Predicting restriction of life-space mobility: a machine learning analysis of the IMIAS study

Area of Science:

• Gerontology
• Artificial Intelligence
• Public Health

Background:

• Machine learning (ML) is increasingly explored for mobility prediction in older adults.
• ML offers potential for analyzing complex life-space mobility (LSM) data.

Purpose of the Study:

• To assess the predictive accuracy of ML algorithms for LSM restriction in the elderly.
• To identify significant risk factors contributing to LSM reduction.

Main Methods:

• Developed a 2-year LSM reduction prediction model using decision tree, random forest, and eXtreme gradient boosting (XGBoost).
• Validated the model on an independent cohort from the International Mobility in Aging Study (IMIAS) (n=372, age ≥65).
• Life-Space Assessment (LSA) questionnaire measured LSM across five levels.

Main Results:

• The XGBoost model demonstrated strong predictive performance with MAE of 10.28 and RMSE of 12.91.
• Identified key predictors: frailty (39.4%), mobility disability (25.4%), depression (21.9%), and female sex (13.3%).

Conclusions:

• ML algorithms effectively predict LSM restriction, identifying high-risk older adults for proactive interventions.
• The XGBoost model provides valuable insights into mobility complexities, complementing traditional statistical methods.