| Part I Analysis of the predictors of pneumonia after lung surgeryObjective:The most common complication of lung surgery is pneumonia,and severe pneumonia is even life-threatening.In this study,we analyzed the data of patients undergoing pulmonary surgery in the thoracic surgery department of Jiangxi Provincial Cancer Hospital,and screened out the independent risk factors affecting postoperative pneumonia after lung surgery.Methods:Case data of patients who underwent lung surgery in thoracic Oncology Department of Jiangxi Cancer Hospital from 2018 to 2021 were analyzed,and eligible patients were selected by inclusion criteria and exclusion criteria.Eligible patients were randomly split into 2 datasets in a 7:3 ratio.In this study,the top 70%of the datasets were used for the predictive factor analysis of the postoperative pulmonary pneumonia,and they were divided into pneumonia and non-pneumonia groups according to whether pneumonia occurred postoperative.Independent predictive factors for postoperative pulmonary pneumonia were finally selected after lasso regression and multivariate logistic regression analysis.Results:The first part of this study included 1615 patients,split into the top 70%dataset,including 1136 patients,including 333 patients in the pneumonia group and 803patients in the no pneumonia group.Lasoo regression analysis selected smoking,transfusion,hypoalbuminemia,hypertension,diabetes,COPD,and emphysema in the multivariate Logistic regression analysis.Seven variables:smoking(P<0.001),transfusion(P<0.001),hypoalbuminemia(P<0.001),hypertension(P<0.001),diabetes(P<0.001),COPD(P<0.001),emphysema(P<0.001),P<0.05.Conclusion:Smoking,blood transfusion,hypoproteinemia,hypertension,diabetes,COPD,and emphysema were independent predictors of postoperative pulmonary pneumonia.Part II Construction and evaluation of a prediction model for pneumonia after pulmonary surgeryObjective:The prediction model of pulmonary postoperative pneumonia was constructed,and the efficacy of the established model was evaluated and presented visually in the form of a nomogram.Method:The first part of Lasso regression analysis and multivariate Logistic regression analysis were used to predict the risk prediction model of pulmonary postoperative pneumonia.Next,the differentiation,calibration,and clinical applicability of the prediction model were evaluated using the area under the subject working curve(AUC),Hosmer-Lemeisaw goodness of fit(H-L)test,and decision curve analysis(DCA).Finally,the model was visually presented using the nomogram.Results:The nomogram of pulmonary infection established in the second part of this study was accurate and the results were simple and understandable.The discrimina-tion degree of the model was evaluated,and the value of the area under the subject working curve(AUC)was 0.750>0.7,indicating that the constructed model has a good discrimination degree.The calibration of the model was evaluated,with H-L test X~2=2.3347,P=0.969>0.05,and the drawn calibration map shows that the calibration curve basically coincides with the reference line,suggesting that the model has excellent calibration.To evaluate the clinical applicability of the model,in the DCA decision curve,the DCA curve is above the extreme line,and when the threshold probability value is 8%-58%,the corresponding net benefit rate is 0-26%,suggesting that the model established in this study has excellent clinical validity.Conclusion:The risk prediction model of pulmonary postoperative pneumonia established in the second part of this study showed good differentiation,excellent calibration,and clinical applicability through differentiation,calibration,and clinical validity evaluation.The prediction model is presented by the nomogram in the form of the nomogram,which is more intuitive and easy to understand and convenient to calculate risks.Part III Internal validation of the prediction model for pneumonia after pulmonary surgeryObjective:The internal validation of the constructed prediction model for postoperative pneumonia in the lung further verified the differentiation degree,calibration degree,and clinical validity of the model.Method:In this part of the study,the degree of differentiation,degree of calibration,and clinical validity of the second risk prediction model were verified in the last 30%data set in 7:3.This dataset contained 479 patients,including 128 patients in the pneumonia group and 351 patients in the no pneumonia group.Results:The discrimination of the model was verified internally,and the area under the receiver working curve(AUC)was 0.754>0.7,and the latter 30%part was compared with the AUC of the prediction model constructed in the top 70%of the data set using the Delong test(Delong test).The P value=0.8973>0.05.The difference showed that the prediction model had good discrimination.The calibration of this model was verified internally,with H-L test internal data set X2=13.104 and P value=0.1083>0.05.The calibration map shows the calibration curve near the reference line,and these results indicate that the prediction model has excellent calibration.The clinical validity of the model is verified internally,the DCA clinical decision curve is plotted,the DCA curve of the internal validation set is located above the extreme line,when the threshold probability value of the internal validation set is 7%-72%,the corresponding net benefit is 0-27%,and these results indicate excellent clinical applicability of the model.Conclusion:The risk prediction model of pulmonary postoperative pneumonia established by this institute can quickly help medical workers to identify high-risk patients of pulmonary postoperative pneumonia,and take timely intervention measures to achieve precise and individualized treatment. |
PDF Full Text Request