Biostatistics Weekly Seminar

In causal inference, it may be of interest to estimate treatment effects of variables that are continuous and measured at multiple time points. However, existing estimation options are limited in this case - especially if the true intervention-outcome relationship (dose-response curve) should be described as closely as possible; that is, the research question should not be changed. These situations may, however, be of relevance: for example, in pharmacoepidemiology, one may be interested in how counterfactual outcomes of people living with -and treated for- HIV, such as viral failure, would vary for time-varying treatments (i.e., interventions) such as different drug concentration trajectories. One challenge for doing causal inference with continuous interventions is that the so-called positivity assumption is typically violated: this is the requirement that individuals have a positive probability of continuing to receive treatment according to the assigned treatment rule, given that they have done so thus far and irrespective of the covariate history. To tackle such positivity violations, we develop different projection functions, which reweigh and redefine the estimand of interest based on functions of the conditional support for the respective interventions. With these functions, we obtain the desired dose-response curve in areas of enough support, and otherwise a meaningful estimand that does not require the positivity assumption. We develop g-computation type plug-in estimators for this case. Those are contrasted with using g-computation estimators in a naïve manner, i.e. applying it to continuous interventions without specifically addressing positivity violations. The ideas are illustrated in the context of longitudinal data from HIV+ children treated with an efavirenz-based regimen as part of the CHAPAS-3 trial, which enrolled children <13 years in Zambia/Uganda.

Zoom Link to Follow
09 November 2022
1:30pm