- 1. 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!).
Lorch: the authors consider SES (including maternal insurance status, education, starting trimester of prenatal care, and age) as a mediating factor. The authors built a model that included socioeconomic factors and interpreted the ORs for racial/ethnic groups as the part of the effect of racial/ethnic group on fetal death that was not mediated by sociodemographic factors. Through this approach, the authors were able to highlight socioeconomic factors as the primary factors for Hispanics associated with higher fetal death rates compared to non-Hispanic Whites. Furthermore, their analysis demonstrated that among all 3 racial/ethnic minority groups, socioeconomic factors mediated at least some percentage of the disparity in fetal death risk.
Headen: the authors here considered SES as a confounder. Their justification for this is that race and SES are highly correlated which makes SES difficult to control for. To measure SES, they used the following data points from the NLSY79 survey: past year employment, participants’ mother’s years of education, and income (accounting for family size). They then controlled for this in their adjusted analysis. They conducted generalized estimating equations and analyzed both the crude and adjusted associations. They authors conducted a sensitivity analysis because they noted that “missing” data may be more likely to come from specific SES and/or race/ethnic groups.
Merlo: In this paper the authors demonstrate how to incorporate SES into a multilevel regression analysis (MLRA) as a way of measuring variation within and between “clusters”. In this approach, SES is considered a “contextual phenomenon”; a concept that is integral to social epidemiology and important in and of itself, not as a data point to “manage” as a confounder or mediator. I think this is the most complex and compelling approach we’ve seen so far, as it forces us to account for SES and social factors in a more thoughtful way than merely imputing them into statistical models. This approach is particularly useful in that it allows us to analyze the effects of social factors at the individual and geographic (community) level.
- 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 area that I am interested in studying is the under-availability of opioid therapy for essential pain management for cancer, HIV or end-of-life care. This is a particular issue in many Eastern-European, south Asian and African countries (and counter to the opioid crisis caused by over-availability in many Western countries). Access to opioids is largely determined at a structural level by policies that restrict access altogether or make accessing opioids exceedingly difficult. Use of opioids at end of life care can also be dictated on an individual level by religion, cultural acceptability and fear of addiction. I think it would be very interesting to assess this issue using a multilevel analytic approach outlined in the Merlo paper. An example of how I could do this would be to pick a country with varying levels of access to opioids (for this example, Uganda, Hungary). I would measure licit opioid consumption on the individual level (measured as morphine equivalent dose per day) and also measure licit opioid consumption by geographic “clusters” at the regional level. Using MLRA I would be able to evaluate changes in opioid consumption at the individual and regional level compared to the national mean consumption. I hypothesize that there will be differences in regional consumption based on proximity to the larger cities. I would expect that rural clusters will have low access to opioids both at the regional levels and at the individual level – I image this would be due both to geographic proximity to large state hospitals and also greater fear around opioids on an individual level in more rural regions.
- 3. Respond to one other person's post on the forum with a comment or suggestion.
(See response to Maricianah)