Structural Equation Model for Periodontal Disease

• Main Authors: Wu-Han Chang, 張婺涵
• Other Authors: Hsiu-Hsi Chen
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/17352763382941608330

碩士 === 國立臺灣大學 === 流行病學與預防醫學研究所 === 99 === Background： Logistic regression model with binary outcome of periodontal disease (PD) has been traditionally used to investigate the associated risk factors. From statistical viewpoint, this approach dose not deal with the highly correlations between risk factors. From the preventive medicine viewpoint, the traditional approach fails to elucidate the causal relationships and pathways interplayed by these risk factors in association with PD. Aim： The aim of our study is to apply a series of structural equation models to build up a pathway diagram leading to PD through a series of causal relationships. Materials and Methods： Our study was based on national survey of the prevalence and risk factors in association with PD for adults aged 18 years on older in Taiwan. The sample size was 4061. Data on periodontal indexes, periodontal knowledge, lifestyle, eating habits and biomarkers were collected. Factor analysis was firstly used to find the construct aggregated by similar variables in each dimension. The measurement model is further built by using confirmatory factor analysis (CFA) to assess the covariance between constructs in addition to the relationship between variables and each construct. Finally, we built up the structural equation model (SEM) for PD by extending CFA. Result： Factor analysis was performed to form four constructs, periodontal knowledge, life style, eating habits and biomarkers. In measurement model, periodontal knowledge has negative relationship with periodontal indexes (-0.6528,p-value<0.0001), lifestyle has posistive relationship with biomarkers (0.6582,p-value<0.0001) and biomarkers has positive relationship with periodontal indexes (0.3705,p-value<0.0001). The exogenous variables in SEM are life style and periodontal knowledge, we found that the periodontal knowledge has direct effect on periodontal indexes (-0.6350,p-value=0.0230). Lifestyle, correlated to periodontal knowledge, may affect biomarkers (1.1255,p-value<0.0001) and, in turn, affect periodontal indexes (0.0663,p-value=0.0038). Conclusion： By the application of structural equation model, we constructed a pathway whereby the direct effect of periodontal knowledge and indirect effect of lifestyle through the influence of biomarkers on PD was demonstrated.