肺癌化疗患者口腔衰弱预测模型的构建及验证 [中文引用][英文引用]

目的  调查肺癌化疗患者口腔衰弱（oral frailty，OF）的现状，分析其影响因素并构建风险预测模型。方法  采用便利抽样法，选取 2024 年 9 月至 11 月江苏省 3 所三级甲等综合医院收治的 431 例肺癌化疗患者作为训练集，分为口腔衰弱和非口腔衰弱组，对两组相关资料进行比较，通过多因素 Logistic 回归分析确定 OF 的影响因素，并据此构建风险预测模型。采用受试者操作特征（receiver operating characteristic，ROC）曲线下面积检验模型预测效果。选取 2024 年 12 月江苏省 1 所三级甲等综合医院的 185 例患者进行预测效果验证。结果  肺癌化疗患者 OF 发生率为 58.93%。高龄（OR=3.420）、低文化程度（OR=0.030）、脑转移（OR=7.880）、高营养风险筛查 2002 评分（OR=1.550）、高 C- 反应蛋白水平（OR=1.100）以及高乳酸脱氢酶水平（OR=1.010）是OF的危险因素（均P<0.05）。构建的预测模型建模集ROC曲线下面积为 0.860（95%CI：0.830~0.900），验证集 0.840（95%CI：0.780~0.900）。Hosmer-Lemeshow 检验训练集 χ 2=4.870，P=0.770；验证集 χ 2=2.770，P=0.950。结论  本研究构建的肺癌化疗患者 OF 风险预测模型具有良好的预测效能，有助于早期识别高风险患者，为临床干预提供科学依据。

Objective To investigate the status of oral frailty（OF）in patients who underwent chemotherapy for lung cancer， identify key factors influencing OF，and develop a risk prediction model. Methods Using convenience sampling，431 lung cancer inpatient were recruited from three Tier-IIIA hospitals in Jiangsu Province between September and November 2024 as the training cohort. The patients were divided into OF and non-OF groups. Relevant data were compared between the two groups. Multifactorial logistic regression analysis was performed to determine factors that associated with OF，and a risk prediction model was created accordingly. Receiver operating characteristic（ROC）curve analysis was used to predict model performance. In December 2024，additional 185 patients from one other Tier-IIIA hospitals were recruited to validate the developed model. Results The prevalence of OF among lungcancer patients undergoing chemotherapy was 58.93%. Following listed items were identified as the risk factors of OF（all P<0.05）：older in age（OR=3.420），poor education（OR=0.030），brain metastasis（OR=7.880），high nutritional risk screening 2002 score（OR=1.550）， elevated C-reactive protein（OR=1.100），and elevated lactate dehydrogenase（OR=1.010）. ROC area under the curve（AUC）of the model was 0.860（95% CI：0.830-0.900）in modelling cohort and 0.840（95% CI：0.780-0.900）in validation cohort. Hosmer-Lemeshow goodness-of-fit test yielded χ 2 =4.870，P=0.770 for the training set and χ 2 =2.770，P=0.950 for the validation set. Conclusion The risk prediction model for OF developed in this study demonstrates a good predictive performance and can facilitate early identification of highrisk patients，thereby providing a scientific basis for clinical interventions.