Abstract:Objective Based on machine learning algorithm, to develop a depression predict model in patients with end-stage renal disease after renal transplantation and verify the prediction effect of the model. Methods A total of 189 patients undergoing renal transplantation in the Urology Department of a tertiary A-level hospital in Nanjing from February 2015 to March 2019 were selected for the model development. According to the one-year follow-up results after operation, they were divided into depression group (n=74) and non-depression group (n=139). After oversampling of the raw imbalanced data, the dataset was randomly stratied to training set and testing set in the ratio of 0.75∶0.25. Using six commonly used machine learning algorithms, 10-fold cross-validation on the training set was performed to test out the most suitable algorithm based on the AUC value and to evaluate them on the test set, screening out the important features related to prediction. Results After 10-fold cross-validation, the random forest algorithm was selected for the development of the final model. On the testing set, the accuracy, sensitivity, and specificity of the final model were 80.6%, 83.3% and 78.1% respectively. The top five features with the highest contribution of prediction model were creatinine, care mode, income, total protein and albumin. Conclusions The random forest model for depression prediction among the renal transplant recipients shows good effect and provides valuable information for specialists to refer to. Additionally, collaborative nursing mode is important in improving the mental health of such patients.