Abstract
BACKGROUND: This study aims to develop and validate an Gradient Boosting algorithm (XGBoost) model for predicting the risk of depression in elderly stroke patients, and simultaneously identify the key risk factors. METHODS: A cross-sectional survey was conducted on 260 elderly patients with stroke. Depression scales were used for screening, and XGBoost was employed to analyze the data to identify the key influencing factors and rank them according to their predictive importance. RESULTS: Among the elderly stroke patients surveyed, the prevalence of depression was 24.615%. According to the XGBoost model, the importance of various factors was ranked as follows: sleep status, social participation, marital status, history of falls, and educational level. CONCLUSION: Depression in elderly stroke patients should not be overlooked. Clinical medical staff pay more attention to factors such as sleep status, social participation, marital status, history of falls, and educational level. Clinical medical staff should formulate individualized intervention strategies based on the specific conditions of elderly stroke patients to effectively reduce the risk of depression.