Application of structural equation modeling in two independent problems in health sciences
Date Issued
April 2025
Author(s)
Advisor
Abstract
Structural Equation Modeling (SEM) is a powerful multivariate statistical technique
that enables researchers to analyze complex relationships among observed and
latent variables. While its origins lie in the social sciences, SEM has gained growing
attention in medical and healthcare research due to its ability to model multifactorial
phenomena and uncover latent structures underlying observable clinical outcomes.
This thesis explores the integration of SEM into various facets of health sciences,
particularly in the evaluation and prediction of cardiovascular disease risk using
non-invasive imaging and clinical indicators. By utilizing latent constructs such as
life-design behavior, biochemical markers, and vascular features—derived from
ultrasound-based texture analysis of the carotid artery—SEM models are developed
to examine the direct and indirect effects of these variables on cardiovascular
health.
Through a series of case studies, including preliminary work on the intima-media
complex (IMC) segmentation and more advanced modeling in published research
(for example, Section 3.2 and Section 3.3), the thesis demonstrates the effectiveness
of SEM in capturing both measurement and structural relationships in healthcare
data. The methodology emphasizes the importance of confirmatory factor analysis,
path analysis, and the testing of model fit indices to ensure robustness and
interpretability. Applications discussed include modeling the influence of
behavioral, demographic, and physiological factors on disease risk, as well as
evaluating the systemic structure of healthcare delivery systems, such as the
General Health System of Cyprus.
Ultimately, this thesis aims to showcase SEM not just as a methodological choice,
but as a strategic approach to advancing medical research through data-driven,
theoretically grounded models. The results underline SEM's potential to improve
diagnostic tools, support clinical decision-making, and contribute to the
development of personalized healthcare strategies.
that enables researchers to analyze complex relationships among observed and
latent variables. While its origins lie in the social sciences, SEM has gained growing
attention in medical and healthcare research due to its ability to model multifactorial
phenomena and uncover latent structures underlying observable clinical outcomes.
This thesis explores the integration of SEM into various facets of health sciences,
particularly in the evaluation and prediction of cardiovascular disease risk using
non-invasive imaging and clinical indicators. By utilizing latent constructs such as
life-design behavior, biochemical markers, and vascular features—derived from
ultrasound-based texture analysis of the carotid artery—SEM models are developed
to examine the direct and indirect effects of these variables on cardiovascular
health.
Through a series of case studies, including preliminary work on the intima-media
complex (IMC) segmentation and more advanced modeling in published research
(for example, Section 3.2 and Section 3.3), the thesis demonstrates the effectiveness
of SEM in capturing both measurement and structural relationships in healthcare
data. The methodology emphasizes the importance of confirmatory factor analysis,
path analysis, and the testing of model fit indices to ensure robustness and
interpretability. Applications discussed include modeling the influence of
behavioral, demographic, and physiological factors on disease risk, as well as
evaluating the systemic structure of healthcare delivery systems, such as the
General Health System of Cyprus.
Ultimately, this thesis aims to showcase SEM not just as a methodological choice,
but as a strategic approach to advancing medical research through data-driven,
theoretically grounded models. The results underline SEM's potential to improve
diagnostic tools, support clinical decision-making, and contribute to the
development of personalized healthcare strategies.
File(s)![Thumbnail Image]()
Name
PhD-Thesis GE.pdf
Size
4.95 MB
Format
Adobe PDF
Checksum (MD5)
0f73fd5706d38ffadc117a065c123196