Construction and validation of a risk prediction model for hypostatic pneumonia in patients with severe craniocerebral injury undergoing tracheotomy

TAN Wanlan; ZHAI Juan; YOU Yiqing; DENG Qin; YANG Yujie

doi:10.12104/j.issn.1674-4748.2025.24.005

Abstract：Objective:To establish a risk prediction model for hypostatic pneumonia in patients with severe craniocerebral injury undergoing tracheotomy,and to verify its effect.Methods:Convenient sampling was used to select 188 and 55 patients with severe craniocerebral injury who underwent tracheotomy at Friendliness Hospital Yangzhou in Jiangsu Province from January 2022 to October 2024 and November 2024 to January 2025,respectively.A modeling group and a verification group were established respectively.The patient information was collected by general data questionnaire and the number of patients with hypostatic pneumonia was counted.The risk factors of hypostatic pneumonia in patients with severe craniocerebral injury were screened by univariate and multivariate logistic regression analysis,and the risk prediction model was constructed based on the results of regression analysis,and its effect was verified.Results:The statistics showed that 22 patients had hypostatic pneumonia,the incidence was 11.70%.Univariate and multivariate logistic regression analysis showed that age≥60 years old,diabetes mellitus,Glasgow Coma Scale(GCS)score<7,hypoproteinemia,use of more than 2 types of antibiotics,and regular sputum aspiration were the risk factors for hypostatic pneumonia in patients with severe craniocerebral injury undergoing tracheotomy(P<0.05).The area under the receiver operating characteristic(ROC)curve of the subjects in the modeling group and the verification group were 0.821[95%CI(0.772,0.869)]and 0.817[95%CI(0.814,0.821)],respectively,and the Youden index were 0.568 and 0.600,respectively.The Hosmer-Lemeshow test of the modeling group and the verification group showed that P was>0.05,and the difference was not statistically significant.The calibration curve showed that the calibration curve was close to the reference line,and the prediction probability was in good consistency with the actual occurrence probability.This suggests that the model had good calibration degree and clinical practicability.Conclusion:Patients with severe craniocerebral injury have the risk of hypostatic pneumonia,and its risk factors are complex.The risk prediction model of hypostatic pneumonia in patients with severe craniocerebral injury undergoing tracheotomy established in this study has a good calibration degree and differentiation degree,which can effectively predict the risk of hypostatic pneumonia,provide a reliable basis for clinical nursing,and help strengthen the prevention effect,reduce the incidence of hypostatic pneumonia.