Machine learning-based prediction of gout flares during hospitalization in patients with upper gastrointestinal
Rongrong Chen1,2, Sihan Hu3, Shiyun Lu1,2
1Shengli Clinical Medical College of Fujian Medical University, Fuzhou, Fujian, China.
Background And Aims:
Acute gout flares pose a significant therapeutic challenge in hospitalized patients with upper gastrointestinal bleeding (UGIB) due to the contraindications of standard anti-inflammatory treatments. This study aimed to develop and validate a machine learning (ML) model to predict the risk of gout flares in this high-risk inpatient population.
Methods:
A retrospective cohort study was conducted on UGIB patients admitted to the Department of Gastroenterology, Provincial Hospital, Fujian Medical University. Five ML algorithms-Decision Tree (DT), Random Forest (RF), k-Nearest Neighbors (KNN), Naive Bayes (NB), and Extreme Gradient Boosting (XGBoost)-were trained and tested using routinely collected clinical and laboratory data at admission. Model performance was evaluated on an independent test set using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) were used to enhance model interpretability.
Results:
A total of 718 patients were included, with 158 (22.0%) experiencing a gout flare during hospitalization. The RF model exhibited the best predictive performance, achieving an AUC of 0.951 (95% CI: 0.923-0.979), accuracy of 0.901, sensitivity of 0.929, and specificity of 0.873 on the test set. DCA confirmed the clinical utility of all models. SHAP analysis identified six key predictors: serum uric acid (UA), creatinine (Cr), hemoglobin (HB), blood urea nitrogen (BUN), body mass index (BMI), and alcohol consumption history.
Conclusion:
This study successfully developed a robust ML model, with RF as the optimal algorithm, for accurately predicting inpatient gout flares in UGIB patients. This tool facilitates early identification of high-risk individuals, enabling targeted preventive strategies and enhancing clinical management.