Development of an interpretable machine learning model-based online tool for risk identification of anxiety symptoms in Chinese older adults.

BACKGROUND: As the population aging process has accelerated, anxiety symptoms (AS) among older adults have become a critical concern. In addition to developing an online computational tool to identify high-risk persons in community settings, this study intends to construct and verify an interpretable machine learning (ML) model for identifying AS in older adults. METHODS: In accordance with the health ecology model, 38 predictor variables were extracted from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) 2018 database for this study. The dataset was randomly allocated into training (70%) and testing (30%) subsets. Five ML methods-Random Forest (RF), XGBoost, LightGBM, Decision Tree, and Logistic Regression-were utilised to identify important predictive variables and develop predictive models. Optimal hyperparameters were determined and validated. Ultimately, the top-performing model serves as the foundation for an online risk prediction tool. RESULTS: Among 9535 participants, the prevalence of AS was 11.89%. Sleep quality, self-reported health, depressive symptoms, vegetable, ventilation method, and economic situation were important risk factors for AS in older adults. The RF model performed optimally in predictive analysis, achieving an Area Under the Curve of 0.790 in the test set, with corresponding accuracy, sensitivity, and specificity values of 0.729, 0.712, and 0.747, respectively. The online tool implementing this model is available at https://onlineriskpredictionmodel.shinyapps.io/anxiety_prediction_tool/. CONCLUSION: The application of ML methods to predict AS in older adults and develop an online tool facilitates early identification of high-risk groups and enables precise interventions, thereby providing scientific guidance for formulating public health policies and advancing the achievement of healthy aging.