Development of a web platform for predicting fall risk in cardiovascular patients using machine learning.

Area of Science:

• Gerontology
• Cardiology
• Artificial Intelligence

Background:

• Falls are a significant concern for middle-aged and elderly individuals, particularly those with cardiovascular disease (CVD).
• Predicting fall risk is crucial for developing effective prevention strategies and improving patient outcomes.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based model for predicting fall risk in middle-aged and elderly individuals with CVD.
• To identify key predictors of falls in this population and create an interpretable and accurate predictive tool.

Main Methods:

• Utilized data from the China Health and Retirement Longitudinal Study (CHARLS).
• Employed Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify predictive variables.
• Constructed and evaluated six ML algorithms, including Light Gradient Boosting Machine (LGBM), for fall risk prediction.
• Performed internal and external validation and used SHapley Additive exPlanation (SHAP) for model interpretability.

Main Results:

• Analyzed 1784 participants, with 24.3% experiencing falls over two years.
• Identified nine key predictive factors for falls.
• The LGBM model achieved an area under the receiver operating characteristic curve (AUC) of 0.839 (internal validation) and 0.816 (external validation).

Conclusions:

• The developed ML model, particularly LGBM, demonstrates high accuracy in predicting fall risk among CVD patients.
• This model offers a scientific foundation for personalized fall prevention, aiming to decrease fall incidence and enhance patient quality of life.
• A web platform was created based on the best model to facilitate practical application in clinical settings.