Development of a falls risk assessment scale for middle-aged and elderly patients with cardiovascular diseases in...

Falls among middle-aged and elderly patients with cardiovascular diseases (CVD) frequently occur in non-hospital settings. However, there is a lack of simple forecasting tools for fall risk in middle-aged and elderly patients with CVD in Chinese community. We aimed to develop an easy-to-use scoring system, even for non-professionals, to facilitate the implementation of individualized fall prevention measures.

Nationwide cohort study.

The study included 2280 patients over the age of 45 with CVD from the China Health and Retirement Longitudinal Study (CHARLS). The data was divided into training (70 %) and testing (30 %) sets. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used to refine 134 candidate variables in the training set. After that, six machine learning algorithms were employed to construct predictive models for post-diagnosis fall risk. AutoScore, an interpretable machine learning-based clinical score generation algorithm, was then utilized to establish a fall risk prediction score. Finally, the models' performance was evaluated using Area Under the Curve (AUC), Accuracy, Sensitivity, Specificity, and Brier Score.

LASSO regression identified 16 predictive variables for input into various models. In the test set, AutoScore achieved the highest AUC (0.7447) among the six machine learning models, showing statistically significant differences with LightGBM, Random Forest, and XGBoost. Ultimately, a percentage-based questionnaire scale was developed by used AutoScore, including variables such as physical condition, psychological state, and surrounding environment, among others. The optimal threshold for predicting falls was determined to be 31 points.

We have developed a practical and easily interpretable fall prediction score for middle-aged and elderly patients with CVD in community.