Study On Patients Of AECOPD And COPD Combined With Pneumonia By Using Cluster Analysis And Principal Component Analysis And Factor Analysis

Objective: COPD(chronic obstructive pulmonary disease)is a complex syndrome,which is characterized by many pulmonary and extrapulmonary manifestations.From the clinical observations for so many years,we found that COPD is a heterogeneous disease.The primary aim of identifying phenotypes is to provide patients with the best health care possible,tailoring the therapeutic approach to each patient.This study uses polynary statistical method,including principal component analysis(PCA),factor analysis and cluster analysis,to describe the complexity of patients who are admitted to hospital because of AECOPD(acute exacerbation of chronic obstructive pulmonary disease)and COPD with pneumonia.Then we can distinguish patients individual prognosis and characterristics of treatment,and tailor the therapeutic approach to each patient.In addition,we can study the application value of polynary statistical method in this field for further discusions.Method: Collect the clinical datas of 162 inpatients who are in the Department of respiration of Second Hospital Affiliated to Tianjin Medical University from September,2014 to December,2015.The clinical datas including sex,age,somking idex,course of disease,exacerbation rate,complications(hypertension,diabetes,coronary disease),GOLD stages,eosinophile granulocyte,Alb(albumin),BUN(blood urea nitrogen),PCO2(pressure of carbon dioxide),antimicrobial species and curative effect.The clinical datas were processed by SPSS20.0 statistical software.Patient classification was performed using two-step clustering method based on principal component--factor analysis transformed datas.Result: 162 COPD subjects were analysed :79 were male,83 were female;average age was 77.5 years(range 44-93years);median(interquartile range)for course of disease was 10(8,30)yrs,(range1-60)yrs.Principal component--fator analysis showed that 7 independent components accounted for 78.44% of all variances.PCA-based twostep cluster analysis,resulted in the classification of 162 subjects into 4 groups: older/ short duration / medium,younger older / short duration / mild,younger older /long duration / severe,older / long duration / severe.the patients in different groups are significant different in age,course of disease,smoking idex,EO,PCO2,exacerbation rate,GOLD stage,comlications,antimicrobial species and curative effect(P<0.05),Four groups are no difference in sex distribution,Alb and BUN(P>0.05).Conclusion: Principal component analysis can search for a few variables(or common factors)to reflect most of the information of all variables,and these new variables are irrelevant to each other.Clustering analysis can maximize classify the homogeneous patients into the same group,and heterogeneous patients into different groups.The results suggest that polynary ststistical method has certain clinical value in classifing the COPD patients.