[Causal graph model and its application in nutritional epidemiologic research]

D Tang,X Xiao,F Yang,Y F Hu,J Z Yin,X Zhao
DOI: https://doi.org/10.3760/cma.j.cn112338-20200805-01025
2021-10-10
Abstract:Suboptimal diet is one of the most important controllable risk factors for non-communicable diseases. However, randomized controlled trials make it difficult to quantify the causal association between specific dietary factors and health outcomes. In recent years, the rapid development of causal inference has provided a robust theoretical and methodological tool for making full use of observational research data and producing high-quality nutritional epidemiologic research evidence. The causal graph model visualizes the complex causal relationship system by integrating a large amount of prior knowledge and provides a basic framework for identifying confounding and determining causal effect estimation strategies. Different analysis strategies such as adjusting confounders, instrumental variables, or mediation analysis can be created based on other causal graphs. This paper introduces the idea of the causal graph model and the characteristics of various analysis strategies and their application in nutritional epidemiology research, aiming to promote the application of the causal graph model in nutrition and provide references and suggestions for the follow-up research.
What problem does this paper attempt to address?