| Objective:1.To investigate the influencing factors of MCI in COPD patients and build a risk early warning model.2.Classify MCI risk level of COPD patients.Methods:1.Patients with COPD who met the admission criteria were selected from the respiratory department,geriatrics department of two 3A hospitals and community health service center in Taiyuan from November 2019 to November 2020.The Montreal Cognitive Assessment Scale(Mo CA),Social Function Activity Questionnaire(FAQ)and Clinical Dementia Scale(CDR)were used to evaluate whether MCI occurred in COPD patients.Patients with COPD were divided into mild cognitive impairment(MCI)group and normal cognitive function(NC)group.The related information of COPD patients was collected through general data questionnaire,COPD-Q questionnaire,electronic medical record system,modified British MRC dyspnea score(m MRC),etc.The general data questionnaire included general situation,past disease history,clinical data,living habits and so on.Used Chi-square test and Logistic regression to analyze the influencing factors of MCI in COPD patients.Taking the actual outcome of MCI of COPD patients as the target variable,the collected research variables as the input variables are included in the decision tree,and all samples are set as the training set and the test set in a ratio of 7:3.The risk early warning model of MCI decision tree of COPD patients is constructed,and the receiver operating characteristics(ROC)curve is drawn to compare the predictive ability of Logistic regression model and decision tree model.2.According to the predicted probability of MCI in COPD patients,the decision tree CHAID algorithm in SPSS Modeler14.1 software is used to classify the risk level,and the classification effect of the risk level evaluation model is evaluated by the accuracy rate and gain diagram.Results:1.Logistic regression analysis showed that old age,female,low education level,manual workers,low family income per capita,high blood pressure,no physical exercise,no recreational activities,high dyspnea,poor compliance,low health literacy,complicated with pulmonary heart disease and long illness time were independent risk factors for MCI of COPD patients,while never smoking,drinking tea and taking home oxygen therapy as needed were protective factors for MCI of COPD patients.Health literacy,compliance,education level,complicated with pulmonary heart disease,dyspnea grade,age,physical exercise,whether to take home oxygen therapy as needed,smoking years and hypertension entered the decision tree model as nodes,showing 13 prediction paths of MCI in COPD patients.The root node is health literacy,and its importance is 29%.The prediction accuracy of Logistic regression model is 88.37%,that of decision tree model training set is87.47%,that of test set is 92.26%,and the overall prediction accuracy is 88.91%.ROC curve showed that AUC of Logistic regression model was 0.883(95% CI: 0.854-0.909),and AUC of decision tree model was 0.889(95% CI: 0.860-0.914),and there was no significant difference in predictive ability between the two models(P=0.7022).2.The prediction probability of MCI in COPD patients is graded by decision tree CHAID algorithm,and finally the MCI risk rating of COPD patients is divided into five grades: extremely low risk(0 < Pr ≤ 0.706),low risk(0.706 < Pr ≤ 0.762),medium risk(0.762 < Pr ≤ 0.870),high risk(0.870<Pr ≤ 0.934)and extremely high risk(Pr>0.934).The prediction accuracy of the training set and the test set of the risk level assessment model is 80.56% and 80.95%,respectively.The gain diagram shows that the model has certain discrimination ability.Conclusion:1.MCI of COPD patients is influenced by many factors,which interact with each other.Logistic regression model and decision tree model can effectively predict the risk of MCI of COPD patients,and there is no significant difference in reliability between them.The decision tree model uses top-down tree classification to get 13 prediction paths,which is more intuitive in presentation and more detailed in prediction path,which is helpful for clinical workers to find high-risk individuals conveniently and quickly,and provides new ideas for clinical decision-making,so as to intervene intermediate factors and changeable factors as early as possible and delay the occurrence and development of MCI in COPD patients.2.The evaluation of MCI risk level of COPD patients based on decision tree classification has certain discrimination ability,which can evaluate the risk degree of MCI through the personal data of COPD patients and realize early warning.Clinicians can formulate corresponding intervention strategies and take targeted measures according to MCI risk level. |
PDF Full Text Request