Causal Inference in Perinatal Epidemiology: Revisiting the Birth Weight Paradox

Does smoking cause neonatal mortality? Understanding causation is essential to inform the creation of effective interventions aimed at improving human health. Causal diagrams in the form of directed acyclic graphs (DAGs) have changed the way we approach causal inference. This chapter provides an overview of causal thinking and the use of DAGs in helping to design etiologically-oriented epidemiologic research. It introduces the theory of DAGs, and adjustment for variables when estimating total effects, and when estimating direct and indirect effects is discussed. The "birth weight paradox" is used to illustrate the relevancy of causal analysis to reproductive and perinatal epidemiology. DAGs are used to help estimate different effects, to explain why z-scores remove the crossing of curves, and to help define overadjustment. DAGs provide a tool to aid in formulation of a research question and its accompanying analytic plan, and hence help ensure that the intended question is answered and the science interpreted within a formalized causal paradigm.