Biostatistics Seminar Series: Elizabeth Halloran, MD, DSc

Title: From dependent happenings to causal inference with interferenceAbstract: Sir Ronald Ross developed a Theory of Happenings that distinguished independent from depending happenings. Dependent happenings are those in which the frequency of the happening depends on the number of individuals affected such as in most infectious diseases. Due to dependent happenings, interventions in infectious diseases can have several different kinds of effects. Demonstrating indirect effects of vaccination can have important consequences for public health policy. Causal inference using potential outcomes is an experimental paradigm to define estimands and estimators of causal effects of interventions. In infectious diseases, the potential outcome of an individual can depend on the treatment status of other individuals, called interference in causal inference. Defining causal estimands and estimators under interference is complicated. Here we present causal estimands and estimators of direct, indirect, total and overall effects of interventions under interference, with emphasis on vaccination in infectious diseases. We present these in both the two-stage randomized, one-stage randomized, and observational context.  We present examples of estimating these different effects for different vaccines using different approaches. We present recent trends in developing these methods that include using networks to allow for individual interference sets, for structured interference, and randomized experiments in large networks. Demonstrating indirect effects of vaccination can have important consequences for public health policy.