中医四诊信息采集者间一致性评价的统计方法研究

发布时间：2018-10-20 11:09

【摘要】：目的：辨证论治是中医学的精髓,而中医辨证建立客观的量化标准是十分重要的。中医四诊信息的采集是量化的基础,显得尤为重要。以慢性支气管炎为例,本研究采用不同的统计分法考核不同研究者和不同研究单位对中医四诊信息采集的一致性研究,以提高中医临床辨证的精确性和可重复性,使临床试验中疗效评价相对科学、客观。方法：收集两个不同医师分别进行采集的321例慢性支气管炎数据,用Epidata进行双人双机录入并校对,以保证数据的准确性。对慢性支气管炎的四诊信息采用常用的一致性评价方法进行评价,根据Fleiss对kappa系数的分级筛选出一致性程度一般的四诊信息,再对筛选出的四诊信息进行一致性模型的构建,首先应用对数线性模型,根据BIC指标选择出拟合程度最好的模型,对其参数进行估计与解释；最后应用潜在类别模型,得到四诊信息的各等级在潜在类别下的条件概率。研究用SAS9.3软件和Mplus软件相结合,SAS9.3进行一致性描述、一致性的对数线性模型,而Mplus进行潜在类别分析。结果：主要研究结果如下：1、一致性描述：一致性程度一般的有14个,占四诊信息总个数的36.8%,一致性程度较好的有24个,占四诊信息总个数的63.2%。一致性程度一般的四诊信息分别为咳嗽、咳痰、气喘、神疲乏力、少气懒言、自汗、纳呆、精神萎靡、耳鸣、气短、胸闷、胸痛、苔腻、弦脉。一致性程度一般的四诊信息相对于一致性程度较好的四诊信息,属于更为主观的判断指标,没有客观的指标或体征说明。对于kappa系数在0.75以下的变量分别采用对数线性模型及潜在类别模型进一步考虑一致性的评价。2、对数线性模型：考虑等权重与相关性对一致性的影响时,等权重和线性间交互对于所有的四诊信息的数据解释均有意义。考虑等权重、相关性、协变量对一致性造成的影响,各四诊信息等级之间重要程度对的一致性数据解释均具有统计学意义,说明等权重有利于解释一致性数据。除胸闷、神疲乏力外,其余四诊信息在不同医院之间的一致性的差异没有统计学意义,即协变量医院对一致性是没有影响的。3、潜在类别分析：咳嗽在主治医师与主任(副)医师之间的评价标准是存在部分的不一致的。在无的级别上,基本上两类医师的一致性都较好,对于中度或轻度,医师在某些项目(如精神萎靡、自汗等)中的评价存在一定的差异,而对于重度的级别其出现的频率较少,其一致性并不理想。结论：中医四诊信息属于更为主观、不能触摸到或观察到具体特征的指标,只能是医师依靠自身的技能主观判断和医师依据病人对病情的描述。四诊信息的等级间的相关和各等级的重要程度对于一致性都是有影响的。四诊信息各等级的条件概率还可为四诊信息的等级划分提供依据。一致性的分析为提高研究的精确性和临床的研究质量起到重要作用。
[Abstract]:Objective: syndrome differentiation is the essence of traditional Chinese medicine, and it is very important to establish objective quantification standard. The information collection of four diagnoses of TCM is the basis of quantification, and it is especially important. Taking chronic bronchitis as an example, different statistical methods were used to examine the consistency of information collection of four diagnoses of TCM by different researchers and different research units, in order to improve the accuracy and repeatability of TCM clinical syndrome differentiation. The evaluation of curative effect in clinical trial is relatively scientific and objective. Methods: the data of 321 cases of chronic bronchitis collected by two different doctors were collected, and the data were recorded and proofread by double machines with Epidata to ensure the accuracy of the data. The four diagnosis information of chronic bronchitis was evaluated by the commonly used consistency evaluation method. According to the classification of kappa coefficient by Fleiss, the four diagnostics information of general consistency degree was screened out, and the consistency model of the four diagnosis information was constructed. First, the logarithmic linear model is used to select the best fit model according to the BIC index, and its parameters are estimated and explained. Finally, the conditional probability of each level of four diagnostics information under the potential category is obtained by using the potential category model. Using SAS9.3 software and Mplus software, SAS9.3 is used to describe consistency, logarithmic linear model of consistency is used, and Mplus is used for latent category analysis. Results: the main results were as follows: 1. Consistency description: there were 14 cases of general consistency, accounting for 36.8% of the total number of information of four diagnoses, and 24 cases of good consistency, accounting for 63.2% of the total number of information of four diagnoses. The general information of the four diagnoses were cough, expectoration, asthma, fatigue, indolence, sweating, depression, tinnitus, shortness of breath, chest tightness, chest pain, greasy moss and string pulse. The information of four diagnostics with general consistency degree is more subjective than that with good consistency, and there is no objective index or sign explanation. For the variables with kappa coefficient below 0. 75, the logarithmic linear model and the potential category model are used to further consider the evaluation of consistency. 2, logarithmic linear model: considering the influence of equal weight and correlation on consistency, The equal-weight and linear interactions are significant for the interpretation of all four diagnostics. Considering the influence of equal weight, correlation and covariable on consistency, and the importance of each level of information, there is statistical significance in the interpretation of consistency data, which is beneficial to the interpretation of consistency data. In addition to chest tightness and fatigue, there was no significant difference in the consistency of the other four diagnostics among different hospitals. That is, covariant hospitals have no effect on consistency. 3. Potential Category Analysis: there is partial inconsistency in the evaluation criteria of cough between the attending physician and the chief (deputy) physician. Basically, there was a good consistency between the two categories of physicians at the no level. For moderate or mild, there was some difference in the evaluation of certain items (such as mental retardation, self-sweating, etc.), but for the severe grade, the frequency of occurrence was lower. Its consistency is not ideal. Conclusion: the information of the four diagnoses of TCM is more subjective and can not touch or observe the specific characteristics. It can only depend on the subjective judgment of the doctor's own skills and the description of the patient's condition according to the doctor's own skills. The correlation between the four levels of diagnostic information and the importance of each level have an impact on consistency. The conditional probability of each grade of four diagnosis information can also provide the basis for the classification of four diagnosis information. Consistency analysis plays an important role in improving the accuracy of research and the quality of clinical research.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R241

【参考文献】