Determinants And Back Propagation Predictive Model Of Surgical Site Infection In Operational Patients Of Some Tertiary Comprehensive Hospital

BackgroundSurgical site infection(SSI)is the third largest hospital infection following urinary tract infection and respiratory infection,accounting for 13 to 18 percent of all in-patient hospital infections.At the same time,SSI is a common hospital infection and complication after surgery.The occurrence of SSI is a waste of health resources,causing great pain to patients,bringing a certain economic burden,and also affecting the prognosis of patients.Therefore,it is important to predict the SSI and to strengthen SSI prevention and control.PurposesDescribe the level of infection in surgical operative site of a hospital,To analyze the factors affecting the infection of surgical site and establish BP model,To predict the incidence of surgical site infection in surgical patients,Explore measures to prevent and control infection in surgical sites,To provide a basis for effective control of surgical site infection?Materials and contents1.Sample selection:Surgical patients in the hospital from January 1,2012 to December 31,2015;2.Basic description:275 cases of incision infection for the case group,and 266 cases of patients were selected into the control group according to the the ratio of 1:1 matching method with the same period,and the same unit.Surgical infection site detection of pathogen were analyzed.3.SPSS 17.0 software was used to analyze the data.Univariate analysis were used to explore the risk factors of host factors and iatrogenic factors,and then those significant factors were included into a multi-variate logistic regression model to explore possible risk factors for surgical patients with SSI.4.SSI risk prediction,and also the establishment of neural network model for SSI:1)Those patients from 2012 to 2014 were regarded as a training sample,and those in 2015 as a test sample.SSI was the independent variable in this study.If there is a serious imbalance in the data,we have expanded the number of patients infected,so as to balance the data for the model forecast;2)Based on the balance data,the Back Propagation(BP)model is established with the training data,the verification samples are tested and the model parameters are modified,and the certified samples and test samples are classified by the final classification model.Results1.Of the 78,643 surgical aseptic surgery patients,275 cases experienced SSI(0.35%);2.The prevalence of SSI superficial wound infection was 28.36%,followed by soft tissue infection,deep incision infection,and organ cavity infection.Superficial incision and deep incision infection pathogens to G-bacteria-based,soft tissue infection and organ cavity infection G-bacteria and G + bacteria similar.3.Univariate analysis showed that the incidence of surgical infection and age,whether the basic disease,acute surgery or elective surgery,incision healing grade,ASA disease grade,whether the occurrence of bleeding,incision type,surgical cases,the total length of hospital stay,the relationship with death,the manner of anesthesia,whether implanted in artificial,the time of operation of the material,and the prophylactic use of antibiotics.Multivariate analysis showed that age,surgical type,incision healing grade,whether with underlying disease,surgical procedure,operation time,ASA disease grade and total length of hospital stay were the independent risk factors for surgical site infection.According to the size of OR,the degree of influence of each factor from strong to weak followed by surgery(acute or elective),whether suffering from underlying diseases,incision type,ASA disease grade,age,length of stay,operation time,surgical cases.4.The final selection of the calibration sample network was only 0.028%,with an average error of 0.035%.Based on a ROC evaluation,the curve area of 0.943,indicating that the model has a good ability to predict.SuggestionsSSI was associated with surgery(acute or elective),whether suffering from underlying diseases,incision type,ASA disease grade,age,length of stay,operation time,surgical cases.Artificial BP model is a good predictor of hospital infection risk,and is a prospective study of nosocomial infection.It should be further applied to the information network and convenient for the establishment of hospital infection risk prediction system.To a certain extent,it could ensure that the hospital infection pre-monitoring,it also could provide criteria for medical diagnosis and management decisions.