| ObjectivesThis study aims to explore the current status of nosocomial infection in patients after cranial nerve microvascular decompression surgery comprehensively,investigate and analyze the risk factors of nosocomial infection,to construct the Logistic regression model and the decision tree risk prediction model of nosocomial infection.At the same time,the predictive effects of the two models were validated.The study is intended to be used for the screening and management of patients at high risk of nosocomial infection in microvascular decompression surgery,as well as to guide medical and nursing staff in the department to adopt scientific and effective interventions,thereby reducing the rate of nosocomial infection and providing an evidence-based basis for the development of nosocomial infection prevention and control initiatives.MethodsIn this study,907 patients who underwent cranial nerve microvascular decompression in a Grade A hospital in Shandong Province from January 1,2021 to December 31,2022 were selected.With the help of the hospital information system and nosocomial infection surveillance system,combined with bedside rounds,relevant research data were collected by using the self-designed "Targeted Surveillance Questionnaire for Nosocomial Infection in Patients with Microvascular Decompression Surgery".According to a ratio of 7:3,The subjects were randomly divided into the modeling group(635 cases)and the validation group(272 cases).According to the results of univariate analysis of the modeling group,variables with statistically significant differences(P<0.05)were used to construct the Logistic regression prediction model.The model was evaluated by the data of the modeling group and verified by the data of the validation group.The model differentiation was evaluated by the area under receiver operating characteristic curve(AUC),and the fitting effect was tested by Hosmer-Lemeshow.Variables with statistical differences in univariate analysis were used to build a decision tree risk prediction model by using the CHAID algorithm.The evaluation was carried out by AUC and risk statistics,and the verification was carried out by validation group data.Finally,we compared the accuracy,sensitivity,specificity and Youden index of the two models,and we compared the AUC of the two models by the Z-test and to analyze whether there were statistical differences in the prediction effects of the two models.Results1.A total of 907 patients undergoing cranial nerve microvascular decompression were included in this study,including 320 males(35.30%)and 587 females(64.70%),with an average age of 55.01 ± 10.555 years.Among them,379(41.80%)were operated on in 2021 and 528(58.20%)were operated on in 2022.There were 635 patients in the modeling group,with an average age of 55.06±10.715 years old,including 224 males,accounting for 35.28%.There were 272 patients in the validation group,with an average age of 54.90±10.191 years old,among which males accounted for about 35.29%.2.In this study,a total of 89 patients had postoperative nosocomial infection,a total of 102 infections,the nosocomial infection rate was 9.80%,the incidence rate of nosocomial infection cases was 11.25%.In the modeling group,the total postoperative nosocomial infection rate was 9.60%,and the incidence rate of nosocomial infection cases was 11.34%.In the validation group,the total postoperative nosocomial infection rate was 10.30%,and the incidence rate of nosocomial infection cases was 11.03%.3.In this study,there were 44 cases of surgical site infection,the composition ratio was 49.44%,including 38 cases(42.70%)of organ space infection and 6 cases(6.74%)of superficial incision infection.Pulmonary infection(22.47%)was the second highest with a 20 cases.The other infections were 6 cases(6.74%)of urinary tract infection,6 cases(6.74%)of herpes simplex virus infection,1 case(1.12%)of upper respiratory tract infection,and 12 cases(13.48%)of multiple site infection.In the model group,there were 29 cases(47.54%)of surgical site infection,14 cases(22.96%)of pulmonary infection,5 cases(8.20%)of urinary tract infection,2 cases(3.27%)of herpes simplex virus infection,1 case(1.64%)of upper respiratory tract infection,and 10 cases(16.40%)of multiple site infection.In the validation group,15 patients(53.57%)had surgical site infection,6 patients(21.43%)had pulmonary infection,1 patient(3.57%)had urinary tract infection,4 patients(14.29%)had herpes simplex virus infection,and 2 patients(7.14%)had multiple site infection.4.16 strains of pathogenic bacteria were cultured in all of the specimens submitted for examination,among which 14 strains were Gram-positive,the composition ratio was 87.50%.There were 4 strains of Staphylococcus hominis(25.00%),3 strains of Staphylococcus epidermidis(18.75%),1 strain of Streptococcus cristae,1 strain of Staphylococcus saprophyticus,1 strain of Streptococcus Gordonii,1 strain of Staphylococcus capitis,1 strain of Micrococcus Cohn,1 strain of Staphylococcus haemolyticus,and 1 strain of Staphylococcus warneri(6.25%).Two strains of Gram-negative bacteria were cultured,the composition ratio was 12.50%,including 1 strain of Acinetobacter baumannii(6.25%)and 1 strain of Moraxella catarrhalis(6.25%).5.Logistic regression analysis selected 6 independent risk factors for nosocomial infection:actual hospital stay,bone flap reduction,body temperature on the 5th day after surgery,duration of glucocorticoid use,mastoid opening,days of continual drainage through lumbar subarachnoid space catheter.The Logistic regression model was established according to the independent risk factors,and the AUC was 0.918(95%CI:0.881～0.954),the prediction accuracy was 92.50%,the sensitivity was 0.850,and the specificity was 0.849.Hosmer-Lemeshow test suggested that the model had a high goodness of fit.The ROC curve showed that the model had good differentiation and prediction effect.6.Statistically significant variables by univariate analysis were used to build a decision tree model,and 5 variables were screened out,including actual hospital stay,mastoid opening,temperature on the 5th day after surgery,bone flap reduction,and the number of lumbar puncture procedures.Altogether 7 decision rules were generated.The AUC of the model group was 0.906(95%CI:0.875～0.938),indicating that the model was well differentiated,the prediction accuracy was 91.70%,the sensitivity was 0.918,and the specificity was 0.808.Validation group was used to verify the stability and prediction effect of the proposed decision tree model.7.The AUC of Logistic regression model and a decision tree risk prediction model was compared,Z=0.875,P=0.3878,and the difference was not statistically significant.ConclusionsThe incidence of nosocomial infection in patients undergoing cranial nerve microvascular decompression surgery is high,and there are various types of infection.Surgical site infection and pulmonary infection are the main sites of infection.Actual hospital stay,bone flap reduction,body temperature on the 5th day after surgery,duration of glucocorticoid use,mastoid opening,days of continual drainage through lumbar subarachnoid space catheter are independent risk factors for nosocomial infection after cranial nerve microvascular decompression.Scientific and effective measures should be taken as soon as possible to reduce the incidence of nosocomial infection.In this study,Logistic regression prediction model and decision tree risk prediction model were constructed.The models have good differentiation and prediction effect,which can be used as the basis to achieve early prediction,early diagnosis and early treatment of high-risk groups of nosocomial infection,so as to achieve the effect of controlling nosocomial infection. |
PDF Full Text Request