混合型高脂血症证候要素特征的隐结构模型评价

Latent Structure Model Evaluation on Pattern Factor Characteristics in Mixed Hyperlipidemia

目的以多维分层聚类的隐结构模型方法评价混合型高脂血症的中医证候要素特征。方法结合临床专家实践经验筛选血脂异常症状条目,采用隐结构模型方法构建634例混合型高脂血症的隐树结构图,分析隐类概率、条件概率、互信息及累积信息覆盖率等症状/证候量化数据,以人工判读方法诠释混合型高脂血症的主要证候要素特征。结果共以41个症状作为显变量构建634例混合型高脂血症隐结构模型图。634例混合型高脂血症包含2个隐类,其中Y0=S0的概率值P=0.41,Y0=S1的概率值P=0.59,综合判定634例混合型高脂血症人群表现为中、重度症状/证候。与Y0相关的互信息由大到小前3位症状依次为心烦(0.19)、急躁(0.12)、腰酸(0.12)。混合型高脂血症证候要素以气虚、气滞、火热、阴虚为主,其中以阴虚最为重要,证候要素靶位以心、肝、肾为主。结论混合型高脂血症证候要素以气虚、气滞、火热、阴虚为主;说明隐结构模型方法评价混合型高脂血症证候要素特征,其结果与中医学对血脂异常的传统认识基本一致。

Objective To evaluate the pattern factor characteristics of mixed hyperlipidemia with latent structure model. Methods The symptom items of dyslipidemia were screened by experienced clinical experts based on guideline and literature. The implicit tree structure in 634 cases of mixed hyperlipidemia was established with latent structure model. The symptom or pattern quantitative data such as hidden category probability, conditional probability, mutual information and accumulated information coverage was analyzed. The pattern factor characteristics of mixed hyperlipidemia were analyzed with manual interpretation method. Results The implicit tree structure in 634 cases of mixed hyperlipidemia was established with 41 symptoms as significant variables. There were 2 hidden categories in 634 cases of mixed hyperlipidemia. The probability value of Y0 =So and Y0 =SI was P=0.41 and P=0.59 respectively. The moderate to severe symptom or pattern was shown in 634 cases of mixed hyperlipidemia. The top 3 symptoms that mutual information associated with Y0 from big to small were upset （0.19）, irritable （0. 12） and backache （0.12）. The common pattern factors of mixed hyperlipidemia were qi deficiency, qi stagnation, fire heat and yin deficiency, especially yin deficiency. The common pattern factor targets were heart, liver and kidney. Conclusion The mixed hyperlipidemia manifests as moderate to severe symptom or pattern. The common pattern factors of mixed hyperlipidemia are qi deficiency, qi stagnation, fire heat and yin deficiency. The result of pattern factor characteristics of mixed hyperlipidemia with latent structure model is consistent with traditional understanding of Chinese medicine.