基于隐结构结合Logistic回归分析探讨9323例古籍咳嗽医案证候分布

Syndrome Distribution of 9323 Cough Cases in Ancient Chinese Medical Books Based on Latent Structure Model and Logistic Regression Analysis

目的:探索中医古籍咳嗽医案证候分布规律。方法:提取古籍中医肺病数据库内咳嗽病案9323例,用Lantern 5.0(孔明灯)软件构建前50位症状和中药共100个显变量隐结构模型,对不同隐节点进行诠释,根据隐结构综合聚类权重量化揭示症状和证候、中药(方剂)和证候间的辨证规则,通过二分类Logistic回归分析拟合不同病性和证候间的相关关系。结果:9323例咳嗽医案涉及证候204种,频次>100的证候有18种。将前50位症状和50味中药建模得到35个隐变量,98个隐类,10个综合聚类模型,其中Z5的阈值6.7最高,Z6咯痰信息覆盖度52%最高,Z7的分值19最高。二分类Logistic回归模型拟合出不同病性和5类证型的相关关系,其中外风和风热犯肺证的优势比最高达88.919,痰热蕴肺证与病性热、痰的优势比为51.594和15.861,痰湿阻肺证与湿、痰、饮的优势比分别为31.415,34.370和4.936。从证候的频数分布、以症推证、以药(方)测证得出咳嗽的常见证候有14种,分别为外感咳嗽有风寒袭肺、风热犯肺、风燥犯肺,内伤咳嗽有痰湿阻肺、痰热壅肺、肝火犯肺、肺阴亏虚、肺气阴两虚、脾肺气虚、肺肾阴虚、肺热阴虚、外寒内饮、脾胃虚弱、痰瘀阻肺。结论:新发现的咳嗽证候有肺热阴虚、外寒内饮、脾胃虚弱、痰瘀阻肺证,临证多以复合证候为主,如肺气阴两虚、肺肾阴虚。中医证候的辨证带有一定的主观性,权值的高低提示不同症状对证候的贡献度有差异,对证候推断有一定指导意义,隐结构模型结合Logistic回归分析在一定程度上能解决中医辨证的定量问题,可用于疾病证候分布的挖掘。

Objective:To explore the syndrome distribution of cough cases in ancient Chinese medical books.Method:A total of 9323 cough cases in the database of lung diseases in ancient Chinese medical books were extracted.Lantern 5.0 was used to construct the latent structure model for the 100 manifest variables based on the first 50 symptoms and 50 Chinese herbal medicines,and different latent nodes were interpreted.The syndrome differentiation patterns of syndromes with symptoms and Chinese herbal medicine(formula)were quantitatively revealed by the comprehensive clustering weights of latent structure.The correlation of diseases with syndromes was fitted through the binary Logistic regression analysis.Result:There were 204 syndromes involved in 9323 cough cases with 18 syndromes showing a frequency higher than 100.As demonstrated by the model established on the first 50 symptoms and 50 Chinese herbal medicines,35 latent variables,98 latent classes,and 10 comprehensive clustering models were obtained,where Z5 was the highest in the threshold value(6.7),Z6 in the information coverage of productive cough(52%),and Z7 in the score(19).The binary Logistic regression model fitted the correlation between different disease types and five syndromes,where the dominance ratio of external wind to the syndrome of wind-heat invading lung reaching up to 88.919,those of syndrome of phlegm-heat accumulating in lung to diseased heat and sputum 51.594 and 15.861,and those of the syndrome of phlegm-dampness obstructing lung to dampness,phlegm,and fluid retention 31.415,34.370,and4.936,respectively.Conclusion:The newly discovered cough syndromes included lung heat and yin deficiency,external cold and internal fluid retention,weakness of spleen and stomach,and phlegm and blood stasis in lung.In most cases,multiple syndromes were observed clinically,such as syndrome of deficiency of both Qi and Yin in lung combined with yin deficiency in lung and kidney.Since differentiation of traditional Chinese medicine(TCM)syndrome is subjective,the weight can indicate the difference in the contributions of different symptoms to the syndrome,which is of guiding significance for syndrome inference.The latent structure model combined with Logistic regression analysis can solve the problem of quantification in TCM syndrome differentiation and can be used to explore the syndrome distribution of diseases.