Analysis Of Risk Factors And Establishment Of Risk Prediction Model For Klebsiella Pneumoniae Infection In A Third Class A Hospital In Nanchang City

Objective:By analyzing the infection status and risk factors of Klebsiella Pneumoniae KP in a class A hospital of Nanchang city,the independent risk factors of KP infection in Nanchang city were discussed,and the prediction model of KP infection was established by using machine learning method.To provide data support for the formulation of KP prevention strategies,in order to reduce THE INFECTION rate of KP in Nanchang,which has important guiding significance for clinical prevention and improvement of prognosis.Methods:The method of retrospective study was used to collect information of patients in the case group and control group according to inclusion and exclusion criteria through big data center records combined with patient records and telephone return visits.Data of 1295 cases of KP infection in Nanchang city from 2018 to 2021 were collected(excluding repeated cases submitted for examination).Analysis was performed in a 1:1 case-control study,in which case group(N=1295)and control group(N=1295).The first step is to make descriptive statistics of the relevant information of the two groups.In the second step,single factor analysis was conducted on the two groups of information data to screen out the risk factors related to KP infection,Lasso regression analysis was performed to reduce overfitting,and then multi-factor Logistic regression analysis was used to evaluate the selected risk factors to determine the independent risk factors of KP infection.The selected independent risk factors will be used as features to construct the prediction model.The third step applies Logistics model,XG Boost model,Light GBM model,random forest model,Ada Boost model,Gaussian Naive Bayes model and neural network model to construct KP infection prediction model respectively.The proportion of test set selection is 15%,and the remaining 85% samples are used to construct prediction model.The samples outside the test set were divided into 10 pieces and trained for 10 times.Each time,9 pieces were used as the training set and 1 piece was used as the verification set for internal verification to verify whether the model fit was successful.The performance and effect of various prediction models were evaluated comprehensively by analyzing AUC value,accuracy,calibration curve,decision curve DCA and other indicators of model test set.KP infection data(N=331 in the case group and N=331 in the control group)were selected from another third-class a hospital in Nanchang,and the cross-validation method was used to verify the model effect for external verification.Results:1.The detection rate of KP showed an upward trend,increasing from 8.88% in2018 to 12.12% in 2021.Most of 608 strains were detected in sputum,followed by276 in urine.The infection site accounted for the largest proportion of lung infection,followed by urinary tract infection.The resistance of KP to clinically common antibiotics increased year by year.KP had the highest resistance to ceftriaxone(50.11%)and cefazolin(56.77%),and was most sensitive to imipenem(5.67%)and ertapenem(8.92%).The proportion,utilization rate and intensity of antibiotics in the case group were higher than those in the control group.KP mainly distributed in neurosurgery(12.90%),followed by respiratory and critical care(12.28%).2.The independent risk factors for KP infection were age,hospitalization times,length of stay,hyperlipidemia,coronary heart disease,hypertension and central venous canal.3.KP infection prediction model based on XG Boost algorithm has the best effect,with ROC curve area =0.867 and accuracy =0.779.Conclusion:KP infection rate increased year by year,clinical doctors and nurses should first distinguish high-risk patients such as the length of time is too long,older,severe basic diseases,long time use of antibacterial drugs in patients,can be used in this paper,the XG boost early screening,prediction model for patients at high risk of infection in a timely manner to take preventive measures,to prevent and discover the early symptoms,Early targeted treatment is crucial.At the same time to strengthen inasieness operations such as urine tube or the tube,endotracheal intubation and tracheotomy,noninvasive ventilation,such as blood purification of prevention and control,the KP engraftment infection from the source to solve the problem,with the scientific and rational use of antibacterial drugs,avoid multiple drug-resistant strains of drug-resistant strains or affect the treatment effect,effectively improve the prognosis of patients with infection,reduce mortality.