基于多维潜变量模型和数据挖掘的溃疡性结肠炎中医证候量化方法学研究

Chinese medicine had some advantages in treatment of ulcerative colitis (UC) disease, a clinically incurable disease. However, it was lack of uniform quantitative criteria about Traditional Chinese Medicine (TCM) syndromes of UC, which resulted in the inconvenience of clinical diagnosis and clinical treatment for the disease. Latent variable theory can measure the latent variables which could not be measured directly through observed variables. Through the mathematical relationships between observed variables and latent variables, the theory can help to quantify the latent variable. Based on the researches about the early use of latent variable theory in TCM scale, our research team will draft the collection form of UC syndromes, based on literature research, the expert advices and the syndromes theory of TCM. This study will carry out the cross-over study to investigate the UC syndromes of the patients, utilize latent variable models of structural equation modeling and item response theory, construct a suitable multi-classification multidimensional latent variable model to calculate standardized syndrome score for UC patients, and use decision tree method and ROC curve to determine the diagnostic threshold for TCM syndrome. What is more, we use SAS software to establish scoring system of TCM syndromes, and validate the latent variable model. This study will eventually establish the syndromes quantitative indicators, which correspond with the theory of TCM, and embody the features of ulcerative colitis. This study will provide a scientific basis to quantity and standardize the syndromes, in order to improve the diagnosis level of Chinese medicine, and establish an innovative method for syndromes quantitative research.

中医药治疗临床重大疑难疾病溃疡性结肠炎富有特色，但其中医证候缺乏统一的量化标准，给疾病临床诊治带来不便。潜变量理论能将难于直接测定的变量以外显变量表现，通过建立外显变量和潜变量的数学关系，实现潜变量的定量化测量。本课题在前期运用潜变量理论开展中医量表量化研究的基础上，以溃疡性结肠炎为对象，通过文献研究，结合专家咨询和证候理论制定初步采集表；开展临床流行病学调查，应用结构方程模型和条目反应理论等现代多维潜变量理论方法，构建多分类多维条目反应理论模型，建立证候潜变量与外显变量的数学逻辑公式，计算患者的标准化证候得分；采用决策树方法和ROC曲线确定证候的诊断阈值；使用SAS软件编制分析程序，建立证候的评分系统，并对模型进行验证。本研究最终将建立符合中医规律、体现溃疡性结肠炎特色的证候量化指标，为其证候的量化、客观化和规范化提供科学依据，提高证候的诊断水平，并建立一套创新的证候定量化研究方法。

中医证候具有潜变量的特征，虽然不能直接被测量，但可以通过症状等外显指标表达出来。本研究以溃疡性结肠炎（UC）为例，采用定量、定性研究方法和中医理论结合的方法来量化研究中医证型。. UC常见中医证型均来源于共识、标准及教材；经定性优化条目后，共形成大肠湿热、脾虚湿蕴、脾虚湿热证、寒热错杂等证型，105个条目的采集表。临床调查共收集770例UC患者，大肠湿热证249例、脾虚湿蕴证123例、脾虚湿热证117例、寒热错杂证106例、其它证候175。使用频数法、克朗巴赫系数法、相关系数法、因子分析法并结合专业知识，最终确定大肠湿热证的条目。大肠湿热证包括共性模块、热模块和湿模块。共性模块包括腹泻、粘液、脓血便、腹痛4个条目；热模块包括肛门灼热、舌红、黄苔、发热、口干5个条目；湿模块包括苔腻和排便不爽2个条目。结构方程模型得到模型的拟合指数为较优。项目反应理论单参数Logistic模型得到难度参数区间在[-3,3]之间，所有条目质量较好。每个患者的标准化证型得分即被试者的能力值，介于[-1.81，1.09]，标准差为[0.40，0.63]。以专家辩证作为金标准，本模型得分作为诊断得分，构建ROC曲线，根据灵敏度和特异度确定诊断阈值为7，曲线下面积为0.919，P<0.01。采用144例患者进行验证，得到敏感度为0.810，特异度为0. 911，Kappa值为0.784，提示模型的可靠性好。. 结构方程模型及项目反应理论可用于指导中医证候的模型及筛选条目，使用ROC曲线确定中医证候的诊断阈值，最终建立了UC大肠湿热证的中医证候诊断量表；诊断量表经临床验证得到其准确性与可靠性均较高，具备良好的诊断价值，可为医师临床辨证提供有益参考。