1. It is often mentioned that racial/ethnic differences in health should not be investigated without consideration of socioeconomic position/status. Describe two ways to account for SES in an analytic model and the interpretations/implications of each approach.
(I’m not sure whether I am interpreting this question correctly. I am reading it as ‘what analyses can we use’ vs. ‘what measures of SES do we use.’ I am answering the former, though at least one classmate answered latter, and both seem plausible from the wording of this question…)
Assuming that SES is operationalized and measured thoughtfully, one might run analyses, such as a regression, in which racial/ethnic group membership is predicting a health outcome of interest. For example, as part of my dissertation, I examined mental health symptoms among Asian American and White young adults. My analyses showed that Asian American young adults had higher depressive symptoms than White young adults. To account for SES in this type of analytic model, one strategy might be to include SES in the regression model to examine whether the effect of race/ethnicity on depressive symptoms persists, attenuates, or becomes non-significant. If it persists, the findings would suggest that race/ethnicity significantly predicted depressive symptoms, even after accounting for SES. If it attenuates, then SES is partly accounting for the differences in depressive symptoms. If the effect of race/ethnicity becomes non-significant, the findings would suggest that race/ethnicity was confounded with SES, and it was actually SES that had an effect on depressive symptoms. Another way to test this might be to use a mediation analysis, in which race/ethnicity is the independent variable, SES is the mediator, and the health outcome is the dependent variable. If SES is a significant mediator, the interpretation would be that the effect of race/ethnicity on depressive symptoms was explained by differences in SES. E.g., Asian Americans had lower SES than White Americans, which explained relatively higher depressive scores among Asian Americans. The challenge, however, might be in running mediation analyses with categorical data, and the researcher would have to consider using techniques such as structural equation modeling.
2. Select a research question investigating associations between multi-level social factors (operating at least two levels) and a health outcome. State the exposures and outcomes, the additional study covariates that would be included in the analytic model, and a discussion of the analytic considerations in an multi-level investigations.
My area of research is on the influence of food insecurity on cigarette smoking (not a health outcome per se, but a health behavior that is associated with many health outcomes). Food insecurity could operate on multiple levels, as it can be affected by your own income and your other competing expenses, your peer/social networks (e.g., some low-income individuals/families are less food insecure than others because they can rely on their family and friends to help out with food), and where you live (e.g., differences in food access; cost of housing). The additional study covariates might then be income, number of children and adults in the household, SNAP participation, neighborhood SES (e.g., % of poverty), and other variables associated with smoking (such as gender, age, race/ethnicity, substance use, depression). In terms of analytic considerations, I would have to think about multicollinearity in the model, particularly for multilevel variables, as neighborhood SES might be significantly associated with individual income, etc. I would also have to think carefully about the number of independent variables I am including given the sample size.