Machine learning models for predicting renal injury in patients with gout

Area of Science:

• Nephrology
• Medical Informatics
• Machine Learning

Background:

• Renal injury is a serious complication for individuals with gout.
• Developing predictive models for kidney injury in gout patients is crucial.

Purpose of the Study:

• To construct and evaluate machine learning models for predicting renal injury risk in gout patients.
• To identify key variables associated with renal injury in this population.

Main Methods:

• Utilized the NHANES database (2007-2018) to train predictive models.
• Compared Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN) models.
• Assessed model performance using AUC, calibration curves, sensitivity, specificity, accuracy, and F1 score.

Main Results:

• Analyzed 1,203 patients using seventeen variables.
• XGBoost demonstrated the highest predictive performance based on AUC.
• Key predictors identified include blood urea nitrogen, age, uric acid, and urinary albumin.

Conclusions:

• Successfully developed ML models for predicting renal impairment in gout patients.
• The XGBoost model exhibited superior performance.
• A web-based tool was created for estimating renal injury probability in gout patients.