25 Jun 2020 09:30am to 10:30am

Mediation effects that emulate a target randomized trial: Evaluation of ill-defined interventions on multiple mediators

Event Location
Zoom videoconferencing
Margarita completed a PhD in Biostatistics at Université Paris-Sud in 2014. She is currently an ARC DECRA Fellow and conducts methodological research in the areas of causal inference, missing data...

Many epidemiological questions concern potential interventions to alter the pathways presumed to mediate an association. 

For example, we consider a study that investigates the benefit of interventions in young adulthood for ameliorating the poorer mid-life psychosocial outcomes of adolescent self-harmers relative to their healthy peers. Two methodological challenges arise. Firstly, mediation methods have hitherto mostly focused on the elusive task of discovering pathways, rather than on the evaluation of mediator interventions. Secondly, the complexity of such questions is invariably such that there are no existing data on well-defined interventions (i.e. actual treatments, programs, etc.) capturing the populations, outcomes and time-spans of interest. Instead, researchers must rely on exposure (non-intervention) data to address these questions, encountering the problem of ill-defined interventions. We address these challenges by specifying a target trial addressing three policy-relevant questions, regarding the impacts of hypothetical (rather than actual) interventions that would shift the mediators’ distributions (separately, jointly or sequentially) to distributions that can be emulated with exposure data. We then define novel interventional effects that map to this trial, emulating shifts by setting mediators to random draws from those distributions. We show that estimation using a g-computation method is possible under an expanded set of causal assumptions relative to inference with well-defined interventions. These expanded assumptions reflect the lower level of evidence to be expected with ill-defined interventions. Application to the self-harm example using data from the Victorian Adolescent Health Cohort Study illustrates the value of our proposal for informing the design and evaluation of future interventions.

Margarita leads research that focuses on the development of statistical methods that are motivated by the issues arising in clinical and population health research projects on which she collaborates, with particular focus in the areas of causal inference, missing data and survival analysis. Her emerging leadership in biostatistics is reflected by her roles in the management teams of the Victorian Centre for Biostatistics (ViCBiostat) and the MCRI's LifeCourse Initiative, comprised of over 40 longitudinal cohort studies. She is recipient of a Discovery Early Career Researcher Award (DECRA) Fellowship in Statistical Methods from the Australian Research Council.