Analysis Of Risk Factors And Establishment Of A Clinical Prediction Model For Renal Function Progression In Patients With Diabetic Kidney Disease

ObjectiveThis study aims to find the risk factors for the progression of renal function in diabetes kidney disease(DKD)patients and to develop a simple and effective prediction model to assess the risk of renal function progression in DKD patients,so as to provide a reference for the treatment and follow-up of DKD in clinical practice.MethodsThe clinical data of DKD patients with multiple visits to the Second Hospital of Lanzhou University with complete case information from January 1,2015 to September 30,2022 were retrospectively collected for statistical analysis.By simple random grouping in excel,all patients were divided into modeling dataset and validation dataset according to the ratio of 7:3.In the modeling dataset,it is divided into the progressive group and non-progressive group of renal function according to whether the annual decline of estimated glomerular filtration rate(e GFR)was greater than 5ml/(min·1.73m~2).Logistic regression analysis and LASSO regression analysis were used to screen risk factors for progression of renal function in DKD patients.The above risk factors were used to development clinical prediction models respectively,and P<0.05 indicated that there was a statistical significance between the data of the two groups.The discrimination,calibration and clinical utility of the two models were evaluated separately,and the discrimination ability of the models was further evaluated by the net reclassification improvement and the integrated discriminant improvement,and the optimal model was selected.Finally,the optimal models were transformed into prediction equation and Nomogram for the convenience of clinicians.Results1.A total of 808 patients with DKD were included in this study,including 530 males and 278 females.There were two groups according to whether there was progression of renal function,173 cases in the progressive group and 635 cases in the non-progressive group,and the probability of having progressive renal function was 21.41%.All patients were divided into 2 groups according to the ratio of 7:3,with 565 cases in the modeling dataset and 243 cases in the validation dataset,and all indicators were balanced and comparable between the two groups,P>0.05.2.The model developed with the seven risk factors screened by Logistic regression analysis was model 1 and the model developed with the six risk factors screened by LASSO regression analysis was model 2.For model 1,the area under curve(AUC)of ROC of the modeling dataset is 0.812;the AUC of the validation dataset is 0.791;for model 2,the AUC of the modeling dataset is 0.807;the AUC of the validation dataset is 0.768,model 1 has a higher discrimination.The p-values of the Hosmer-Lemeshow test: 0.285 for the modeling dataset and 0.284 for the validation dataset for model 1,and 0.249 for the modeling dataset and 0.337 for the validation dataset for model 2,and combined analysis of the calibration curve results shows that model 1 has better calibration performance.Decision curve analysis(DCA)showed a high risk threshold interval range between approximately 1%-90% for model 1 and between approximately 1%-82% for model 2 in the validation dataset,with model 1 having greater clinical utility.In addition,compared with model 2,the NRI improved by 2.44% and the IDI improved by 1.50% of model 1.Finally,model 1 was determined to be the optimal model,and the variables incorporated into the model included systolic blood pressure(SBP),24-hour urinary protein quantity(UTP),serum uric acid(SUA),calcium(Ca),phosphorus(PHOS),total cholesterol(TC),and combined diabetic retinopathy.Finally,the optimal model was transformed into prediction equation and Nomogram.ConclusionsSBP,UTP,SUA,Ca,PHOS,TC and combined diabetic retinopathy were independent risk factors for progression of renal function in patients with DKD.The clinical prediction Model developed with seven predictors,including SBP,UTP,SUA,Ca,PHOS,TC and combined diabetic retinopathy,has good discrimination,calibration and clinical utility.It has some clinical value in predicting whether patients with DKD develop progression of renal function.