|
Abstract:
|
While causal mediation analysis has been developed to study the mechanisms, or mediators, through which a treatment or exposure affects a health outcome, its extension to a repeatedly measured mediator and/or outcome is less clear. The use of (discrete) structural equation models has the drawbacks that inference is confined to the particular measurement times used and does not recognize the continuous nature of the mediation process over time. To overcome such limitations, we present a new continuous time causal mediation approach that uses differential equations to describe the dynamic relationships among model variables over time. This approach is applied to data from a cohort of children followed from birth to 3 years in which dental caries and behavioral risk factors were measured repeatedly. We assess the direct effect of socioeconomic status on dental caries and its indirect effect through consumption of sugary drinks. Through this example and simulations we demonstrate the potential for the proposed approach to provide a more holistic assessment of mediation, as well as better predictions of the effects of future interventions.
|
