食管癌患者同步放化疗期间医院感染风险预警评分模型的构建及评价

Construction and evaluation of hospital infection risk early warning scoring model for esophageal cancer patients during concurrent chemoradiotherapy

目的探讨食管癌(EC)患者同步放化疗期间医院感染的风险预警评分模型的构建及评价。方法选取2019年1月~2022年8月南通市第一人民医院120例EC患者,根据同步放化疗期间是否发生感染分为感染组(43例)和未感染组(77例),采用logistic回归分析其独立危险因素,构建风险预警模型并评价其预测效能。结果120例EC患者中,医院感染率为35.83%(43/120)。多因素logistic分析结果显示,临床分期、糖尿病、年龄、住院时间、手术时间和抗菌药使用时间是EC患者同步放化疗期间发生医院感染的独立危险因素(P<0.05,OR>1)。构建的列线图评分模型ROC曲线下面积为0.952(P<0.001,95%CI:0.915~0.989),最大约登指数为0.834,灵敏度为97.70%,特异性为85.70%,建模组、验证组的校正曲线总体趋势均接近理想曲线,Hosmer-Lemeshow拟合优度检验χ2=1.485,P=0.993;决策曲线(DCA)分析显示阈为0.10~0.86时,该预测模型表现为正的净效益(模型公式Z=1.124×临床分期+0.064×年龄+0.090×住院时间+1.181×手术时间+2.530×糖尿病+0.140×抗菌药使用时间-17.686)。结论EC患者同步放化疗期间医院感染率较高,临床分期、糖尿病、年龄、住院时间、手术时间和抗菌药使用时间是其独立预测因子,以此构建的风险预警评分模型具有良好的预测效能和临床实用性。

Objective To explore the construction and evaluation of risk warning scoring model for nosocomial infection in patients with esophageal cancer(EC)during concurrent chemoradiotherapy.Methods A total of 120 EC patients in Nantong First People's Hospital from January 2019 to August 2022 were selected and divided into infection group(n=43)and non-infection group(n=77)according to whether hospital infection occurred during concurrent chemoradiotherapy.Multivariate logistic regression was used to analyze the independent risk factors,and a risk early warning model was constructed and its predictive efficacy was evaluated.Results Among 120 EC patients,the nosocomial infection rate was 35.83%(43/120).Multivariate Logistic analysis showed that clinical stage,diabetes,age,hospitalization time,operation time and antibiotic use time were independent risk factors for nosocomial infection during concurrent chemoradiotherapy in EC patients(P<0.05,OR>1).The area under the ROC curve of the nomogram scoring model was 0.952(P<0.001,95%CI:0.915~0.989),the maximum Youden index was 0.834,the sensitivity was 97.70%,and the specificity was 85.70%.The overall trend of the calibration curve of the modeling group and the verification group was close to the ideal curve.The Hosmer-Lemeshow goodness of fit testχ2=1.485,P=0.993;the decision curve(DCA)analysis showed that when the threshold was 0.10~0.86,the prediction model showed a positive net benefit(model formula Z=1.124×clinical stage+0.064×age+0.090×hospitalization time+1.181×operation time+2.530×diabetes+0.140×antibiotic use time-17.686).Conclusions The hospital infection rate of EC patients during concurrent chemoradiotherapy is high.Clinical stage,diabetes,age,hospitalization time,operation time and antibiotic use time are independent predictors.The risk early warning scoring model constructed based on this has good predictive efficacy and clinical practicability.