The Associated Factors of Fatigue with Statistical Applications among Postgraduate Students

• Main Authors: Yi-Chin Lee, 李依錦
• Other Authors: 陳秀熙
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/36114931557032534359

博士 === 國立臺灣大學 === 流行病學與預防醫學研究所 === 101 === Background: In addition to the established risk factors, the contribution of co-morbid diseases or metabolic syndrome to fatigue, particularly for young post-graduate students who are one of groups vulnerable to fatigue, is still poorly understood. However, the measurement of fatigue may get involved with multi-dimension property on different aspects of clinical attributes. Beside, as the subjects of interest were postgraduate students, the data is a hierarchical structure such as students nested within class or colleges. Moreover, since the risk factors responsible for fatigue are multifarious and the reciprocal relationships are complex, we therefore applied several statistical techniques to accommodate these properties while the relationships between putative factors and fatigue were investigated. Methods: We conducted a survey for identifying risk factors associated with fatigue among 1806 new entrants of post-graduate students in the year 2004. We identified comorbidity information on these participants three years prior to the study. In the first part, we applied a Poisson regress model to evaluate the association between comorbid disease and fatigue by treating the visits due to comorbidity as counts in order to build up risk score taking co-morbidity into account. In the second part, multivariate analysis was adopted to evaluate the association between metabolic syndrome and four-dimension of fatigue score. In the third part we built up multi-level regression models inherent from hierarchical data structure to evaluate the association between associated factors and fatigue. Finally, we used the structural equation model making allowance for hierarchical data structure to disentangle the reciprocal relationships across these multi-attributes. Results: Students who had been diagnosed as one of three systematic diseases led to severe total fatigue by 14% for respiratory system disease (International Classification of Disease, 9th Revision, Clinical Modification code(ICD) 460-519 , relative risk (RR) =1.14, 95% CI: 1.00, 1.31), by 49% for genitourinary system disease (ICD 580-629, RR=1.49, 95%CI: 1.01, 2.20), and by 30% for skin and subcutaneous tissue disease (ICD 680-709, RR=1.30, 95%CI: 1.00, 1.69) when potential confounders were adjusted. By the stratification of gender, males with Hotelling-Lawley Trace test revealed significant multivariate effects relevant to four-dimensional fatigue for waist circumference (p=0.016), systolic blood pressure (p=0.022), and diastolic blood pressure (p=0.015) respectively. The corresponding results for females failed to show multivariate effects for each metabolic component. In multilevel regression model, while each specific fatigue score was taken as a dependent variable, the random intercept model with adjustment for baseline fatigue score across colleges shows a non-statistical significant difference for all the independent variables (such as physical activity). The results based on random slope model and both random intercept and slope models were not statistically significant. For analysis of structural equation model, the significant path coefficient included comorbidity score to healthy life style (path coefficient (standard error), -0.0294 (0.0144)), and three paths to fatigue, including nutrition status (0.0624 (0.0298)), healthy life style (-0.6823 (0.0715)), and sleep time (-0.0426 (0.0183)) It is noticed that the contribution of metabolic factors to any pathway was not statistically significant. Conclusion: This thesis elucidated the associations between fatigue and risk factors such as co-morbidity and metabolic factors that have been barely addressed before. Different statistical methods have been proposed to analyze the postgraduate data characterized by four-dimension property and also multi-level property. The structural equation model was further applied to building up the pathways leading to fatigue by considering various latent variables and their associated manifested variables.