基于决策树分类模型的糖尿病肾病进展预测量表的研制

Development of Predictive Scale for Diabetic Kidney Disease Progression Based on Decision Tree Classification Model

目的 基于决策树分类模型建立糖尿病肾病(diabetic kidney disease,DKD)进展预测量表。方法 回顾性收集六盘水市第二人民医院内分泌科2020年7月至2021年7月收治的308例糖尿病肾病患者作为研究对象,并将其分为微量蛋白尿组(n=224)与显性蛋白尿组(n=84)。对2组患者的人口学资料、既往基础病史等指标进行单因素和多因素Logistic回归分析,并运用决策树分类模型建立DKD进展预测量表。结果 308例研究对象中84例(27.27%)为显性蛋白尿,224例(72.73%)为微量蛋白尿。多因素Logistic回归分析显示收缩压(OR=1.022,P=0.003)和血肌酐(OR=1.012,P <0.001)和总蛋白水平(OR=0.953,P=0.003)是引起显性蛋白尿的独立风险因素。运用决策树分类模型建立预测量表,量表总分为60分,诊断阈值为33分,决策树模型ROC曲线面积(0.781)大于多因素Logistic回归(0.769),灵敏度为95.2%,特异度为78.9%。结论 DKD进展预测量表能够较准确的评估DKD进展,对于早期预测DKD进展具有较好的临床价值。

Objective To establish a diabetic kidney disease(DKD)progression prediction scale based on the decision tree classification model.Methods A retrospective analysis was conducted on 308 patients with diabetic kidney disease admitted to Department of Endocrinology,the Second People's Hospital of Liupanshui from July 2020 to July 2021.The patients were divided into two groups:microalbuminuria group(n=224)and macroalbuminuria group(n=84).Univariate and multivariate Logistic regression analysis were performed on demographic data,past medical history and other indicators of the two groups of patients,and a DKD progression prediction scale was established using the decision tree classification model.Results Among the 308 subjects,84(27.27%)had macroalbuminuria and 224(72.73%)had microalbuminuria.Multivariate Logistic regression analysis showed that systolic blood pressure(OR=1.022,P=0.003)and serum creatinine(OR=1.012,P<0.001)and total protein levels(OR=0.953,P=0.003)were risk factors for macroalbuminuria.The decision tree classification model was used to establish a prediction scale with a total score of 60 points and a diagnostic threshold of 33 points.The area under the ROC curve of the decision tree model(0.781)was greater than that of the multivariate logistic regression model(0.769).The sensitivity was 95.2%and the specificity was 78.9%.Conclusion DKD progression prediction scale can accurately assess the progression of DKD and has good clinical value for the early prediction of DKD progression.