Development and validation of machine learning models to predict frailty risk for elderly

Area of Science:

• Gerontology
• Artificial Intelligence
• Public Health

Background:

• Early identification of frailty in the elderly is crucial for reducing healthcare burdens and enhancing quality of life.
• Frailty is a significant concern in aging populations, necessitating effective predictive tools.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting frailty risk in the elderly.
• To identify key risk factors associated with frailty in older adults.

Main Methods:

• A prospective cohort study utilizing data from the Chinese Longitudinal Healthy Longevity Study (waves 6-7).
• Four ML algorithms (random forest, support vector machine, XGBoost, logistic regression) were employed to predict frailty, assessed using the osteoporotic fractures index.
• Models were trained and internally validated on a cohort of 4385 individuals, with external validation on 6997 individuals.

Main Results:

• Random forest (RF) and logistic regression (LR) models demonstrated strong predictive performance.
• The RF model achieved the highest Area Under the Curve (AUC) of 0.75, closely followed by LR (0.74).
• Final models showed similar AUC values (0.77 for RF, 0.76 for LR) and identical accuracy (87.4%).

Conclusions:

• Preliminary ML-based prediction models for elderly frailty risk have been successfully developed.
• These models can assist healthcare providers in predicting frailty probability and guiding interventions for older adults.