NSCLC患者并发医院肺部感染的风险预测列线图模型构建研究

Construction of nomogram model for risk prediction of nosocomial lung infection in NSCLC patients

目的探讨非小细胞肺癌(NSCLC)患者并发医院肺部感染的影响因素,并据此构建预测列线图模型。方法回顾性分析商丘市第一人民医院2020年9月至2022年9月收治的279例NSCLC患者的临床资料,按照2∶1的比例将其分为建模组(186例)和验证组(93例),并根据患者是否并发医院肺部感染将建模组患者分为感染组和未感染组。对比感染组、未感染组临床资料;采用多因素Logistic回归性分析法分析NSCLC患者并发医院肺部感染的影响因素,并构建预测列线图模型;采用Bootstrap法进行内部验证,绘制校准曲线;采用受试者工作特征(ROC)曲线以评估列线图模型的预测效能;采用决策曲线(DCA)验证模型的临床净获益率。结果279例NSCLC患者并发医院肺部感染的有67例,其中建模组186例患者有47例并发医院肺部感染,感染率为25.27%;验证组93例患者有20例并发医院肺部感染,感染率为21.51%。感染组中年龄≥75岁、吸烟史、临床分期Ⅲ~Ⅳ期、化疗周期≥4个、化疗药物联合、住院时间>21 d、化疗前体力状况评分<80分、糖尿病、白细胞计数<4×10^(9)/L、中性粒细胞计数<1×10^(9)/L、白蛋白<30 g/L占比均高于未感染组(P<0.05);经多因素Logistic回归性分析,年龄≥75岁、吸烟史、临床分期Ⅲ~Ⅳ期、化疗周期≥4个、化疗药物联合、住院时间>21 d、化疗前体力状况评分<80分、糖尿病、中性粒细胞计数<1×10^(9)/L、白蛋白<30 g/L均是NSCLC患者并发医院肺部感染的独立危险因素(P<0.05);将上述影响因素作为预测指标,构建NSCLC患者并发医院肺部感染的风险预测列线图模型,建模组和验证组列线图模型预测校准曲线均接近标准曲线;ROC曲线分析结果显示,建模组列线图预测模型的曲线下面积(AUC)为0.839,灵敏度为82.61%,特异度为88.00%;验证组列线图预测模型AUC为0.822,灵敏度为80.43%,特异度为85.33%;建模组和验证组DCA结果均显示列线图模型具有良好的临床效益。结论年龄≥75岁、吸烟史、临床分期Ⅲ~Ⅳ期、化疗周期≥4个、化疗药物联合、住院时间>21d、化疗前体力状况评分<80分、糖尿病、中性粒细胞计数<1×10^(9)/L、白蛋白<30g/L均是NSCLC患者并发医院肺部感染的影响因素,据此构建的列线图预测模型可帮助临床早期识别高风险人群,临床实用性高。

【Objective】To investigate the influencing factors of nosocomial lung infection in patients with non-small cell lung cancer(NSCLC),and to construct a predictive nomogram model.【Methods】Clinical data of 279 patients with NSCLC admitted to the First People's Hospital of Shangqiu City from September 2020 to September 2022 were retrospectively analyzed,and they were divided into modeling group(186 cases)and validation group(93 cases)according to a ratio of 2:1.Patients of the modeling group were also divided into infected group and uninfected group according to whether they had nosocomial lung infection.The clinical data of the infected group and the uninfected group were compared.Multivariate Logistic regression analysis was used to analyze the influencing factors of nosocomial lung infection in patients with NSCLC,and a predictive nomogram model was constructed.Bootstrap method was used for internal verification and calibration curve was drawn.Receiver operating characteristic(ROC)curve was used to evaluate the predictive efficiency of the nomogram model.Decision curve analysis(DCA)was used to verify the clinical net benefit rate of the model.【Results】Among the 279 NSCLC patients,67 cases were complicated with nosocomial lung infection.Among the 186 patients in the modeling group,47 cases were complicated with nosocomial lung infection,and the infection rate was 25.27%.In the verification group,20 cases of the 93 patients had nosocomial lung infection,and the infection rate was 21.51%.The proportions of age≥75 years old,having smoking history,clinical stageⅢ–Ⅳ,chemotherapy cycle≥4,combination of chemotherapy drugs,length of hospital stay>21 days,physical condition score before chemotherapy<80 points,diabetes mellitus,white blood cell count<4×10^(9)/L,neutrophil count<1×10^(9)/L,albumin<30 g/L in infection group were higher than that in non-infection group(P<0.05).After multivariate Logistic regression analysis,age≥75 years old,having smoking history,clinical stageⅢ–Ⅳ,chemotherapy cycle≥4,combination of chemotherapy drugs,length of hospital stay>21 days,physical condition score before chemotherapy<80 points,diabetes mellitus,neutrophil count<1×10^(9)/L and albumin<30 g/L were all independent risk factors for nosocomial lung infection in NSCLC patients(P<0.05).The above influencing factors were used as predictive indicators to construct the risk prediction nomogram model of NSCLC patients complicated with nosocomial lung infection.The nomogram model prediction calibration curves of both the modeling group and the verification group were close to the standard curve.ROC curve analysis results showed that the area under the curve(AUC)of the nomogram prediction model of the modeling group was 0.839,and the sensitivity was 82.61%,and the specificity was 88.00%.The AUC of nomogram prediction model of the validation group was 0.822,and the sensitivity was 80.43%,and the specificity was 85.33%.The DCA results of both the modeling group and the validation group showed that the nomogram model had good clinical benefit.【Conclusion】Age≥75 years old,having smoking history,clinical stageⅢ–Ⅳ,chemotherapy cycle≥4,combination of chemotherapy drugs,length of stay>21 days,physical condition score before chemotherapy<80 points,diabetes mellitus,neutrophil count<1×10^(9)/L,albumin<30 g/L are all factors affecting nosocomial lung infection in NSCLC patients.The prediction model based on the column graph can help to identify high-risk groups in early clinical stage and has high clinical practicability.