A risk prediction system for depression in middle-aged and older adults grounded in machine learning and visualization technology: a cohort study.

Introduction: Middle-aged and older adults are highly susceptible to depression. For this reason, early identification and intervention can substantially reduce its prevalence. This study innovatively proposed a visual risk prediction system for depressive symptoms and depression in middle-aged and older adults, rooted in machine learning and visualization technologies. Methods: Using cohort data from the China Health and Retirement Longitudinal Study (CHARLS), involving 8,839 middle-aged and older adult participants, the study developed predictive models based on eight machine learning algorithms, primarily including LightGBM, XGBoost, and AdaBoost. To enhance the interpretability of the XGBoost model, SHAP technology was employed to visualize the prediction results. The model was then deployed on a web platform to establish the risk prediction system. Results: Among the models, XGBoost demonstrated the best performance, achieving an average ROC-AUC of 0.69, and was ultimately selected as the predictive model for depressive symptoms and depression risk in this population. The developed risk prediction system can output the probability of users developing depressive symptoms or depression within five years and provide explanations for the prediction results, improving user accessibility and interpretability. Discussion: Rooted in China's national longitudinal cohort, this platform integrates machine learning analytics with interactive visualization, with web deployment enhancing its clinical translational value. By enabling early depression detection and evidence-based interventions for middle-aged and older adult populations, it establishes a novel health management paradigm with demonstrated potential to improve quality of life.