A0286
Title: Structural equation modelling for diagnostic test accuracy meta-analysis
Authors: Zelalem Negeri - University of Waterloo (Canada) [presenting]
Abstract: Diagnostic test accuracy meta-analysis is a rapidly growing and active area of research. Standard approaches to this type of meta-analysis utilize hierarchical or bivariate random effects models that account for the within- and between-study heterogeneity in test characteristics. However, these approaches usually fail to converge for sparse data types or lead to biased inferences for meta-analysis with few studies. Moreover, the established methods cannot handle complex relationships between primary studies (e.g., direct and indirect effects or moderators and mediators), missing data, and adjusting for potential confounding variables. Therefore, the aim is to develop a structural equation modeling-based framework for aggregate data meta-analysis of diagnostic test accuracy studies to overcome these limitations of the standard methods. The proposed method is demonstrated and validated via extensive simulation studies and real-life data examples.
