基于Logistic回归分析和决策树算法构建泌尿系统结石患者术后复发风险的预警模型
Construction of warning model for the risk of postoperative recurrenceinpatients withurinary calculi based on logistic regression analysis and decision tree algorithm
叶慧娟 1张功芳1
作者信息
- 1. 福建省立医院泌尿外科,福州 350000
- 折叠
摘要
目的:基于多因素Logistic回归分析和决策树算法构建泌尿系统结石患者术后复发的风险预警模型.方法:采用便利抽样法选取福建省立医院泌尿外科2021年6月至2023年6月住院治疗的泌尿系统结石患者300例,依据术后1年结石复发情况分为复发组(n=80)与非复发组(n=220),多因素Logistic回归分析泌尿系统结石患者术后复发的危险因素,采用SPSSModeler软件构建术后复发的决策树模型,并分析两种模型的预测效能.结果:300 例泌尿系统结石患者 1 年后 80 例术后复发,复发率为26.67%(80/300).多因素Logistic回归分析结果显示,术后结石残留、术后出现泌尿系统感染、术前尿常规白细胞"+++"、感染性结石、输尿管出现梗阻、高蛋白饮食是术后复发的危险因素;构建了术后复发的决策树模型,决策树模型共 4 层,17 个节点,决策树模型曲线下面积(AUC)是 0.780(95%CI:0.729～0.825),多因素Logistic回归模型的AUC是 0.753(95%CI:0.701～0.801),两种模型的Delong检验结果为Z=1.669,P=0.095.结论:本研究基于多因素Logistic回归分析和决策树法构建的泌尿系统结石患者术后复发的模型预测效能较好,可为患者复发的防治及制定护理策略提供理论参考.
Abstract
Objective:Based on multivariate logistic regression analysis and decision tree algorithm,the warning model for the risk of postoperative recurrence in patients with urinary calculi was established.Methods:A total of 300 patients with urinary calculi hospitalized in the Department of Urology of Fujian Provincial Hospital from June 2021 to June 2023 were selected by convenience sampling method.According to the recurrence of calculi one year af-ter surgery,they were divided into recurrent group(n=80)and non-recurrent group(n=220).Multivariate logistic regression was used to analyze the risk factors of postoperative recurrence in patients with urinary calculi.SPSS Mod-eler software was used to construct a decision tree model of postoperative recurrence,and the predictive efficacy of the two models was analyzed.Results:Among 300 patients with urinary calculi,80 recurred 1 year later,and the re-currence rate was 26.67%(80/300).Multivariate logistic regression analysis showed that postoperative stone resi-due,urinary system infection,white blood cell"+++"in preoperative urine routine,infectious stones,ureteral ob-struction and high-protein diet were risk factors for postoperative recurrence.A decision tree model for postoperative recurrence was constructed,which had 4 layers and 17 nodes.The AUC of the decision tree model was 0.780(95%CI:0.729～0.825),and that of the multivariate logistic regression model was 0.753(95%CI:0.701～0.801),the Delong test results of the two models are Z=1.669,P=0.095.Conclusion:The predictive performance of the models constructed based on multivariate logistic regression analysis and decision tree methods for postoperative recur-rence in patients with urinary tract stones is satisfactory.There models can provide theoretical references for the pre-vention,treatment,and nursing strategies for recurrence in these patients.
关键词
算法/Logistic模型/泌尿系统结石Key words
algorithms/logistic models/urinary calculi引用本文复制引用
出版年
2024
NETL
NSTL
万方数据