Application Of Cox's Proportional Hazards Regression Model Based On Principal Component Analysis In The Prognosis Of Hemodialysis

Objective: To explore application of Cox's proportional hazards regression model based on principal component analysis in the prognosis of hemodialysis?Methods: The paper was a retrospective research which analyzed the prognosis of maintain hemodialysis,patients were from June 2006 to November 2015 in the Second Affiliated Hospital of Nanchang University, there were 194 cases of patients.The baseline data of 194 cases of hemodialysis patients and 14 prognostic factors were analyzed with stata12.0,principal component analysis ?Cox regression and Cox regression model based on principal component analysis were analyzed with the statistical software of SPSS 22.0,and software of SAS 11.0 was uesd to verify whether the results of the principal component analysis was consistent. Colinear diagnosis was suggesting that the colinearity phenomenon existed between the independent variables, then the Cox regression model based on principal component analysis was used to analyze and explain the final regression model.Results: The survival rate of 1 to 4 year was 92.7% ? 80.4% ? 70% ?56.7%,respectively,which were Obtained by Kaplan-Meier method.14 prognostic factors(X1:gender;X2:dialysis frequency;X3:primary disease;X4:age;X5:serum albu-min;X6:serum uric acid;X7:blood calcium;X8:blood phosphorus;X9-X12 sequen-tially refers to:serum total cholesterol,triglycerides,high density lipoprotein, low density lipoprotein;X13:hemoglobinm;X14:intact parathyroid gland hormone) were analyzed by single factor analysis(Kaplan-Meier method, log-rank test) and 11 factors were included in the multivariate analysis(P<0.05).Univariate Cox proportional hazards regression analysis model was performed on 14 factors,X1 to X3 were not statistically significant(P > 0.05), but X4 to x14 were statistically significant(P < 0.05); the regression coefficient of X13 was negative,which were protective factor;the the regression coefficient of others were all positive, which were risk factors.And multivariate Cox proportional hazard regression model analysis included the 11 factors,and the results suggested: the coefficient of X4?X7?X8 and X14 werenegative,which were protective factors,the coefficient of X13 was positive,which was risk factor,the results were inconsistent with the univariate Cox proportional hazards regression analysis,and these were not meeting with the professional practice,it indicated that multicollinearity exists between the independent variables.seven principal components which extracted from principal component analysis were analysed by Cox proportional hazards regression model.One main component were into the Cox proportional hazard regression model equation.Conclusion:The survival rate of 1 to 4 year was 92.7% ? 80.4% ? 70% ?56.7%,respectively. The Cox proportional hazard regression model based on principal component analysis can solve the problem of multicollinearity between the prognostic factors of hemodialysis patients. Principal component analysis of Cox proportional hazards regression model showed that with the age of hemodialysis patients(15-29/30-44/45-59/60-69/70-79/80+ years old) levels rised in turn,the risk of death increased 9%,respectively; the death risk of hemodialysis patients that serum albumin did not meet with standard( <35g/L) increased 42% than patients which was consistent with target level; the death risk of hyperuricemia is 3.46 times as much as which were normal serum uric acid in hemodialysis patients; the death risk of serum calcium ? phosphorus and i PTH which were not up to the standards(Ca:2.2-2.5mmol/L,P:1.1-1.7 mmol/L,i PTH:150-300 pg/m L) increased 36% ? 36% ? 51%,respectively; while the death risk of the total cholesterol? triglyceride? high density lipoprotein ? low density lipoprotein which were inconsistent with the standards(TC<5.17mmol/L,TG<1.69mmol/L,HDL>1.03mmol/L,LDL<3.36mmol/L) were 3.82times?2.51 times?1.97 times?4.06 times as much as which were consistent with standards,respectively;with the degree of anemia reduced in turn, the risk of death decreased 27%, respectively.