| [Background]With the current social economic development, social rhythm increasing the competitiveness of the crowd will also increase accordingly, the job stress, emotional tension. Such adverse living conditions have become the norm status. At the same time, many kinds of chronic illnesses are appeared by this unhealthy life status, especially of the gastrointestinal diseases. Functional gastrointestinal disease is characterized by long cycle and symptoms, so it is hard to arouse the attention of patients to medical treatment without serious organic diseases. The diagnosis of functional gastrointestinal diseases is often according to the experience of the doctor. That is easily to lead to missed diagnosis. Nowadays, some questionnaire based on the symptoms has been considered as the effective diagnostic tool for the functional gastrointestinal diseases. But in China, some commonly used questionnaires such as RDQ, ROMEēcombination questionnaire are lack of application. There are not some Chinese version of the questionnaires had received formal large sample of cognitive testing and reliability and validity detect.About the research methods, the traditional statistical methods generally use the Cronbach coefficient and exploratory factor analysis to test the questionnaire reliability and validity, but the application of these methods always exist some shortcomings. During the analysis of questionnaire results, the traditional methods can not solve the problems of setting models between different questionnaires, and unable to overcome the interaction and mixed effects. Structural Equation Model introduces the latent variables to the path analysis, and combines the latent variable and observation variables effectively by factor analysis methods. It can analyses not only relationships between latent variables and can also analyses the relationship between the latent variable and observation variables. At the same time, structural equation model provides the method for the expression of abstract variables and measurement the accuracy of the test. Using the measurement model part in structural equation modeling, which is the confirmatory factor analysis method, can analyze the scale's construct validity and reliability.[Objective]This topic has the following specific purposes: ?Used the delete method, EM algorithm, FIML algorithm to handle missing data, and explore a suitable method for this study data, and comparing the merits between these three methods; ?Used the confirmatory factor analysis and the exploratory factor analysis to evaluate the reliability and validity of the survey instrument, and comparing the two types of factor analysis methods; ?Used the structural equation model to set up the quality of life structural equation model under the role of gastroesophageal reflux disease, according to the internal model parameters to learn the risk factors of GERD, the impact of GERD to the quality of life and the risk factors'direct and indirect impact to the quality of life.[Methods]1. The methods for missing data treatmentUsed of the delete method, the expectation maximization method and the full information maximizing-likelihood method to deal with missing data. These three kinds of methods to deal with missing data represented the three main areas, which were the "delete", "fill" and "no deal". Through the fit index of SF-36 scale measurement model, to determine which method were fit of this study.2. The methods for the reliability and validityUsed of the confirmatory factor analysis and the exploratory factor analysis to evaluate the validities of the survey instrument. And estimated the internal consistency reliabilities according to the Cronbach'? coefficients and the R~2 which were calculated by the confirmatory factor analysis,3. The methods for establishing the modelUsed of the structural equation model to set up the quality of life structural equation model under the role of gastroesophageal reflux disease. The model could make the relationships between the measurement variables and latent variables, as well as the latent variables more clearly.[Results]1. Data qualityThere are few missing in the survey data. The quality control was reasonable and the sample had a strong representativeness.2. Deal with the missing dataUsed of the delete method, the expectation maximization method and the full information maximizing-likelihood method to deal with missing data. The confirmatory structural equation models were better fitting based on these three kinds of missing data treatment methods. Only from the Root Mean Square Error of Approximation (RMSEA) and X~2 , the FIML algorithm seemed more suitable for this study. But through the factor loading of the models, it can be learn that the maximum factors load values appeared in the EM algorithm model. All the factors were greater than 0.5, and it was more fitting to the theoretical model.The result with FIML method and delete method were basically the same. Their factor loading coefficients on the individual dimensions were less than 0.5. We could see that the missing values indeed had impacts on the models, and different treatment methods of missing value fit out the very different models. Considered the results, this study was carried out EM algorithm to deal with missing data.3. Scale evaluationThe validity, from the exploratory factor analysis, RDQ had a good construct validity. In SF-36 scale, some individual variables in the vitality (VT) dimension and mental health (MH) dimension divided into the wrong attribution of common factor. The other results were in line with the SF-36 scale theoretical framework, to illustrate the scale good construct validity. From the confirmatory factor analysis, the RDQ and the SF-36 scale models indicated that the theoretical models were well fitting to the actual data. All parameters estimated had reached a significant level of 0.01, and the estimated standard error of the parameters was very small. So the construct validities of RDQ and SF-36 scales were good.Therefore, both confirmatory factor analysis and exploratory factor analysis results showed that the RDQ and the SF-36 scales had good construct validities.The reliability, whether the application of confirmatory factor analysis or Cronbachαcoefficient, only the SF-36 scale of social functioning (SF) dimension was less than the 0.6 standard, other dimensions and RDQ were proved to had good reliabilities.4. General modelFrom the model, GERD risk factors were divided into three aspects, "Sociology of population characteristics," "Living status characteristic" and "family history". "Living status characteristics" was the greatest impact factor with GERD, followed by "family history", "Sociology of population characteristics" was the minimal impact factor with GERD.GERD was most affected the patients with body pain (BP) dimensions, followed by general health (GH). Although the affect with psychology existed, it was less than the physiology. GERD affected the patient's mental health (MH), followed by the vitality (VT).The 3 major risk factors ("Sociology of population characteristics", "Living status characteristic" and "family history") impact on the quality of life, divided into the overall effects and indirect effects. From the overall effects, "Family history" affected the general health (GH) most, "Living status characteristic" affected the body pain (BP) and vitality (VT) most, "Sociology of population characteristics" affected the mental health (MH) most. From the indirect effects, GERD caused by "Living status characteristics" affected the quality of life most. GERD caused by either the risk factors ("Sociology of population characteristics", "Living status characteristic" and "family history") affected the body pain (BP) dimension most.[Conclusion]The conclusions of this study divided into two areas, the conclusions by empirical research and the conclusions by methods research.1. The conclusions by empirical researchThe survey design was strict quality controlled. The data was with high quality and good representativeness. The questionnaire has good reliability and validity.Different from traditional research, through structural equation model in this study, it not only explored the risk factors such as "age", "BMI index", "smoking", "frequency of exercise", "mentality", " social status "," ability to live and work "and" family history ", but also can find the different risk factors make the different role with GERD according to the path coefficients in model. The "mentality", "social status", " ability to live and work" which on behalf of "Living status characteristics" affected GERD rather than by the "age", "BMI index", "smoking tobacco", "frequency of exercise" which represented by " population characteristics of sociology". It can be said that if patients who were with difficult to reduce their own weights, less smoking, more exercises, then, by adjusting the mentality, more to communicate with the outside world, increasing the ability to live and work can also play a role in the prevention of GERD happen.As for the impact of the quality of life in patients with GERD, the most related studies showed GERD can affect significantly 8 dimensions on SF-36 scale in very general exposition. Through this study, it can be learned that GERD affected the physiology more than the psychology, especially in the physical pain (BP) dimensions. Through the impact on patient physiology then the psychology could be affected. Gave the patients enormous the spirit burden and affected their daily mental status (MH) and vitality (VT).So, it was not necessarily entirely due to the disease itself caused by the psychological burden. Also there were some part role with the physical pain which caused by GERD to affects the psychological. By the traditional methods of statistical analysis, the two types of effects didn't be distinguished, and the indirect effects were easy to overlook.2. Methods study conclusions(1) The method for missing data: The 3 missing data treatment methods which used in this study were all with advantages and disadvantages. Particularly for the use of the full information maximizing-likelihood method by structural equation model to deal with missing data, used the missing data itself as the data information. Introduced the new idea that directly analysis the model with the missing data. In addition, the study also found that the good results could be achieved by the delete method, the condition of completely random missing.(2) The method for reliability and validity: The confirmatory factor analysis and exploratory factor analysis were compared in this study. In the construct validity testing for SF-36 scale, not all variables were attributed to their corresponding common factors by using the traditional method. It was with vague and subjective by using this kind of result to evaluate of validity. But the confirmatory factor analysis was base on the professional structure theory. It could test the accordance between the survey data and the theoretical models. It was with clear and objective to assess the construct validity by a series of fit indexes.At the same time, the confirmatory factor analysis and Cronbach ? coefficient were compared in this study. The confirmatory factor analysis which based on structural equation model had the obvious advantages, compared with the Cronbach ? coefficient, for calculating each variable's reliability, not requiring the same direction in every measurement, overcoming the disadvantage in Cronbach ? coefficient lower than the real internal reliability.(3)The method for exploring the risk factors of GERD and GERD on the quality of life impact: It could only estimate the risk factors on the direct effect of GERD by traditional statistical methods, or the directly affecting on quality of life by GERD. It was difficult to combine the risk factors, GERD and quality of life and do the comprehensive evaluation. In other words, it is difficult to put the general situation questionnaire, RDQ and SF-36 scale in one model with general considerations.The structural equation model which was set up in this study was using the interrelationship between 8 latent variables extracted from the three questionnaires. The path coefficients can show different risk factors on GERD by different intensity, and GERD on quality of life are also different. The risk factors, the direct relationship between GERD and quality of life, the indirect relationship between risk factors and quality of life by GERD could be shown by the path graph.The structural equation model compared with other models also had the advantages of allowing comparisons between different models, rich fitting indicators, and intuitive expression by path graph.3. The enlightenments by establishing the modelThrough this study, structural equation modeling has a strong confirmatory function, so using the structural equation model to develop the new scales also can play a good role. In model modification, the nature of structural equation modeling is to explore gradually by the certification, but it still should not deviate from the basic theory. Structural equation modeling can not only carry on by the professional software such as LISREL, Amos, EQS, but it is also embedded in some large software packages like SAS, SPSS. Because there were so many variables in this study, the LISREL software was used to establish the models. First, set up the theoretical model by path graph, then updated the model by the program which generated by the path graph.4. Outstanding issuesFor it was difficult to iterative interview the same respondents with the same questionnaire, at a certain period of time. The latent variable growth curve model was not used for the test-retest reliability in this study. The superiority of Latent variable growth curve model in dealing with the longitudinal data had not been verified in this study. In this study, the psychological problems was treated as resultant variables, it was not explore the psychological factors of GERD reaction. It was recommended that psychological scale should be added in the follow-up study, and carry on the cognitive psychology - behavioral intervention, then do the follow-up study.5. The limitations of structural equationThe data used for structural equation modeling need a larger sample size. There must be a relevant professional and reasonable theoretical basis when using of structural equation model. In the use of structural equation modeling, the investigator must have the in-depth theoretical basis and the application conditions of it. Avoid of the abuse and misinterpretation of the structural equation modeling.
|
PDF Full Text Request