Hierarchical Models and Bayesian Structural Equation Models with Application in Health Services Research

Health Services Research applications often involve multivariate data which also have inherent complex hierarchical structures. Examples are patients clustered in medical centers, multiple clinical treatments in a patient , and so on but also in other research areas there are ample examples to find, such as children in schools, teeth clustered in mouths, teats clustered in udders, etc. Ignoring this correlation among observations in a cluster may lead to incorrect estimation of parameters of interest because of ignoring the correlation structure in the data. As a consequence the conclusions may be wrong and therefore misleading. The increasing complexity in study designs and research questions add to the need of novel statistical tools that fully exploit the nature of the data and adequately answer the research questions. In this thesis, we exploited the correlated structure of the data and made use of existing statistical procedures, and when necessary, have extended them. We have two aims. First, to advance the knowledge of the relationship between organizational features of nursing care and patient outcomes by making use of Bayesian multilevel structural equation models (SEMs) that reflect the complex hierarchical structure of the motivating data. Second, we aim to critically evaluate the performance of existing Bayesian SEM methods. In particular, we aim to explore the performance of joint multilevel factor analytic models and multilevel covariance regression models. We analyse the European Union's three-year (2009-2011) Registered Nurse Forecasting (RN4CAST) study data and data from a three-year (2017-2019) epidemiological study of adolescent girls and young women in Nairobi slums. We also carry out a simulation study to assess the merit of the proposed method. We implement the models in Mplus, SAS and in JAGS running interactively from R . The relevant code can be found online as supplementary material of the published papers or from the first author upon request. The approaches considered here are motivated in the context of Health Services Research, but they are relevant to a wide variety of fields.

Jaar van publicatie:2022

Toegankelijkheid:Open