A Sensor-Driven Visit Detection System in Older Adults' Homes: Towards Digital Late-Life Depression Marker Extraction

Area of Science:

• Gerontology
• Digital Health
• Mental Health Technology

Background:

• Late-life depression is a prevalent mental health disorder in older adults, often underdiagnosed due to stigma and misconceptions.
• Social isolation, exacerbated by a lack of home visits, is a significant risk factor for late-life depression.
• Sensor technology offers potential for objective health monitoring and digital measures in older adults.

Purpose of the Study:

• To develop a robust home visit detection system using sensor data that generalizes to diverse living environments.
• To evaluate the system's performance in identifying home visits as a proxy for social isolation.
• To assess the correlation between sensor-derived home visit data and established depression screening tools.

Main Methods:

• A self-training-based domain adaptation strategy was employed to build a generalized visit detection system.
• The system was trained on semi-annotated data and tested for its ability to generalize to unseen apartments.
• The extracted home visit information was correlated with the Geriatric Depression Scale (GDS) screening tool.

Main Results:

• The developed visit detection system achieved a robust performance with an ROC AUC of 0.773.
• The sensor-derived home visit data showed a strong negative correlation with the Geriatric Depression Scale (ρ = -0.87, p = 0.001).
• The findings support the utility of home visit detection as a digital measure for social isolation.

Conclusions:

• Sensor-based home visit detection can serve as a valuable digital measure for social isolation in older adults.
• This approach shows promise as a digital biomarker for monitoring the risk of late-life depression.
• The technology offers a non-intrusive method to support mental health monitoring in the growing elderly population.