Objective:

To develop and validate an explainable model for predicting likelihood of favorable outcome under integrative medicine treatment (IMT) exposure in adult diabetic kidney disease (DKD) patients.

Approach:

• Cohort Analysis: A retrospective cohort of 7,400 DKD patients from 2010 to 2018 was analyzed, with 3,900 cases divided into training and test sets, and 3,500 cases for temporal validation.
• Feature Selection and Model Development: Feature selection was performed using the Boruta algorithm, followed by LASSO regression, Random Forest, and XGBoost-SHAP analysis. Predictive models were evaluated based on AUC and other performance metrics.
• Web Application Deployment: The optimal XGBoost classifier was deployed as an interactive web application using the R Shiny framework.

Key Findings:

• XGBoost achieved the best performance with AUC = 0.783 in training, 0.715 in test, and 0.762 in validation sets.
• Ten key variables were identified: creatinine, uric acid, age, red blood cell count, urea, glucose, platelet count, calcium, white blood cell count, and sodium.
• The web application allows for real-time prediction of favorable outcomes under IMT.

Interpretation:

The model effectively predicts likelihood of favorable outcomes under IMT exposure in DKD, aiding personalized treatment.

Limitations:

• The study is retrospective and may be subject to biases inherent in such designs.
• The model's applicability may be limited to the specific population studied.

Conclusion:

The developed model provides a tool for predicting outcomes in DKD patients receiving IMT, facilitating personalized treatment strategies.