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.
Approach 1: Adjust for SES (whichever way it was measured) like you do with other covariates. Methods of measuring SES may include income, education, housing type, occupation, some combination of the above or others. A benefit of this approach is that you preserve the sample size and thus the ability to detect associations. However, the role of SES in the association being studied just gets “buried” and lost in the model like all the other covariates, and prevents further discussion of a very important variable with important implications in terms of designing interventions to correct the associations being reported on in the analysis.
Approach 2: Stratify the analysis by SES and develop separate models to examine the associations of interest in each strata of SES (again, however it was measured). A benefit of this approach is that you can understand the associations for communities/types of people (e.g., people who make more money versus less), which may have implications for the interventions being proposed. HOWEVER, the sample size is reduced and may make it harder to detect associations as well and thus understand true effect sizes.
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 multi-level investigations.
Question = Is SES associated with omega-3 fatty acid levels in patients with coronary artery disease (CAD)? The exposure is SES, measured in four ways: Education level, Household income, Occupation, and Housing status. The outcome will be measured levels of omega-3 fatty acids in the blood. The population of patients will be those with known CAD. Traditional cardiovascular risk factors do not completely explain the epidemiological association between SES and cardiovascular risk. Covariates for adjustment will include traditional cardiovascular risk factors , demographics and factors implicated in omega-3 fatty acid levels such as age, sex, ethnicity, smoking, marital status, regular alcohol use, BMI, and physical activity, statin use, renal function. Analytically, I would first describe how omega-3 levels vary with increasing levels of each of the four SES variables. Next, I would show unadjusted associations between each of the four SES predictors and omega-3 levels. Finally, I would show adjusted levels for those same associations. I would not create a composite outcome variable since each of these four SES predictors is capturing a different aspect of SES that requires its own interpretation. Again the purposes here is to dissect the variability of omega-3 levels with respect to SES. It is reasonable to guess that food scarcity is not just related to income but may also be related to housing, for example. So access to healthy food needs to be understand from multiple angles and not with a composite outcome.