Research On Quantitative Analysis Methodology Of TCM Syndrome Of FGIDs Based On Multidimensional Computer Adaptive Testing

ObjectiveThis article aims to establish new methodology for quantifying traditional Chinese medicine(TCM)syndrome s using classical statistic methods and modern analytic technologies since there is a lack of scientific methods avaliable for syndrome differentiation due to its complexity in theory and speciality in clinical practice.Treatment based on syndrome differentiation is both the main theoretical characteristic and the basic clinical practice method of traditional Chinese medicine.And syndrome differentiation play an important part in Chinese medicine.With the purpose of resolving the diversity of disease syndromes and the complexity of clinical manifestations,government and researchers has made great effort in establishing standardized specifications and practical guidelines for syndrome differentiation in recent decades.However,unified practice guidances and quantitative analysis methods for syndromes that meet the case of clinical applications are still in lack therefore become obstacles to the research about Chinese medicine modernization.It is urgent to develop research methods to meet the objective requirements about multidimensional quantitative evaluation of TCM syndromes.Setting functional dyspepsia(FD)and diarrhea-predominant irritable bowel syndrom(IBS-D)in functional gastrointestinal disorders(FGIDs)as example,this study attempts to establish syndrome models with the classical statistical methods and modern measurement theory such as multidimensional item response theory,stuctural equation modelling and computer adaptive test.Through multidisciplinary fusion technology,this article try to set up a new multidimensional quantitative analysis approach for TCM syndrome evaluation.On basis of the multidimensional quantitative model,this article then construct a new network platform as technical support for scientific research and clinical practice about syndrome differentiation.Methods1.Research resource integration In previous study,a series of instrument s and data resources around syndrome differentiation and quality of life measurement for FGIDs were developed.As the first step of this study,item bank and database are to be set up for FGIDs syndrome analysis by integrating and optimizing all the research resources.Scope of the study is limited to FD and IBS-D that fall within definition of Weitong,Weipi and Xiexie in TCM theory and clinical practice.Through literature review,group discussion and experts review,typical syndromes and related symptoms and signs of the diseases are selected from clinical practice guideline and expert consesus materials as basic construction of theoretical TCM syndrome model.Referring to advanced theory and technology of TCM syndrome differention research including Syndrome Factor Theory and Pathogenesis Factor Theory,syndrome factors are extracted to establish multidimensional hierarchical structure of theoretical model of syndrome so as to offer basic model for quantitative analysis.2.Clinical investigation A two-stage investigation combining a retrospective and a prospective study is carried out for clinical data collection.In the retrospective study,clinical data including personal details,clinical features and diagnostic informations are collected to construct the quantitative model.Data collected in the prospective study are used for evaluation of the syndrome model about its accuracy,practicality and understandability.The investigation are carried out in the First Affiliated Hospital of Guangzhou University of Chinese Medicine from October 2015 to October 2017.After singing consent form,all cases are enrolled in the study once they meet the criteria designed according to the discription of diagnostic criteria for FD and IBS-D in the Rome IV Criteria for Functional Gastrointestinal Disorders.And their TCM syndromes are diagnosed by TCM experts with senior titles.Since data are retrived and checked,demographic characteristics,features and clinical characteristics of Syndrome were analyzed in detail to offer enough information for model construction.3.Theoretical model optimization On basis of the theoretical model established and clinial distribution character analysed resutl in the front stage,the model is optimized as adapt ation for the core TCM syndromes.At the same time,according to the definition of syndromes and the pathogenesis of various syndromes,the relationship between clinical characterization and various syndrome differentiation elements is established.Combined with the clinical distribution characteristics of each clinical feature,we evaluated the clinical characterization of the syndrome factors by a panel of experts,and finally established the hierarchical model and made concerntration for it by getting rid of less common syndromes,syndrome factors or symptoms.4.Quantitative evaluation of syndrome model Based on the clinical data distribution,Logistic regression analysis,structural equation model(SEM)and multidimensional item response theory(MIRT)are used integratedly for extracting and quantifying the theoretical model.In SPSS 22,Logistic regression analysis on the correlation between the symptoms and the syndrome factors.As to analysis details,the step forward method(Forward:LR)of maximum likelihood estimation is used as a stepwise regression variable selection method.The significant level of the selected variables was 0.05,and the significant level of eliminating the variables was 0.10.Based on the statistical efficiency of each variable,positive correlations are extracted and screened.In this way,the model structures of variables are reduced and optimized.According to the results of Logistic regression analysis,the structure equation model is used for evaluating the degree of fitting between clinial data and theoretical structure of the syndrome model in AMOS 21.The theoretical model is used as the initial structure,and the clinical data were used for evaluating.The evaluation criteria of the model structure referring to indices that commonly used for the confirmatory factor analysis including RMSEA,GFI,ACIA and so on.The models are then adjusted and optimized with modification indices provided by AMOS.MIRT is used for further analyzing the multi-dimensional measurement capability of each observation variable with clinical investigation data.In R program,Quasi Monte-Carlo EM algorithm is applied as the core algorithm for estimating diagnostic properties of the items.For items with binary options,two parameters Logistic regression model is used for parameter fitting while grade response model is used for items with graded options.Thus multidimensional correlation between syndrome differentiation factors and syndromes are establish with quantitative properties.Factor loadings and multidimensional discrimination(MDISC)are used as the main indicators for item quality evaluation.Finally,item character curve and item character surface are plotted as visualization of measuring quality of the items.5.Computer adaptive testing Based on the multidimensional quantitative modeling established by MIRT,computer adaptive test(CAT)program is developed for automatic quantitative analysis of syndromes.Accessing logic and procedure of CAT is designed consisted of four part including starting item setting,adaptive criteria,latent trait estimating method and stopping criteria.Items representing typical symptoms with diseases distinguishing ability are selected as starting items.And Maximum determinant of the information matrix is chosed as adaptive criteria.Maximum a posteriori algorithms is used for computing severity scores of individual’s syndrome factors.This test stopped either all items are answered or the change in latent trait values less than 0.05.Taking the diagnostic criteria of TCM clinical experts as the gold standard and comparing the results of automatic evaluation,the reciever operating characteristic(ROC)curve is performed to evaluate the stability of FGIDs TCM syndrome differentiation model by evaluating the accuracy and feasibility of CAT and its consistency with clinical diagnosis result.Pairing with Kappa test and chi-square test automation diagnosis and clinical diagnosis of consistency and difference analysis,thus the accuracy and feasibility of the model and the scientific comprehensive evaluation.6.Development of syndrome quantitative testing platform According to the standard process and technical method of software engineering development,the test platform is constructed in rapid prototyping pattern.According to the application requirements of clinical situation,the Browser/Server framework is used for constructing the platform.Three layer architecture of Modeler-View-Controller is designed as core logical structure of the platform.The technical solution of programming is planned using Python,R and HTML5 and other programming language program functions integratedly considering advantages of comprehensive utilization of Python in agile development and efficiency of data processing and plenty of statistical support integrated in R.Design of functional modules is mainly based on the user requirements analysis book that formulated according to ideas of TCM clinicians.CAT is designed as the core programe of the platform.Follow the logic of Syndrome Evaluation computer adaptive test design,multidimensional quantitative severity of syndrome could be calculated.And visualization module also designed to provide intuitive data presentation for clinical practice.ResultsThere are 6 scales and 4 item banks designed for FGIDs syndromes measuring together with 1 association table indicating relationship between symptoms,syndrome factors and syndromes collected as original material for building instrument in this study.Through literature search,8 treatment guidelines and expert consensus for standard clinical diagnosis and treatment about IBS-D and FD covering 28 TCM syndromes.According to definition and identification standard of various syndromes introduced in the materials,symptoms and signs are extraced matching the item bank including 177 items for Weitong,155 items for Weipi and 174 items for Xiexie.Finally,the Instrument for FD and IBS-D TCM syndrome related symptoms and signs collection is formed covering 209 items as integration of the three TCM diseases.In the clinical investigation,1454 patients are enrolled including 478 cases of Weitong patients taking over 32.86%,502 cases of Weipi patients taking 34.54% and 474 cases of Xiexie patients taking over 32.60%.Among the 28 syndromes Liver depression spleen deficiency,Spleen deficiency dampness resistance,Spleen and stomach deficiency,Liver and stomach disharmony,Spleen and stomach damp heat,Liver depression Qi stagnation,Spleen deficiency and Qi stagnation and Spleen stomach deficiency cold are commonly seen by taking over 77.92% of the total cases accumulativly.By extracting typical syndromes,syndrome factors are evaluated as typical characteristic of TCM diseases.As to Weitong,there are 7 essential factors including Stomach,Spleen,Liver,Qi stagnation,Qi Deficiency,Yin Deficiency and Damp.As to Weipi,there are 7 essential elements including Stomach,Spleen,Liver,Qi stagnation,Qi deficiency,Heat and damp.As to Xiexie,there are 8 essential factors including Stomach,Spleen,Liver,Qi stagnation,Qi deficiency,Yang deficiency,Damp and Heat.Accordingly,the theoretical model covering 22 syndrome factors of the 3 TCM disease is constructed under the category of FD and IBS-D.In the framework of the theoretical model,symptoms positivly correlated to syndrome factors are picked via Logistic regression analysis.For Weitong,114 related syndromes are finally obtained among which 99 items are positive.For Weipi,107 syndromes are botained with 90 positive.For Xiexie,113 syndromes are botained with 97 positive.After removing duplicate entries,simplified model containing 73 items for Weitong,61 tiems for Weipi and 72 items for Xiexie.The regression model of most of the syndromes of each disease is consistent with the theory of traditional Chinese medicine and the law of clinical syndrome differentiation,and the explanatory character of the model is adequate.However,there are more variable with cross properties such as Stomach,Liver and Spleen facotrs,indicating demand for modification of the model.Combined with syndrome differentiation theory model and Logistic regression analysis result,MIRT analysis is performed for quantitative analysis of syndrome differentiation model.For Weitong model,54 items are considered to be of adequate quality with MDISC over 0.5 and 16 items with goood quality with MDISC over 1.5.For Weipi model,39 items are considered to be of adequate quality with MDISC over 0.5 and 15 items with goood quality with MDISC over 1.5.For Xiexie model,53 items are considered to be of adequate quality with MDISC over 0.5 and 17 items with goood quality with MDISC over 1.5.Since three quantitative models are estimated,CAT program is developed based on the MIRT analysis results.All clinical cases data are re-evaluated automatically in teh program and the quantitative scores of TCM syndromes are evaluated.Desired result of the ROC curve analysis are achieved with most of the AUC indices over 0.8 indicating the model of automatic identification and clinical recognition with good consistency to clinical disgnosis.It was found that that adaptive testing diagnosis of Wei,Qi-deficiency,Liver-Qi stagnation and Wet elements were not statistically significant in the diagnosis of Weitong,and that the Kappa coefficient of the Wet element was highest than 0.8,indicate that the self-adaptive diagnostic result of this element was highly consistent with the diagnosis of the clinical physician.The Kappa coefficient of Hot element was the highest,while the Kappa coefficient of Hot and Wet in Xiexie were above 0.75,suggesting that the adaptive testing diagnosis of each factor was highly consistent with the diagnosis of clinicians.However,there are two reason for the lack of consistency in the diagnosis of some factors.On the one hand,since the sample size is insufficient,the accuracy of the model still needs to be optimized with a larger sample.On the other hand,the threshold form and its determination method need to be further optimized to meet the clinical diagnosis requirement.By referring to the clinical syndrome differentiation of TCM,the computer programming transformation and realization of TCM syndrome differentiation theory model are realized of which the MCAT program desinged as the core logic of the test.The platform give adequate support for disease identification by offering interactive interface to patients.According to the clinical symptoms and severity of symptoms or signs,the multidimensional severity level of syndrome are evaluated not only as results of the test but also used for the selection of the next items to promote the evaluation process.Since the evaluation is completed,the result is presented in the form of radar graph through the visual interface,which provides a comprehensive and intuitive quantitative display for the evaluation results.In addition,the system presents the patient’s multiple evaluation results in the form of a graph,providing auxiliary media for periodical evaluation and evaluation of curative effect.ConclusionThis study innovatively put forward a clinical demonstration project of TCM syndrome quantifying research.From the perspective of clinical application environment and business needs,combined with the new trend of the development of the Internet and big data technology,the integrated use of mathematical statistics and computer technology,visualization technology and modern advanced measurement technology,for the integration of resources and the number of cross time,cross regional data sharing scheme provides the engineering practice.It effectively fills gaps in the research and development platform of domestic scale research,and provides new technical plan and working mode for promoting the clinical demonstration of TCM Syndrome Scale.