辨证选用中成药的辅助决策模型研究

Proprietary Chinese medicine，which is effective, safe and easy to use, is favored by patients. Itis often used not only by TCM doctors but also by Western medicine physicians.who are usually lack of TCM theory. The blindness in the use of proprietary Chinesemedicines resulted in more and more adverse events in recent years which hasbecame a major issue affecting the efficacy evaluation of proprietary Chinesemedicines or even endangering the safety of drugs. Improving the accuracy ofWestern doctors' use of proprietary Chinese medicines is a problem need to besolved and is also a key link to promote the proper use of traditional Chinesemedicine.Webelieve that achieving the maximum treatment effect is the prerequisite of thedevelopment and application of proprietary Chinese medicines. For TCM clinicalexperts, the process of proprietary Chinese medicine selection, based on fourdiagnostician information and biological indicators, is a Statistics Assessmentprocess. Based on this, we are going to establish a clinical informationcollection standard which is easy to understand and operate by Medical Doctorswith item response theory and selective integration classifier technology.Amathematical model, which includes TCM doctors' experience and could embody therelation between four diagnostic information and proprietary Chinese medicines, will be developed to help Western doctorswith right decisions in the use of proprietary Chinese medicines and alsoprovide a scientific basis and technical support for the safe, effective andstandardized use of proprietary Chinese medicines.

中成药以"疗效确切、服用便捷、使用安全"等优势深受广大患者青睐，在临床实践中被西医医师广泛应用。但西医医师往往不具备中医辨证论治理论知识，使用中成药时盲目性较大，导致近年来因不合理用药产生的安全事件逐年增多，这已成为影响中成药疗效评价、甚至危及用药安全的重大问题。提高西医医师选用中成药的准确性是目前临床亟待解决的问题，也是促进中成药合理化使用关键环节。.中成药的研制与应用是以针对特定人群能够取得最大治疗效应为前提的，中医临床专家根据患者临床四诊信息和生物学指标，方证对应地选择中成药的思维过程，是一个统计判别过程。基于此，本课题拟采用项目反应理论、选择性集成分类器等技术建立易于为西医医师理解和操作的临床信息采集标准；构造富集中医辨证选药经验，能够再现中医思维下临床辨证信息与中成药之间的对应关系的数学模型，用于辅助西医医师正确决策用药，为安全、有效、规范地使用中成药提供技术支撑。

辨证施治是中医临床取得最佳治疗效果的关键。中成药是依中医理法方药而制定的汤方，其应用必然遵循辨证论治。临床上，西医师应用中成药十分广泛，但临证派药较为盲目，导致中成药的临床疗效严重扭曲，甚至危及用药安全。相对个体化汤剂，中成药体现特定人群的类证候。中医选用中成药的思维过程，是一个内隐的统计判别过程。本研究以中风病为例，构建准确性、鲁棒性、泛化能力高的富集中医专家辨证选药经验的辅助决策模型，为规范使用中成药提供技术支持。.1.筛选中成药适宜人群。.1.1采用Delphi法选出中风病常用中成药并进行药物分类。得到21种中成药，分中药提取物类、活血化瘀类、益气活血类、化痰类和补益类五大类。药物证候分类，对临床选药有一定指导意义。.1.2采用最高分、阈值、重复率、顺序法判断适宜人群。最高分法效果好(83.82%)；常用益气活血类、熄风化痰类，具体药物是天丹通络胶囊(22.76%)和牛黄清心丸(21.38%)。以“support=30&lift=1.0”为标准，获取药物关联规则20条：益气活血类、活血化瘀划分合理，而熄风化痰药物存在与益气活血、活血化瘀共用现象。标注中成药的适宜人群，挖掘药物关联关系,为用药决策提供参考。.2.建立再现中医专家临床辨证选药思维过程的辅助决策模型。.2.1基于问诊信息构建中成药分类模型，总结症-药规律。采用logistic回归建模，交叉验证模型准确性。天丹通络胶囊适应症为口唇麻木、头晕等症状。考虑研究者效应，分析专家用药规律。培元通脑胶囊核心症为视物模糊、口渴等症状。单标签药物判别模型，为临床辨证提供了客观依据。.2.2优化中风病四诊信息表。建立Rasch模型采用CMLE估计，考评指标区分度及信息量。总体信息量较高(标准误>0.5)，精度高；但失眠、多梦等24个症状需要调整(区分度<0.2)。问诊指标为药物划分提供较多信息，改进指标内容可提高分类预测精度。.2.3构建辨证使用中风病中成药辅助决策模型。采用CfsSubsetEval法特征选择，获得48个主要指标；建立基于流形排序的判别模型，药物、患者平均准确率为86%、50%，影响因素分别是舌象、四诊信息。药物平均准确率与查全率估算需150例/种。半监督方法为证候量化研究提供了方法学支撑。.在非线性、高维小样本情况下，构建了中成药辅助决策模型，为帮助临床西医医师合理选用中成药提供了新的思路和方法。