailty.
• Early identification of mild cognitive impairment (MCI) or prefrailty is crucial for intervention.
• Digital sensor technology and predictive modeling show promise for early detection.

Purpose of the Study:

• To distinguish normal aging from MCI, dementia, prefrailty, or frailty in older adults using home sensors.
• To predict transitions from normal aging to these conditions.
• To develop a continuous, home-based monitoring system for early detection.

Main Methods:

• Longitudinal cohort study of 200+ adults aged ≥65 years.
• Multi-sensor system (motion, door, bed, medication, wearables) capturing spatiotemporal activity, mobility, sleep, and medication adherence.
• Annual clinical assessments including cognitive tests, frailty measures, and psychosocial indicators.
• Supervised machine learning models (logistic regression, random forests, gradient boosting, deep learning) integrating sensor and clinical data.

Main Results:

• Enrollment ongoing (138/200 participants as of June 2025).
• Full data analysis has not yet commenced.
• Study aims to establish model performance using cross-validation and independent test sets.

Conclusions:

• Development of a reliable in-home sensor system for early detection of cognitive and physical decline.
• Contribution to understanding digital biomarkers for aging-related conditions.
• Potential for prompt intervention to delay or reverse disease progression, enhancing functional years.