1. What are 3 different ways to account for SES in analytic models when investigating racial/ethnic health disparities? Briefly discuss the interpretations/implications of each approach as it relates to understanding health disparities by race/ethnicity.
Three different ways to account for SES when investigating racial/ethnic disparities in health outcomes are as follows:
SES as a confounder: Here we are saying that SES is causally associated with race and that SES is causally associated with health disparities but that it does not lie in the causal pathway (i.e. as an intermediate variable) between race/ethnicity and health disparities. As such, the interpretation is that the relationship between race/ethnicity and health disparities is spurious and that it accounts for part of or all of the association. Therefore, controlling for SES would remove the effect of race/ethnicity on health disparities.
SES as a mediator: Here we are saying that SES is associated with both the exposure and outcome and is an intermediate variable in the causal pathway between race/ethnicity and health disparities (since race/ethnicity systematically relates to SES opportunities). In this scenario, we would do a mediation analysis to account for SES, where adjusting for SES could tell us the direct effect of race/ethnicity on health disparities (apart from its effect on SES). From this we could determine how much of the effect of race/ethnicity on health disparities is mediated by SES by comparing the total effect (without adjusting for SES) to the direct effect (after adjusting for SES).
SES as an effect modifier: Here we are saying that the effect of race/ethnicity on health disparities meaningfully differs according to SES level. Treating SES as an effect modifier would require stratifying the data by SES categories and evaluating the association between race/ethnicity and health disparities in each SES category so see if there is any significant difference in the magnitude of the associations.
2. Describe a potential effect modifier, mediator, or contextual variable (for definition of contextual variable, see Diez-Roux reading) for an association of interest to you and relevant to health disparities. For example, for investigating the association between education and hypertension, I might be interested in evaluating whether the association between years of education and hypertension is different for Black men than for White men. Describe how you would study whether this relationship exists.
Racial/ethnic minorities, particularly American Indians and Alaska Natives, may disproportionately suffer from higher rates of mental illness (eg. depression, substance abuse, PTSD etc.) compared to the general US population. It would be interesting to examine if discrimination is a mediator in the relationship between race/ethnicity and major depression in a diverse cohort. Race/ethnicity can be self-reported, depression could be measured by the Patient Health Questionnaire (PHQ-9), which is a nine-item depression screening instrument, and discrimination could be measured by the Everyday Discrimination Scale. To determine if discrimination is indeed a mediator in the relationship between race/ethnicity and depression, we could do a mediation analysis as described in the previous answer. First, we would confirm that race/ethnicity is associated with depression; race/ethnicity is associated with discrimination and that discrimination is associated with depression. We would then create a model estimating the overall/total effect of race/ethnicity on depression including the effect via discrimination. We would also create a model estimating the direct effect of race/ethnicity on depression, which would exclude discrimination. We could then determine how much of the effect of race/ethnicity on depression is mediated by discrimination by comparing the total effect model to the direct effect model.