Hello all -
I am so sorry about the Zoom problem today that led our session to be even shorter than it already was - making it even more challenging to cover everything about analytic approaches to health disparities in one session! I have confirmed that the link for next week is correct. For those of you who weren't present, the video should be posted later today.
I encourage you to look back over the slides to get the full richness of the content Dr. Vable was presenting. Some take home points we are hoping you take from this:
1. DAGs, DAGs, DAGs. Understanding the relationship between your variables, including the temporality of them and what pathway you are interested in, as well as what assumptions you are making when drawing the DAG, will help to make sure you are doing the correct analysis (including whether variables should be considered confounders or mediators) and drawing the correct conclusions from that analysis.
2. Assessing mediation is an important analytic approach in health disparities research, as it helps to elucidate mechanisms by which inequities occur, pointing to potential interventions. Using the Baron-Kenny approach is a strategy for rigorously assessing mediation that is preferable over the frequently used sequential modeling (note that it also not substantively more complex, but does rely on a good understanding of your relationships/DAG).
3. Variables that are caused by the exposure and cause the outcome can be treated in three different ways: 1) not included in the analysis, if you are interested in the association between exposure and outcome inclusive of the pathway through that variable; 2) included in a formal mediation analysis; 3) included in the analysis without a formal mediation analysis, if you are interested in the association exclusive of the pathway through that variable, but are not making explicit mediation claims.
4. Effect modification is a real phenomenon in the world, and particularly with respect to health disparities research, where interactions between dimension of oppression are common place. Deciding, on a conceptual basis, which variables should be assessed for interaction as part of the analysis plan allows you to get closer to capturing the reality of the world.
5. Multilevel analysis (which was covered very briefly) allows you to bring contextual-level variables into your models, and avoid the "atomistic fallacy" which assumes all variables can be measured at the individual level. This can be a way to capture the structural and social determinants of health which lead to health inequalities, including structural racism and neighborhood effects.
Please let me know if you have questions!