| ObjectiveTo explore the influencing factors of nosocomial infection in patients after gastrointestinal surgery,to establish a simple and effective risk prediction model which is convenient for clinical implementation and application,and to verify the prediction effect of the model.Methods1.Determine the candidate predictive variables.Computer search of CNKI,Wanfang Data Knowledge Service Platform,VIP Database,CBM,Pub Med,Embase,Web of Science,Cochrane Library.The search time was from the database establishment to February 2022.Two researchers independently screened the literatures according to inclusion and exclusion criteria,and extracted the data.Rev Man5.4software was used for meta-analysis of the data,so as to obtain a comprehensive and reliable risk factor of nosocomial infection after gastrointestinal surgery.The candidate predictive variables of the model were determined according to the results of the meta-analysis and expert opinions,and the Collection Table of Hospital Infection Information for Patients after Gastrointestinal Surgery was designed.2.The development of risk prediction model.A retrospective nested case-control study was adopted in this study.The patients undergoing gastrointestinal surgery in three hospital districts of a class III A hospital in Guangzhou from July 1,2016 to June 30,2022 were selected as the research subjects.The patients who underwent gastrointestinal surgery from July 1,2016 to June 30,2021 were used as the modeling group for retrospective analysis.Case data were collected according to the Collection Table of Hospital Infection Information for Patients after Gastrointestinal Surgery.In the modeling group,the patients were divided into an nosocomial infection group and a non-nosocomial infection group according to whether nosocomial infection occurred or not.Propensity score matching was used to pair the nosocomial infection group and the non-nosocomial infection group in a ratio of 1:1 with admission time and department as covariates.Lasso regression was used to screen the variables,and Logistic regression was used to construct the risk prediction model of nosocomial infection in patients with gastrointestinal postoperative,and the model was visualized by nomogram and web calculator.3.The validation of risk prediction model.Internal validation: internal validation was performed in the modeling group using the enhanced Bootstrap method to test the repeatability of the model development process.External validation: patients who underwent gastrointestinal surgery in three hospital districts of a class III A hospital in Guangzhou from July 1,2021 to June 30,2022 were used as the external validation group.Temporal validation was used to test the portability and generalization ability of the model over different time periods.4.The evaluation of risk prediction models.The discrimination of the model was reflected by calculating the C-index,drawing the curve of the receiver operating characteristic and calculating the area under the curve.The Calibration plot was drawn to reflect the calibration of the model,and the clinical applicability of the model was reflected by decision curve analysis.Results1.The results of candidate predictive variables.A total of 30 articles were finally selected involving 32630 patients undergoing gastrointestinal surgery.Meta-analysis results showed that there were 22 risk factors for nosocomial infection in patients undergoing gastrointestinal surgery:Advanced age,male,body mass index,NNIS risk index,ASA≥3,plasma albumin,diabetes,COPD,preoperative anemia,surgical incision length,operation duration,laparotomy operation,emergency operation,surgical incision grade ≥Ⅲ,length of stay,operating room without laminar flow,retention of drainage tube,intraoperative blood loss,no preventive medication before surgery,the use of antibacterial drugs is not standardized,application of hormones and total gastrectomy(P<0.05).The protective factors of nosocomial infection after gastrointestinal surgery were laparoscopic or robotic surgery(P<0.05).In addition,total parenteral nutrition is not associated with nosocomial infection after gastrointestinal surgery(P>0.05).Based on the results of the meta-analysis,seven new influencing factors were added according to the experts opinions,including whether to be admitted to the ICU,whether to use the ventilator,whether to intubate the central vein,whether to indwell the urinary catheter,whether to intubate the trachea,whether to indwell the gastric tube,and whether to suffer from malignant tumor.According to the objective situation of study site and expert opinion,the 5 influencing factors including NNIS risk index,body mass index,laminar flow-free operation room,nonstandard use of antibacterial drugs and application of hormones were deleted.Finally,24 influencing factors are included in the candidate predictive variables of the model,which consists of three parts.The first part is the patient-related factors,including age,sex,ASA ≥ 3,low preoperative Alb and low preoperative Hb,with a total of five items.The second part is the disease-related factors,including diabetes,COPD and malignant tumor,which consists of three items.The third part was about treatment-related factors,including length of stay,admission to ICU,central vein catheters,indwelling urinary catheter,tracheal cannula,use of ventilator,operation time,type of incision ≥ III class,incision length,intraoperative blood loss,emergency surgery,total gastrectomy,retention of drainage tube,indwelling gastric tube,preventive use of antibacterial drugs,and laparotomy,a total of 16 items.2.The results of the model development.In the modeling group,according to the inclusion and exclusion criteria,9,196 patients with gastrointestinal surgery were finally included,including 244 patients with nosocomial infection and 8,952 patients without nosocomial infection,with the incidence of nosocomial infection of 2.65%.With the admission time and department as covariates,244 pairs were successfully matched by 1:1 propensity score.After matching,a total of 488 cases were included in the modeling group,including 244 cases in the nosocomial infection group and244 cases in the non-nosocomial infection group.Influencing factors of Lasso regression final modeling include: length of stay≥15.5days,stay in ICU,surgical incision length≥10 cm,malignant tumor,and low preoperative albumin.According to the results of Logistic regression analysis,the formula for establishing the risk prediction model of nosocomial infection in patients after gastrointestinal surgery was as follows:Z=2.7463×value of length of stay+ 1.7143× value of stay in ICU+0.7003×value of surgical incision length+1.7199×value of malignant tumor+1.2406×value of low preoperative albumin.Based on this formula,a nomogram and web calculator are developed.3.The results of the model validation and evaluation.Using the enhanced Bootstrap method,internal verification is carried out by repeated sampling for 1000 times.The C-index value of the model was 0.932,and the area under the ROC curve(AUC)was 0.9352(95% CI = 0.9133～0.957).According to the Calibration plot,the risk probability of nosocomial infection predicted by this model is in good agreement with the actual risk probability.When the threshold probability of the model is in the range of 1%～52%,the model has better clinical applicability.In the external validation group,according to the inclusion and exclusion criteria of the study,a total of 2205 patients underwent gastrointestinal surgery,of which 51 cases had nosocomial infection and 2154 cases had no nosocomial infection,with the incidence of nosocomial infection of 2.31%.The 50 pairs were successfully matched through 1:1propensity score with the admission time and department as covariates.After matching,a total of 100 cases in the external validation group,including 50 cases in the nosocomial infection group and 50 cases in the non-nosocomial infection group.In external verification,the C-index value of the model was 0.890,and the area under the ROC curve(AUC)was0.8862(95% CI = 0.8187～0.9537).According to the Calibration plot,the risk probability of nosocomial infection predicted by this model is in good agreement with the actual risk probability.When the threshold probability of the model is in the range of 1%～14%,the model has better clinical applicability.Conclusion1.According to the results of the meta-analysis and expert opinions,24 candidate predictive variables for nosocomial infection in patients after gastrointestinal surgery were ultimately included in this study,which included three dimensions:patient-related factors,disease-related factors and treatment-related factors.2.The risk prediction model of nosocomial infection in patients with gastrointestinal postoperative was developed and validated.The model includes five variables: length of stay≥15.5 days,stay in ICU,surgical incision length ≥10 cm,malignant tumor and low preoperative albumin.The model has good discrimination,calibration and clinical practicability in internal verification and external verification.The nomogram and the web calculator are developed and produced,which is helpful for clinical medical staff to predict the probability of nosocomial infection after gastrointestinal surgery more accurately and conveniently,identify high-risk patients as early as possible,take timely intervention measures and reduce the occurrence of nosocomial infection. |
PDF Full Text Request