Xuehui Fan Ruixue Ye Yan Gao Kaiwen Xue Zeyu Zhang Jing Xu Yulong Wang ^*

• Shenzhen Second People's Hospital, Shenzhen, Guangdong Province, China

The Department of Rehabilitation Medicine is key to improving patients' quality of life. Driven by chronic diseases and aging population, there is a need to enhance efficiency and resource allocation of outpatient facilities. This study analyzes treatment preferences of outpatient rehabilitation patients using data and a grading tool to establish predictive models. The goal is to improve patient visit efficiency and optimize resource allocation through predictive models. Data were collected from 38 Chinese institutions, including 4244 patients visiting outpatient rehabilitation clinics. Data processing was conducted using Python software. Pandas library was used for data cleaning and preprocessing, involving 68 categorical and 12 continuous variables. Steps included handling missing values, data normalization, and encoding conversion. Data were divided into 80% training and 20% test sets using Scikit-learn library to ensure model independence and prevent overfitting. Performance comparisons among XGBoost, random forest, and logistic regression were conducted using metrics, including accuracy and receiver operating characteristic (ROC) curves. The imbalanced learning library's SMOTE technique was used to address sample imbalance during model training. The model was optimized using confusion matrix and feature importance analysis, and partial dependence plots (PDP) were used to analyze key influencing factors. XGBoost achieved the highest overall accuracy of 80.21% with high precision and recall in Category 1. Random forest showed similar overall accuracy. Logistic Regression had significantly lower accuracy, indicating difficulties with nonlinear data. Key factors include distance to institutions, arrival time, hospital stay, and diseases like cardiovascular, pulmonary, oncological, and orthopedic conditions. The tiered diagnosis tool helped doctors recommend institutions based on rehabilitation grading. Ensemble methods like XGBoost excel in complex datasets, and addressing imbalance improves performance. Understanding patient preferences aids healthcare policy to optimize resources, improve service quality, and enhance satisfaction.