- What are the different ways to account for SES in an analytic model when investigating racial/ethnic health disparities? (Hint: you should have three options). Discuss the interpretations/implications of each approach as it relates to the interest in understand health disparities by race/ethnicity. Draw a DAG for each option and reference it in your response (you do not have to post this!).
Different indicators of SES include income, education, insurance status. SES can be accounted for in various ways in an analytic model when investigating health disparities. The first is that SES can be thought of as a confounder, the second is SES can be a mediator, and the third way to account for SES is to stratify based on SES.
When thinking of SES as a confounder, the DAG would illustrate a forward arrow from race/ethnicity (the effector) to health disparities (the outcome of interest). Then, SES would be a separate node with arrows from SES to both race/ethnicity and to the health disparities. This suggests that SES influences race/ethnicity and also influences health disparities. It can be understood how SES can influence health disparities either on the individual or population level. However, it is challenging to see how SES would influence a person’s race/ethnicity. Though if race/ethnicity is self-reported and someone is of mixed race, they may choose to report their ethnicity to be only that of the neighborhood where they live. In the analysis, SES would need to be controlled for since it acts here as a confounder.
When thinking of SES as a mediator, the DAG would illustrate a forward arrow from race/ethnicity (the effector) to health disparities (the outcome). SES would be a separate node and the arrow would point from race/ethnicity to SES and another arrow from SES to the health disparity of interest. If this were the case, it would imply that race/ethnicity influences SES and in turn, SES influences the health disparity. This positions SES along the causal pathway between race/ethnicity and the health disparity. Here, we can understand that race/ethnicity influences where someone lives, how much income they can accrue, and what the quality/level of education achieved is. Likewise, SES can play a direct role in the health disparity of interest. In the analysis, SES does not need to be controlled for, since it is a mediator.
A third way to account for SES in an analysis of the effect of race/ethnicity on health disparities is to stratify the analysis by SES (either education, income, zip code, etc.). This is one way to elucidate the role that SES plays in this relationship.
2. Think about multilevel influences on a health outcome of interest to you. Discuss how you would study this, including measurement and analytic approaches you would use to account for exposures across multiple levels.
One health outcome that interests me is outcome after primary brain tumor diagnosis as influenced by SES. My hypothesis is that patients with low SES have worse outcomes than those with higher SES. In the multilevel influence model addressing this issue, it is important to account for community/neighborhood level influences and individual level influences. Neighborhood of residence level influences include distance to hospital that provides the necessary health care services (i.e. medical oncology, radiation oncology, surgical care, primary care, MRI), available transportation to the health care facilities, environmental exposures that may predispose people to develop primary brain tumors. Individual level influences include income, health literacy, language spoken (and if this is congruent with the health care provider), and level of education attained. The Merlo et al article from this week’s readings addresses the concept of multilevel regression analyses as a way to study this type of phenomenon. The authors describe using an interclass correlation analysis to look for clustering of outcomes (in this example, it would be outcome after brain tumor diagnosis) by neighborhood rather than by individual. Another option to analyze the relationships would be to stratify the analysis based on SES. For instance, both the crude and stratified (by SES) risk ratios for the different outcome measures following primary brain tumor diagnosis could be calculated and presented.