Week 8: Assignment:
- 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!).
ANSWER: SES covariates can be accounted for in an analytic model (such as a multivariate regression equation) by treating the SES covariate as a confounder (test for association between SES and race/ethnicity, test for ~ between SES and health outcome, test for ~ between race/ethnicity and health outcome. If you add the SES to the model and retest for ~ between race/ethnicity and health outcome and the association is gone, then SES is most likely predicting the health outcome, not race/ethnicity (in reality, race/ethnicity does not predict the health outcome).
SES covariates can be accounted for by testing to see if the SES is a mediator, in which the association between race/ethnicity and the outcome is real, but can only occur or be mediated via the SES covariate.
Several SES covariates can be taken into account using a multilevel regression analysis. So a SES covariate like neighborhood can be aggregated, variance calculated and the health outcome compared within & between neighborhood groups and between individual level variance. The total variance of this health outcome can be described as the sum of the neighborhood and individual level variances. That tells the researcher out of all of these evaluated SES (system, local community or individual) or race/ethnicity, which one contributes the most to the variability seen in the health outcome.
- 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.
ANSWER: Multilevel multivariate statistical analysis in order to account for variability in my health outcome (change in HbA1c due to metformin treatment) that comes from SES and NOT genetics. I would choose the following community level and individual level covariates to add to the model shown below. The MLSA analysis would explain for each covariate the amount of variance in “change in HbA1c” that the covariate accounts for AND whether or not those covariates correlate with each other (ICC).
Local community level covariates = neighborhood, distance from a produce-selling grocery store (would measure both of these SES by obtaining residential zip code of each subject. Individuals who live in the same zip code would be grouped together in neighborhoods. Grocery store information would be collected via google maps and distance from the grocery store address to the zip code would be calculated)
Individual level covariates = weight/BMI, baseline HbA1c, race/ethnicity, income level (BMI, self-reported race and baseline HbA1c would be gathered from the EMR, and BMI calculated from height and weight information. Income level would be gathered via tax returns).
- Respond to one other person's post on the forum with a comment or suggestion.
ANSWER: In response to Tene Cage: In addition to SES that affect access to healthcare, would you think about adding SES that affect level of stress? Particularly, when it comes to a cancer diagnosis, the patient's perceived stress level (family stress, mental stress, physical stress, etc) can affect treatment and response to treatment.