| Objective:Logistic regression and classification tree model were used to analyze the incidence of PEW and related risk factors in non-dialysis patients with CKD,aiming to provide theory for clinical prevention and treatment.in accordance with.Methods:The clinical data of CKD1-5 non-dialysis patients from January 2017 to June 1818 in our hospital were collected and included in the study according to the inclusion criteria.According to the clinical diagnostic criteria of PEW,they were divided into PEW group and non-PEW group.The differences of clinical indicators between the two groups were compared.Logistic regression analysis and classification tree model were used to analyze the risk factors affecting PEW,and the effectiveness of the two methods was evaluated by ROC curve.The explanatory variables selected by the classification tree model were studied.The relationship between the correlations and the correlation analysis selects Pearson/Spearman correlation analysis according to the specific conditions of the variables.Results:1.(1)In this study,400 patients with non-dialysis patients with CKD1-5 were enrolled.There were 217 patients with PEW,121 males and 96 females with an average age of(52.13±17.08)years;non-PEW patients had There were 183 cases,101 males and82 females,with an average age of(48.72±15.27)years old;the overall prevalence of PEW was 54.25%.(2)Compared with non-PEW group,PEW group had age(P=0.03),CRE(P=0.012),Urea(P=0.048),UA(P=0.006),PTH(P=0.038).Patients with UACR(P<0.001)and CRP(P=0.002)were higher than those in the non-PEW group;patients with Hb(P<0.001)and e GFR(P=0.003)were significantly lower than those in the non-PEW group.It is statistically significant.However,there were no significant differences in gender,MAP,LDL-C,glycosylated hemoglobin(Hb Alc),GLU,Ca,P,Mg and other items.2.(1)Logistic regression analysis showed that Hb(P=0.003),CRE(P<0.001),HCO3(P<0.001),UA(P=0.006),and CRP(P=0.002)were independent risk factors for PEW.(2)The constructed classification tree model consisted of 3 layers,including 15 nodes,9 of which are terminal nodes,and 5 explanatory variables are screened out from more than 20 predictors:PTH,Hb,CRE,UACR and CRP;(3)The classification tree model screened five explanatory variables,and the Pearson/Spearman correlation analysis was performed on the variables that may be correlated in the adjacent two layers.The results showed:PTH and Hb(r=-0.257,P<0.001),CRE and Hb(r=-0.409,P<0.001)was negatively correlated;PTH and CRE(r=0.304,P<0.001),CRE and CRP(r=0.100,P=0.046),CRE and UACR(r=-0.225,P <0.001)is positively correlated.3.Predicted probability obtained by two analytical methods to plot ROC curves.The area under the ROC curve(AUC)value of the logistic regression analysis was 0.722,and the AUC of the classification tree model ROC curve was 0.859.After the Z test,the difference in AUC between the two models was statistically significant(Z>1.96,P<0.05).Conclusion:1.In the CKD non-dialysis patients,the incidence of PEW is high,and factors such as CRE,Hb,CRP and PTH are closely related to the occurrence of PEW.2.Compared with logistic regression analysis,the classification tree model has a better predictive value for the occurrence of PEW in CKD patients. |
PDF Full Text Request