Senel Cagdas, Erkan Anil, Keten Tanju, Aykanat Ibrahim Can, Ozercan Ali Yasin, Tatlici Koray, Basboga Serdar, Saracli Sinan, Guzel Ozer, Tuncel Altug
Department of Urology, Balikesir University School of Medicine, Balikesir, Turkey.
Department of Urology, Balikesir University School of Medicine, Balikesir University Hospital Second Floor Block C, Altieylul, Balikesir, Turkey.
Urolithiasis. 2024 Dec 24;53(1):15. doi: 10.1007/s00240-024-01685-x.
The current study aimed to determine the risk factors and define a new scoring system for predicting febrile urinary tract infection (F-UTI) following retrograde intrarenal surgery (RIRS) by using machine learning methods. We retrospectively analyzed the medical records of patients who underwent RIRS and 511 patients were included in the study. The patients were divided into two groups: Group 1 consisted of 34 patients who developed postoperative F-UTI, and Group 2 consisted of 477 patients who did not. We applied feature selection to determine the relevant variables. Consistency subset evaluator and greedy stepwise techniques were used for attribute selection. Logistic regression analysis was conducted on the variables obtained through feature selection to develop our scoring system. The accuracy of discrimination was assessed using the receiver operating characteristic curve. Five of the 19 variables, namely diabetes mellitus, hydronephrosis, administration type, a history of post-ureterorenoscopy (URS) UTI, and urine leukocyte count, were identified through feature selection. Binary logistic regression analysis showed that hydronephrosis, a history of post-URS UTI, and urine leukocyte count were significant independent predictors of F-UTI following RIRS. These three factors demonstrated good discrimination ability, with an area under curve value of 0.837. In the presence of at least one of these factors, 32 of 34 patients (94.1%) who developed postoperative F-UTI were successfully predicted. This new scoring system developed based on hydronephrosis, a history of post-URS UTI, and urine leukocyte count can successfully discriminate patients at risk of F-UTI development after RIRS.
本研究旨在通过机器学习方法确定逆行性肾内手术（RIRS）后发热性尿路感染（F-UTI）的危险因素并定义一种新的评分系统。我们回顾性分析了接受RIRS的患者的病历，共有511例患者纳入研究。患者分为两组：第1组由34例术后发生F-UTI的患者组成，第2组由477例未发生F-UTI的患者组成。我们应用特征选择来确定相关变量。一致性子集评估器和贪婪逐步技术用于属性选择。对通过特征选择获得的变量进行逻辑回归分析以建立我们的评分系统。使用受试者工作特征曲线评估判别准确性。通过特征选择确定了19个变量中的5个，即糖尿病、肾积水、给药类型、输尿管肾镜检查（URS）后尿路感染病史和尿白细胞计数。二元逻辑回归分析表明，肾积水、URS后尿路感染病史和尿白细胞计数是RIRS后F-UTI的重要独立预测因素。这三个因素具有良好的判别能力，曲线下面积值为0.837。在存在至少其中一个因素的情况下，34例术后发生F-UTI的患者中有32例（94.1%）被成功预测。基于肾积水、URS后尿路感染病史和尿白细胞计数建立的这种新评分系统能够成功区分RIRS后有发生F-UTI风险的患者。
