基于多因素分析构建口腔癌患者认知衰弱的风险预测模型及其临床验证  

Construction and clinical validation of a risk prediction model for cognitive decline in oral cancer patients based on multiple factor analysis

作　　者：高阳 唐慧[2] 蒋小剑[1] GAO Yang;TANG Hui;JIANG Xiaojian(School of Nursing,Hunan University of Chinese Medicine,Changsha 410000,China;Department of Oral and Maxillofacial Surgery,the Second Xiangya Hospital of Central South University,Changsha 410011,China)

机构地区：[1]湖南中医药大学护理学院,湖南长沙410000 [2]中南大学湘雅二医院口腔颌面外科,湖南长沙410011

出　　处：《东南大学学报(医学版)》2025年第2期279-286,共8页Journal of Southeast University(Medical Science Edition)

摘　　要：目的:基于多因素分析结果构建口腔癌患者认知衰弱的风险预测模型,并验证该模型的预测效果。方法:选取三级甲等医院270例口腔癌患者作为研究对象,按7∶3比例将患者分为建模集(n=189)和验证集(n=81)。在入院确诊1周内采用Fried衰弱表型评分记录认知衰弱情况。采用多因素Logistic回归模型分析建模集患者认知衰弱的相关因素,并采用R语言构建列线图预测模型,以Bootstrap法对模型进行内部验证。基于验证集数据,采用受试者工作特征(ROC)曲线对模型进行验证。结果:建模集中72例发生认知衰弱,发生率为38.10%;验证集中31例发生认知衰弱,发生率为38.27%。Logistic多因素分析显示,年龄≥60岁、饮酒史、日常锻炼、使用ARB类药物、Barthel指数≥95分及营养不良均是口腔癌患者认知衰弱的独立影响因素(P<0.05)。基于多因素分析结果建立Nomogram列线图预测模型,绘制校准曲线,结果显示风险预测值与实际预测值的平均绝对误差为0.018。ROC曲线分析结果显示,列线图判断建模集患者发生认知衰弱的AUC为0.851(SE=0.029,95%CI 0.795~0.907,P<0.05),敏感度为0.861,特异度为0.744。列线图判断验证集口腔癌患者发生认知衰弱的AUC为0.877(SE=0.038,95%CI 0.803~0.952,P<0.001),敏感度为0.935,特异度为0.720。结论:口腔癌患者认知衰弱风险较高,认识衰弱的发生与患者年龄、饮酒史、日常锻炼和使用ARB类药物情况、Barthel指数及营养状况相关,基于上述因素建立的列线图预测模型对判断口腔癌患者认知衰弱风险具有较高准确性。Objective:To construct risk prediction model for cognitive decline in oral cancer patients based on the results of multiple factor analysis,and to verify the predictive performance of the model.Methods:270 oral cancer patients from tertiary hospitals were selected as the research subjects,and the patients were divided into a modeling set(n=189)and a validation set(n=81)in a 7∶3 ratio.The cognitive impairment was recorded by the Fried frailty phenotype score within one week of admission diagnosis.Logistic multiple regression model was used to analyze the factors related to cognitive decline in the modeling set of patients,and column chart prediction model was constructed by R language,the model was internally validated using Bootstrap method.The model was externally validated by receiver operating characteristic(ROC)curve of the validation set data.Results:There were 72 cases in the modeling set experienced cognitive impairment,with an incidence rate of 38.10%.31 cases of cognitive impairment occurred in the validation set,with an incidence rate of 38.27%.Logistic multivariate analysis showed that age≥60 years old,alcohol consumption history,daily exercise,use of ARB drugs,Barthel index≥95 points,and malnutrition were independent influencing factors for cognitive decline in oral cancer patients(P<0.05).Based on the results of multiple factor analysis,a Nomogram column chart prediction model was established,and a calibration curve was drawn.The results showed that the average absolute error between the risk prediction value and the actual prediction value was 0.018.The ROC curve analysis results showed that the AUC of Nomogram to judge cognitive impairment in patients in the modeling set was 0.851(SE=0.029,95%CI 0.795-0.907,P<0.05),the sensitivity was 0.861,and the specificity was 0.744.The ROC curve analysis results showed that the AUC of Nomogram to judge cognitive impairment in patients in the validation set was 0.877(SE=0.038,95%CI 0.803-0.952,P<0.001),the sensitivity and specificity were 0.935 and 0.720,

关 键 词：口腔癌 认知衰弱 预测模型 

分 类 号：R739.8[医药卫生—肿瘤]