Investigation Of Nosocomial Infection In Patients With Laryngeal Carcinoma Resection And Construction Of Pulmonary Infection Risk Prediction Models

Objectives:We investigated the current situation of nosocomial infection by targeting patients with laryngeal carcinoma resection.To explore the risk factors of pulmonary infection in patients after laryngeal carcinoma resection.The nomogram prediction model and decision tree risk prediction model for the pulmonary infection after laryngeal carcinoma resection have been constructed.The research aims to provide a convenient early warning model for clinical staff,control the incidence of pulmonary infection,continuously improve the quality of medical care.Methods:The subjects were patients who underwent laryngeal carcinoma resection in the otolaryngology department of the hospital from January 1,2019 to December 31,2022.Relevant information was collected by using the self-designed Targeted Monitoring Questionne of Nosocomial Infections in Patients with Laryngeal Carcinoma Resection,in order that we investigated the current situation of nosocomial infection in patients after laryngeal carcinoma resection.According to the results of single-factor analysis and the selection of appropriate independent variables in combination with clinical practice,multi-factor logistic regression analysis is included to screen out independent risk factors for,pulmonary infection.We constructed the nomogram risk prediction model for the pulmonary infection.In addition,the CHAID algorithm has been used to construct the decision tree risk prediction model.Then,used the area under curve,the Hosmer-Lemeshow test and the decision curve analysis to carry out the differentiation,calibration and clinical effectiveness of the model.AUC and error classification table had been used to evaluate the predictive effect of the decision tree model.Compared the predictive performance by analyzing the AUC of the two risk prediction models.Results:1.This study included a total of 535 patients,including 512 men and 23 women,with an average age of(62.01±8.55)years.2.In all patients undergoing laryngeal carcinoma resection,49 patients had nosocomial infections,a total of 57 nosocomial infections occurred.The incidence of nosocomial infections was 9.16%.3.Among the types of nosocomial infections in this study,there were 36 cases of pulmonary infection(73.47%),9 cases of surgical site infections(18.37%),of which 2 cases had laryngeal fistula(4.08%),3 cases of multiple infections(pulmonary infection+surgical site infections)(6.12%),and 1 case of upper respiratory tract infections(2.04%).4.In this study,a total of 41 strains of pathogens and 31 strains of Gram-negative bacteria(75.61%),of which 8 strains of Acinetobacter baumannii(19.51%),7 strains of Pseudomonas aeruginosa(17.07%),6 strains of Klebsiella pneumonia(14.63%),and 9 strains of Gram-positive bacteria(21.95%),including 5 strains of Streptococcus pneumoniae(12.20%),3 strains of Bacillus G+(7.32%),1 strain of methicillin-resistant Staphylococcus aureus(2.44%),and 1 strain of fungi(Dida albicans)(2.44%).5.Logistic multi-factor regression analysis showed the patient’s age,combined with coronary heart disease,operation time,sputum suction times and postoperative static human blood albumin were the independent risk factors affecting the pulmonary infection of patients with laryngeal carcinoma resection.AUC of the nomogram prediction model was 0.810(95%CI:0.774～0.843),the χ2 in the Hosmer-Lemeshow fitting excellence test was 6.305,P=0.390.It shows that the model has good differentiation and calibration.The DCA curve shows that the clinical effectiveness of the model is good.6.In this study,the decision tree model screened out four prediction variables of postoperative static drop human blood albumin,age,NNIS and sputum suction frequency,and generated 5 decision rules.AUC of the decision tree prediction model was 0.810(95%CI:0.774～0.843).It shows that the prediction effect of the model is good.7.Compared the AUC of the two prediction models,the results showed that Z=0.272,P=0.785(P>0.05).Conclusions:The incidence of nosocomial infection in patients undergoing laryngeal carcinoma resection is relatively high,mainly pulmonary infection.Nosocomial infection management departments and clinical medical staff should pay attention to it.Age(≥60 years old),coronary heart disease,duration of operation(≥200min),sputum suction(>36 times),postoperative intravenous human albumin and the NNIS were the independent risk factors for postoperative pulmonary infection in patients with laryngeal carcinoma resection.Clinical medical staff should pay attention to it when formulating nosocomial infection prevention measures.The postoperative pulmonary infection risk prediction models had a good prediction effect,which was helpful for clinical medical staff to identify the risk of pulmonary infection at early stage.