Risk factors of bladder stones in patients with benign prostatic hyperplasia and development of a SHAP-based interpretable predictive model
10.3969/j.issn.1009-8291.2025.08.004
- VernacularTitle:良性前列腺增生并发膀胱结石的危险因素分析及基于SHAP的可解释性预测模型构建
- Author:
Huan ZHAO
1
;
Song OUYANG
1
;
Hongchao DONG
1
;
Qinzhang WANG
1
Author Information
1. 石河子大学第一附属医院泌尿外科,新疆石河子 832008
- Publication Type:Journal Article
- Keywords:
prostatic hyperplasia;
bladder stones;
nomogram;
prediction model;
SHAP algorithm
- From:
Journal of Modern Urology
2025;30(8):653-661
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the independent risk factors of bladder stones(BS)in patients with benign prostatic hyperplasia(BPH),and to construct a predictive model and an easy-to-use website.Methods The clinical data of 460 BPH patients treated during Jan.2022 and Jan.2025 in the First Affiliated Hospital of Shihezi University were retrospectively analyzed.The independent risk factors of BS in the training set were identified with univariate logistic regression and the Boruta algorithm,based on which a nomogram was constructed.The predictive performance of the model was evaluated with receiver operating characteristic(ROC)curve,sensitivity,specificity,positive predictive value(PPV),negative predictive value(NPV),F1 index,calibration curve and clinical decision curve analysis(DCA).The contribution of different variables to BS was evaluated with SHapley Additive exPlanations(SHAP)algorithm.A web page was established.Results Among the 460 BPH patients,144(31.3％)had BS,and 6 independent risk factors were identified,including neutrophil level,urine culture results,intravesical prostatic protrusion(IPP),urine nitrite test results,urine leukocytes test results,and urine occult blood results.In the test set,the area under the ROC curve(AUC)was 0.887(95％CI:0.816-0.947),sensitivity 0.705,specificity 0.968,PPV 0.912,NPV 0.876,and F1 score 0.795.Calibration and DCA indicated good discrimination and clinical applicability.SHAP results showed that the risk factors mentioned above were the most important for concurrent BS.The resulting website(https://wutiaowu2.shinyapps.io/bladderrrr/)was publicly accessible.Conclusion The neutrophil level,urine culture results,IPP,urinary nitrite test results,urinary leukocytes test results,and urinary occult blood test results were identified as the independent risk factors of BPH complicated with BS.The model and website developed based on these factors demonstrate high usability and accuracy,possessing significant clinical value.
