The Construction Of Risk Prediction Model Of Ventilator-associated Pneumonia In Neurosurgical ICU Patients

Research Purpose:To analyze the risk factors of ventilator-associated pneumonia in neurosurgical ICU patients,construct a Logistic regression risk prediction model,draw and verify nomogram based on this model and develop a We Chat computing program.To help clinical nurses identify high-risk individuals for VAP in neurosurgical ICU patients accurately and quickly,improve the work initiative and efficiency of clinical nurses,increase the compliance of VAP prevention and control measures,and ultimately reduce the incidence of VAP,allocate human and material resources reasonably,improve the prognosis of patients.Research Methods:In this study,425 neurosurgical patients hospitalized in ICU of two tertiary general hospitals in Huzhou from January 1,2017 to December 31,2020 were selected into the modeling group.According to the occurrence of VAP,the patients were divided into case group and control group.The research indicators included in this study include general information,patients’ own factors,medical environment factors,drug factors and surgical factors.Use Epidata3.1 to enter information by two persons,using SPSS 26.0 for statistical analysis.Firstly,the indexes included in the study were analyzed by single factor analysis.Secondly,the indexes with statistical significance in the results of single factor analysis（with P<0.05 as the inclusion standard）were included in binary logistic regression analysis to clarify the risk factors of VAP in neurosurgical ICU patients（with P<0.05 as the difference is statistically significant）,and then the early prediction model was constructed,According to the risk prediction model,the nomogram was drawn with R 4.1.2 software and a We Chat computing program was developed.Bootstrap method was used to verify the model internally.The neurosurgical postoperative patients hospitalized in ICU of two tertiary general hospitals in Huzhou from January 1,2021 to December31,2021 were selected as the research object of the validation group to externally validate the model.ROC curve,Calibration curve and Hosmer-Lemeshow test were used to evaluate the discrimination,calibration and predictive effectiveness of the model,the DCA curve was drawn to evaluate the clinical effectiveness of the prediction model.Research Results:The incidence of VAP in neurosurgical ICU patients is 39.8%.Traumatic brain injury,ICU length of stay,propofol accumulation dose,tracheotomy,deep venous catheterization are independent risk factors for VAP in ICU patients of neurosurgery,while serum albumin is protective factor.The regression equation constructed based on these six factors is: Logit（p）=-2.440 + 2.640 ×（traumatic brain injury）+ 3.258 ×（ICU length of stay）+ 0.273 ×（propofol accumulation dose）+ 1.666 ×（tracheotomy）+ 2.355 ×（deep venous catheterization）-3.633 ×（serum albumin）.Hosmer-Lemeshow test value of the model is 6.910（P = 0.546 > 0.05）,and the fitting degree of the model is good.the ROC curve is 0.966（95% CI: 0.948 ～0.985）,the sensitivity is 89.9%,the specificity is 94.9%.The best cut-off value is 0.461 corresponding to the maximum Youden index 0.858.Internal verification shows that the area under the corrected ROC curve was 0.958（95% CI: 0.945～0.971）,Hosmer-Lemeshow test P = 0.941.External verification shows that Hosmer-Lemeshow test P = 0.759,the area of ROC is 0.954（95% CI: 0.916～0.951）,the distiction and calibration is better.The prediction results show that the sensitivity of the model is 91.2%,the specificity is 88%,and the accuracy is 89%.Decision Curve Analysis shows that the red line representing this model is not close to two extreme cases,the whole is inclined to the upper right corner,indicating that the model constructed in this study has a good clinical application value and a certain clinical benefit.Research Conclusion:Traumatic brain injury,ICU length of stay ≥ 7 days,high propofol cumulative dose,tracheotomy,deep venous catheterization,serum albumin <30g / L are risk factors of VAP in neurosurgical ICU patients,attention should be paid to and timely evaluated in clinical work,and targeted preventive measures should be given.The VAP risk prediction model of neurosurgical ICU patients constructed in this study has good discrimination and calibration.The developed risk prediction model calculation program can facilitate nurses to screen patients with VAP risk,provide the basis for early prevention of VAP in neurosurgical ICU patients,and reduce the incidence of VAP.