Interpretable classifiers for prediction of disability trajectories using a nationwide longitudinal database.

Area of Science:

• Gerontology
• Biostatistics
• Artificial Intelligence in Healthcare

Background:

• Understanding long-term disability trajectories is crucial for the elderly population.
• Heterogeneity in disability progression necessitates advanced predictive modeling.
• Chinese elderly face unique demographic and health challenges impacting disability.

Purpose of the Study:

• To explore heterogeneous disability trajectories in Chinese elderly.
• To construct explainable machine learning models for predicting long-term disability.
• To understand the mechanisms behind disability prediction models.

Main Methods:

• Retrospective data from the Chinese Longitudinal Healthy Longevity and Happy Family Study (2002-2018).
• Included 4149 subjects aged 65+ with longitudinal activities of daily living (ADL) data.
• Utilized mixed growth models and five machine learning models with explainable AI.

Main Results:

• Identified three disability trajectories: normal (77.3%), progressive (15.5%), and high-onset (7.2%).
• Machine learning models, particularly random forest and extreme gradient boosting, showed strong predictive performance.
• Key predictors included ADL, age, leisure activity, cognitive function, and blood pressure.

Conclusions:

• Machine learning demonstrates high efficacy in analyzing quality indicators for disability trajectory prediction.
• Findings support the use of ML for personalized intervention strategies in elderly care.
• Explainable AI provides insights into the factors driving disability predictions.