Logistic regression versus CART decision tree model for predicting pulmonary infection in elderly patients with heart failure with reduced left ventricular ejection fraction
10.3969/j.issn.1006-5725.2024.23.011
- VernacularTitle:老年左室射血分数降低型心力衰竭并发肺部感染预测的logistic与CART决策树模型对比
- Author:
Min LI
1
;
Hongqiang ZHAO
;
Bin CAO
;
Lili LIU
;
Yuzhen BAO
;
Fengyong YANG
Author Information
1. 山东第一医科大学附属济南人民医院全科医学科(山东济南 271100)
- Publication Type:Journal Article
- Keywords:
heart failure;
lung infection;
influencing factors;
decision tree
- From:
The Journal of Practical Medicine
2024;40(23):3349-3355
- CountryChina
- Language:Chinese
-
Abstract:
Objective To analyze the risk factors of pulmonary infection in elderly patients with heart fail-ure with reduced left ventricular ejection fraction heart failure,and establish a risk predicting model of pulmonary infection in those patients by decision tree CART algorithm.Methods 320 elderly patients with heart failure with reduced left ventricular ejection fraction admitted from January 2020 to December 2022 were retrospectively selected as study objects,and were divided into an infection group and a non-infection group according to whether the patients were complicated with pulmonary infection.Logistic regression model and decision tree CART model were used to construct a prediction model of heart failure with reduced left ventricular ejection fraction complicated with pulmonary infection,and 5-fold cross-validation method was used for internal verification.The prediction effi-ciency of the models was compared.Results In the 320 patients,the incidence of pulmonary infection was 30.94％.The data on age,smoking history,diabetes mellitus,cardiac function grades,COPD,invasive procedures,length of hospital stay were compared between the infection and non-infection groups(P＜0.05).logistic regression analysis showed that age of ≥ 75 years smoking history,complications with diabetes or/and COPD,cardiac function gradeⅢ/Ⅳ,invasive procedures,and hospital stay of ≥ 14 days were independent risk factors for pulmonary infection in the patients(P＜0.05).Probability forecasting model P=1/[1+e(-3368+0.763*X1+0.814*X2+0.652*X3+1.05*X4+0.865*X5+1.027*X6+0.652*X7)],with an overall accurate rate of prediction of 80.9％.The Omnibus test showed P＜0.001.The accuracy of predic-tion was 73.6％after the cross-validation of 5 fold.The decision tree model showed that invasive procedures were the most important influencing factors for pulmonary infection in elderly patients with heart failure with reduced left ventricular ejection fraction,with an information gain of 0.280.The ROC showed that the AUC value of logistic regression model was slightly higher than that of the decision tree(Z=2.850,P=0.004),and the prediction efficiency of both models was medium.Conclusions Age,smoking history,complications with diabetes mellitus or/and COPD,cardiac function grades,invasive procedures,and length of hospital stay are all influencing factors for pulmonary infection in elderly patients with heart failure with reduced left ventricular ejection fraction.The deci-sion tree model constructed in this study has a better efficiency for risk prediction,and it can provide reference for early clinical screening and intervention of heart failure with reduced left ventricular ejection fraction.
