Construction And Verification Of Prediction Model Of Diabetic Nephropathy Progressing To End-stage Renal Disease

Objective: This study intends to establish and verify the prediction model based on multiple logistic regression equation,discover and analyze the clinical risk factors of traditional Chinese and western medicine in diabetic nephropathy,provide the basis for risk prediction of DKD patients developing ESRD,and provide support for the formulation of corresponding prevention and treatment programs.Methods: In this study,801 patients with DKD hospitalized in the Department of Nephrology,Affiliated Hospital of Jiangxi University of Traditional Chinese Medicine from January 1,2015 to December 31,2021 were included in a retrospective analysis.Part of the case data was included in the training set to build the prediction model,and the rest of the case data was included in the validation set to verify the model.The TCM syndrome type determination scale developed was used to determine the TCM syndrome type of all cases,making each case belong to one standard syndrome and one primary syndrome.The cases in the training set were divided into DKD group and ESRD group,and SPSS22.0 statistical software was used for single factor logistic regression analysis of the data,to obtain the risk factors of TCM syndrome types,basic information and laboratory test indicators,and then the obtained risk factors were diagnosed collinearity to avoid multicollinearity.Multivariate logistic regression analysis was carried out on the data.Independent risk factors with significant differences were used to construct prediction models.ROC curves were used to evaluate model performance,determine the diagnostic threshold of the models,and calculate the sensitivity,specificity and Yoden index of the models.The data of 50 cases in the verification set were substituted into the completed equation model.The mixed matrix and ROC curve analysis were used to obtain the relevant data of model validation,namely the area under ROC curve,sensitivity,accuracy,specificity,Yoden index and Kappa coefficient.Results: A total of 801 diabetic nephropathy patients were included in this study,and 751 cases were included in the training set.According to the estimated glomerular filtration rate(e GFR)< 15,which means entering the ESRD outcome,the training set was divided into DKD group and ESRD group,with 527 cases in the DKD group and 224 cases in the ESRD group.A total of34 factors were included in the analysis,and the syndrome type information of TCM was the deficiency of Yin and dry heat,the deficiency of qi and Yin,the deficiency of spleen and kidney and the deficiency of Yin and Yang,the standard syndrome of dampness-heat,the standard syndrome of cold and dampness,the standard syndrome of blood stasis and the standard syndrome of phlegm and blood stasis.Basic information and laboratory test indicators were sex,age,systolic blood pressure,diastolic blood pressure,diabetes course,diabetic retinopathy,diabetic neuropathy,diabetic cardiovascular disease,diabetic cerebrovascular disease,hemoglobin,fasting blood glucose,glycocated hemoglobin,serum albumin,alkaline phosphatase,total cholesterol,high density lipoprotein cholesterol,low density lipoprotein Cholesterol,triglyceride,blood urea nitrogen,blood creatinine,blood uric acid,D-dimer,fibrinogen,partial prothrombin time,prothrombin time,urinary protein.Univariate logistic regression analysis showed that there was a statistically significant difference between Yin and Yang deficiency syndrome of TCM syndrome type(P < 0.001),and the standard syndrome of cold and dampness syndrome had a significant difference(P <0.001).The difference of disease duration and diabetic retinopathy in basic information was statistically significant(P < 0.001).The differences of hemoglobin,fasting blood glucose,glycosylated hemoglobin,serum albumin,high-density lipoprotein cholesterol,blood urea nitrogen,serum creatinine,blood uric acid,D-dimer,fibrinogen and urinary protein in laboratory test information were statistically significant(P < 0.001).The obtained risk factors were diagnosed as collinearity to avoid multicollinearity between variables.The conditional index of each variable was less than 10,indicating that there was no multicollinearity between variables.All variables were included in multivariate logistic regression analysis.Multivariate logistic analysis showed that Yin and Yang deficiency syndrome(regression coefficient 0.863),hemoglobin(regression coefficient1.972),urine protein(regression coefficient 1.293)and blood urea nitrogen(regression coefficient 3.595)were independent risk factors in the development of DKD to ESRD,and the constant was-6.289.The equation model jointly constructed P=exp(-6.289+0.863 Yin and Yang deficiency syndrome+1.972 × HGB+1.293 × PRO+3.595 × BUN)/[1+exp(-6.289+0.863 Yin and Yang deficiency syndrome +1.972 × HGB+1.293 × PRO+3.595 × BUN))].The model diagnosis threshold was 0.3478,the sensitivity was 0.853,the specificity was 0.808,the Yoden index was 0.661,and the area under ROC curve was0.888.After the verification set data were substituted into the equation,the positive prediction rate was 91.18%,the negative prediction rate was87.50%,the model sensitivity was 0.813,the specificity was 0.941,the accuracy was 0.9,the Yoden index was 0.754,the Kappa coefficient was 0.774,and the area under the ROC curve was 0.936,indicating that the model had good consistency and good performance.Conclusion: In the progression of diabetic nephropathy to end-stage renal disease,the deficiency of Yin and Yang syndrome,hemoglobin,urine protein and blood urea nitrogen are independent risk factors for its progression.In clinical diagnosis,treatment and decision-making,equation model is a relatively simple and easy to operate means to predict the progression of patients’ disease,and can provide evidence and help for early intervention and treatment.