摘要： Objective To establish an effective prediction model to evaluate the risk of central nervous system infection (CNSI) after craniotomy and to verify its feasibility. Methods A total of 1 020 patients with craniocerebral surgery in our hospital were selected. The indexes of postoperative infection were compared between the infection group (n=61) and the non infection group (n=959). Multivariate Logistic regression was used to establish risk prediction model and area test model under ROC curve to predict the effect. The effectiveness of the prediction model was preliminarily verified by 500 patients with craniocerebral operation. Results CNSI occurred in 61 cases (5.98%) of 1 020 patients undergoing craniocerebral surgery. Multivariate Logistic regression analysis showed that six risk factors including postoperative hospital stay, number of external ventricular drainage (EVD) use ≥1, EVD indwelling duration, operation duration, indwelling permanent implant and graft operation were included in the prediction model. The formula of the prediction model was as follows: postoperative CNSI=-3.025+1.354× postoperative hospital stay +1.225× number of EVD use +1.625×EVD indwelling time +1.427× operation time +1.221× implantation of permanent implants +1.218× consecutive surgery. The AUC under the ROC curve was 0.849 (95%CI: 0.761-0.915), the sensitivity was 81.56% and the specificity was 65.78%. In the preliminary validation cohort, 34 patients developed postoperative CNSI (6.8%), and the model predicted postoperative CNSI in 30 patients (6.0%), with a sensitivity of 91.48% and a specificity of 91.53%. Conclusion This model is suitable for the perioperative evaluation of patients with craniocerebral surgery, and can identify the high-risk population of postoperative CNSI in time.

中图分类号: 

• R651.1