目的 探究直肠癌患者术后发生医院感染的病原菌分布，并构建发生医院感染的预测模型。 方法 回顾性收集1 537例在2021年1月至2022年12月于浙江大学医学院附属邵逸夫医院进行直肠癌手术治疗的患者，并根据其是否发生医院感染，采用倾向性评分匹配法（propensity score matching，PSM）进行1∶1匹配，最终纳入感染组及对照组各83例。对发生医院感染的患者，进行菌种分布及耐药性分析。采用最小绝对收缩和选择算子（least absolute shrinkage and selection operator，LASSO）回归筛选直肠癌患者术后发生医院感染的危险因素，运用多因素logistics回归建立预测模型，构建Nomogram图，利用受试者工作特征（receiver operating characteristic，ROC）曲线、校准曲线和决策曲线分析评估模型的预测价值。 结果 83例医院感染患者中，共分离病原菌93株，其中87株为细菌，包括革兰阴性菌62株（66.67%），以大肠埃希菌和铜绿假单胞菌为主；革兰阳性菌25株（26.88%），以粪肠球菌为主；真菌6株（6.45%），均为白假丝酵母菌。通过LASSO回归筛选变量并使用多因素logistics回归建立模型，结果显示吸烟史（OR=3.97，95%CI=1.27~12.43）、引流管置管时间（OR=1.19，95%CI=1.08~1.30）、手术前后中性粒细胞差值（OR=1.23，95%CI=1.01~1.49）、手术前后C反应蛋白差值（OR=1.05，95%CI=1.03~1.07）为直肠癌患者术后发生医院感染的独立危险因素。基于以上因素构建的列线图模型ROC曲线下面积为0.933（95%CI=0.896~0.969）。校准曲线分析显示，该模型的实际、校正曲线拟合好，且接近于理想曲线。决策曲线分析结果提示，该模型具有良好的临床效用和较高的净临床效益。 结论 本研究构建的预测模型对直肠癌患者术后发生医院感染具有良好的预测价值。

Objective To examine the distribution of pathogens that cause postoperative nosocomial infections in patients with rectal cancer（RC） and to construct a predictive nomogram for nosocomial infection. Methods The clinical data of 1537 RC patients admitted to Sir Run Run Shaw Hospital between January 2021 and December 2022 were collected. Patients were assigned 1∶1 by propensity score matching（PSM） to the infection group（n=83） and control group（n=83） based on the occurrence of nosocomial infection. The distribution and drug resistance of bacteria in patients with nosocomial infection were analyzed. Risk factors for postoperative nosocomial infection were identified by least absolute shrinkage and selection operator（LASSO） regression，and a predictive nomogram was constructed using multivariate logistics regression. The predictive performance of the model was evaluated by receiver operating characteristic（ROC） curve，calibration curve，and decision curve analysis（DCA）. Results A total of 93 strains of pathogens were isolated from the 83 infected patients，including 62 strains of Gram-negative bacteria （66.67%；predominantly Escherichia coli and Pseudomonas aeruginosa），25 strains of Gram-positive bacteria（26.88%； mainly Enterococcus faecalis），and 6 strains of fungi（6.45%； all Candida albicans）. LASSO and multivariate logistics regression showed that smoking （odds ratio［OR］=3.97，95%CI=1.27-12.43），the dwelling time of drainage tube（OR=1.19，95%CI=1.08-1.30），difference in preoperative and postoperative neutrophil counts（OR=1.23，95%CI=1.01-1.49），and difference between preoperative and postoperative C-reactive protein levels（OR=1.05，95%CI=1.03-1.07） were independent risk factors for postoperative nosocomial infection in RC patients. The area under the ROC curve of the nomogram constructed based on the above factors was 0.933（95%CI=0.896-0.969）. The calibration curve showed that the predicted risk was in good agreement with the actual observed risk of infection. In addition，DCA demonstrated that the nomogram has good clinical utility and high net clinical benefits in predicting nosocomial infection. Conclusion The nomogram constructed in this study has a good predictive performance in postoperative nosocomial infection in RC patients.