R for HTA showcase
I was recently invited to present at the 2021 R for Health Technology Assessment (R-HTA) showcase. I talked about the importance of marginalization in HTA and proposed a simple method for this purpose.
In a nutshell, the treatment coefficient of a multivariable regression of outcome on treatment and covariates often informs average effectiveness in HTA. However, this has a conditional interpretation, as opposed to the population-level interpretation that is required for reimbursement decisions made by HTA bodies.
The presentation addresses a specific scenario, involving the marginalization of covariate-adjusted treatment effects in pairwise indirect treatment comparisons. In this case, marginalization over the target covariate distribution is required for compatible comparisons that avoid bias.
For more details on the dangers of combining incompatible effect estimates in indirect treatment comparisons, see my paper (arXiv). For how to address this issue using marginalization, see my papers here and here.