Examination of Social Participation in Older Adults Undergoing Frailty Health Checkups Using Deep Learning Models.

Area of Science:

• Gerontology
• Artificial Intelligence
• Public Health

Background:

• Social participation is crucial for older adults' well-being but is often limited by frailty.
• Frailty health checkups present an opportunity to identify individuals at risk of reduced social engagement.

Purpose of the Study:

• To predict social participation in older adults using machine learning (ML) models.
• To identify key predictive factors for social participation using deep neural network (DNN) analysis.

Main Methods:

• Utilized logistic regression (LR), nonlinear support vector machine (NLSVM), and DNN models.
• Evaluated models using precision, accuracy, sensitivity, specificity, F1 score, and area under the curve (AUC).
• Analyzed 18 attributes including demographic, physical, cognitive, and social factors from 295 older adults.

Main Results:

• ML models demonstrated moderate discriminative performance (AUC 0.776-0.795).
• Deep neural network (DNN) offered the most balanced performance, excelling in sensitivity.
• Information-collection ability, walking speed, and number of cohabitants were significant predictors.

Conclusions:

• Machine learning models can moderately predict social participation in frailty-screened older adults.
• Model selection should align with specific community health screening objectives.
• Information-collection ability is a critical factor influencing social participation in this demographic.