Interventional Effects with Latent Mediators: An Example from Life Course Epidemiology

Department of Epidemiology Seminar Series

Open to the public.

Speaker:

Bianca De Stavola, PhD
Professor of Medical Statistics
Population, Policy & Practice Department
UCL Great Ormand Street Institute of Child Health

Abstract:

Life course epidemiology is important for understanding the effects of lifetime exposures on health outcomes, and for suggesting interventions to improve health,  but it is complex and challenging.

Conceptual models can help  make sense of the many possible pathways via which an exposure in early life may affect health outcomes in adulthood, while counterfactual reasoning can help specify relevant targets of estimation.

Using a study of eating disorders in adolescent girls, the talk will discuss alternative conceptual models to represent the role that body size at increasing ages in childhood may play in the onset of eating disorders in adolescence. Different causal questions can be asked and expressed for example in terms of natural effects of early childhood size mediated or not mediated via later childhood sizes. These questions will then be expanded to consider the best time for intervening to prevent eating disorders. Here, estimands such as the direct and indirect interventional effects will be discussed, involving a single or multiple mediators, as well as observed and latent mediators.