诊断分类模型在中医证候量化及临床诊断的应用研究

Hypertension, the common disease of western medicine, is the main research object of this article. Hypertension in the traditional Chinese medicine has two features. One is the indirectly observed nature of syndrome and syndrome elements, and the other is the four physical examinations information which physicians collected is in the form of classified or ranked. Because of the two features, diagnostic classification model is used to study the TCM syndromes elements extraction and individualized diagnosis.. Firstly, polychoric correlation matrix will be used in the exploratory factor analysis and structural equation model to get the preliminary syndrome analysis, the relationship between syndrome research and four physical examinations information, and the relationship between syndrome itself. Meanwhile, data simulation will be done to explore the difference between traditional structural equation model and polychoric-used structural equation model under various conditions. And the practical application in clinical conditions will be theoretically discussed. With the construction of Chinese-medicine-expert consultation questionnaire and the result from the structural equation model, the different Q-matrix of the diagnostic model can be gotten. Then, diagnostic classification models will be used to explore hypertension syndrome elements extraction and statistical diagnosis of the individual. The methods may provide new ideas and methods for the research of TCM,the result of the research have reference value to the theory and clinical practice of the traditional Chinese medicine, clinical diagnosis, and the process of latent variables of classified data in efficacy evaluation. It also has theoretical guiding significance in application condition of Chinese medicine research for medical statistics.

课题以西医高血压病为病种，针对中医证候、证候要素不可直接观测的性质，医师采集病人的四诊信息为两分类或等级分类的特点，以诊断分类模型研究高血压病的中医证候要素提取及个体化诊断。首先采用多项相关系数下探索性因子分析及结构方程模型对高血压病的中医证候进行初步分类、研究证候与四诊信息关系及证候间的关系。同时，通过数据模拟，以期得到不同条件下，等级资料利用多项相关系数与传统结构方程模型的结果是否存在差异，从理论上探讨临床实际中的应用条件。通过中医专家咨询调查表的构建及结构方程模型结果，得到诊断模型中不同的Q-矩阵。然后，采用诊断分类模型探索高血压病的中医证候要素提取及对个体进行统计诊断。研究结果所建立的方法，可能为中医量化提供了新的思路与方法。不仅对中医学理论和临床实践有参考价值，对临床医学诊断及疗效评定中分类资料的潜在变量问题处理都具有参考价值，对于医学统计学在中医研究应用条件有其理论上的指导意义

中医收集的四诊信息往往是两分类或等级资料，无法满足因子分析、结构方程等模型的显在变量与潜在变量均为连续变量的前提条件。期望通过基于多项相关系数结构方程模型的拟合、修正及优化，解决高血压病证候的主要四诊信息与证候要素和证候间的关系；以诊断分类模型研究高血压病的中医证候要素提取及个体化诊断。.研究结果表明，高血压病主要的中医证型可以分为五类，分别为：肝火亢盛、痰浊内蕴、肝肾阴虚、阴虚阳亢、心肾两虚。对比两种探索性因子分析方法的结果，传统协方差的方法与多项相关性系数的方法的结果差别不大。但多项相关系数方法的模型拟合指数结果略优。二阶证实性因子分析结果表明，高血压病的5个证候分别可以提取2-3个证候要素，并从病位和病性两方面对高血压病的证候要素进行命名，其中病位类证候要素为肝、肾、脾、肝肾、心肾等，病性类证候要素为气虚、气郁、阴虚、阴虚、阳亢、实热、湿盛等。.Monte Carlo的模拟结果为：在处理分类或者等级资料时，基于多项相关系数的证实性因子分析与结构方程模型在数据样本量较小、因子载荷系数较小、模型因子数较多时，优于基于协方差的证实性因子分析与结构方程模型；当样本量较大、因子载荷系数较大、模型因子数较少时，两种方法结果基本一致。.通过中医专家咨询调查表的构建及结构方程模型的结果，分别得到五个证候的诊断模型中的Q-矩阵。然后，利用诊断分类模型中 DINA模型、DINO模型、GDINA模型及Reduced RUM模型分别探索高血压病的中医证候要素，并对个体进行统计诊断，计算每个个体的Bayesian后验概率。构建的模型对中医个体化证候要素的判断提供参考依据，对疗效评定中分类资料的潜在变量问题处理也具有参考价值。所建立的模型为我们正在建立的中医证候诊断分类辅助平台提供理论的基础。