Multivariable regression - adjusting for covariates in an RCT
Bireshwar Sinha will speak about: Multivariable regression to analyze data from a Randomized controlled trial (RCT) of community-initiated Kangaroo Mother Care and postartum depression: Why and when to adjust for covariates in an RCT?
A good analytic epidemiologic study should ensure that the estimated association, such as the effect of an intervention, is not affected by selection/confounding bias or by information bias. The purpose of multivariable models is to control for potential confounding bias, thereby enhancing the likelihood that the measure of association is a true representation of a causal effect.
In a perfectly conducted large randomized controlled trial (RCT), all factors other than the intervention are expected to be distributed equally between its two arms. But, especially in trials with small sample sizes, there is a certain chance that a potentially confounding variable may be unequally distributed between the two arms. The question is then, should we adjust for such a potential confounder and, if so, how should we adjust for it?
In this presentation, I plan to discuss why and when to adjust for covariates in an RCT using an example where we estimated the effect of promoting community-initiated Kangaroo mother care on the risk of postpartum depressive symptoms
We use Adobe Connect for the webinars; the link for these meetings is: https://connect.uninett.no/cismac