Statistical Methods For Agreement Evaluation Of The Four Diagnostic Information Collection In Traditional Chinese Medicine

Objective:Syndrome differentiation and treatment is the essence of traditional Chinese Medicine (TCM). It is very important to establish an objective quantitative standard for TCM syndrome differentiation. Collection of Four diagnostic information is a foundation in quantitative research of TCM. Taking chronic bronchitis as an example, different agreement statistical methods were used to assess agreement of the TCM four diagnostic information among the different researchers and different research units in order to improve the clinical TCM precision and repeatability. Ultimately making evaluation of efficacy in clinical trials relatively scientific and objective.Methods:In order to ensure the accuracy of the data, the TCM four diagnostic information of chronic bronchitis were inputed and checked by double people with different compuuter by Epidata 3.1. Commonly agreement methods were used to evaluate four diagnostic information of chronic bronchitis firstly, the agreement of variables were evaluated according to Fleiss levels of kappa coefficient. If it is the fair agreement, the models of agreement were constructed. The optimal model were selected from different log linear models accord BIC and AIC indexes, and agreement were interpretated by its parameters. Finally, latent class model was used to evaluate each caterory’s agreement by conditional probability. SAS9.3 software and Mplus software were used to analyze. SAS9.3 can be used to analyze the description of agreement and the log-linear model of agreement, and using Mplus analyzed the latent class model.Results:1. Agreement description:14 items were fair agreement,36.8% of the total syndrome elements,24 items were the good agreement,63.2% of the syndrome elements. The good agreement items of syndrome elements were cough, sputum, asthma, God fatigue, less gas lazy talk, spontaneous sweating, loss of appetite, dispirited spirit, tinnitus, chest tightness, shortness of breath, chest pain, greasy moss, and taut pulse. The fair agreement of syndrome elements belonged to a more subjective indicators, with respect to the degree of agreement of the syndrome elements, they were no objective indicators or signs.2. Loglinear model:Constructing the log-linear models of agreement for the fair syndrome elements (except for the pulse string), the weight and correlation were considered for the factors of agreement. Lassitude of the rank correlation and chest pain of various degree of importance had a significant for agreement, the others syndrome elements had significant in rank correlation and the degree of importance. When the covariate (hospitals) was considered, the syndrome elements in the hospital were no significance, in addition to chest tightness and fatigue of the gods.3. Latent class analysis:The evaluation criteria for cough was different between the attending physician and the director (deputy). Agreement between two doctors are good in none of degree. For the moderate and mild of degree, there had some differences in the evaluation of some items, such as listlessness, sweating, etc. For the frequency of severe degree was sparse, its’ agreement coefficient is notsuitable.Conclusion:Four diagnostic information of TCM is the more subjective, and is unable touched or observed the specific features of syndrome elements, is only relied on physicians skills or based on the patient description of the disease.The correlation and the degree of importance between grades have influence on agreement in syndrome elements.The each grade of conditional probability can provide classification basis for the syndrome elements.Agreement analysis play an important role in improving the accuracy of the research on the quality of clinical.