头颈部恶性肿瘤根治术预防性气管切开并发肺部感染危险因素及其预测模型构建

Risk factors for pulmonary infection in patients with head and neck malignant tumors undergoing radical surgery and preventive tracheotomy and establishment of prediction model

目的 探究头颈部恶性肿瘤根治术预防性气管切开并发肺部感染的危险因素并构建预测模型。方法 回顾性纳入徐州市中心医院2018年3月-2023年3月局部中晚期口腔癌根治术预防性气管切开的患者138例为研究对象,按照3∶1比例,用R语言软件随机分为训练集(n=104)和验证集(n=34)。将训练集纳入统计分析,分为肺部感染组(n=53)和非感染组(n=51)。采用多因素Logistic回归分析训练集患者气管切开后肺部感染危险因素,用R软件构建列线图预测模型,并绘制该模型受试者工作特征(ROC)曲线、临床决策曲线和校准曲线,用验证集数据对构建的模型进行验证。结果 Logistic回归法分析显示,糖尿病史、套管留置时间延长为危险因素(P<0.05),预后营养指数(PNI)升高为保护因素(P<0.05)。列线图模型的ROC曲线下面积在训练集与验证集中分别为0.813和0.858,模型内部验证曲线拟合良好。结论 该列线图模型对于预测头颈部恶性肿瘤根治术预防性气管切开患者术后发生肺部感染有较好的临床应用价值。

OBJECTIVE To explore the risk factors for pulmonary infection in the patients with head and neck ma-lignant tumors undergoing radical surgery and preventive tracheotomy and establish the prediction model.METHODS Totally 138 patients with locally advanced oral carcinoma who underwent radical surgery and preven-tive tracheotomy in Xuzhou Central Hospital from Mar 2018 to Mar 2023 were recruited as the research subjects and were,in a 3∶1 ratio,randomly divided into the training set with 104 cases and the validation set with 34 ca-ses by using R language software.The patients of the training set were brought into statistical analysis and were divided into the pulmonary infection group with 53 cases and the non-infection group with 51 cases.Multifactor lo-gistic regression analysis was performed for the risk factors for the pulmonary infection in the tracheotomy pa-tients of the training set.A nomogram prediction model was built by using R software;the receiver operating characteristic(ROC)curve,clinical decision curve and calibration curve of the model were drawn,and the predic-tion model was validated by using the validation set data.RESULTS The logistic regression analysis showed that the history of diabetes mellitus and extension of cannula indwelling time were the risk factors(P<0.05),and the rise of prognosis nutrition indexes was the protective factor(P<0.05).The area under ROC curve of nomogram model was 0.813 in the training set,0.858 in the validation set,indicating that the internal validation curve of the model fit well.CONCLUSION The nomogram model has high clinical value in prediction of the postoperative pul-monary infection of the patients with head and neck malignant tumors undergoing radical surgery and preventive tracheotomy.