Week 8 HW

Week 8 HW

by Timothy -
  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.
    1. As demonstrated in the Lorch paper, SES can be accounted for as a mediator on the casual pathway from racial/ethnic health disparities to the outcome of interest. Lorch conducts a mediation analysis to attempt to attribute a percentage of racial/ethnic disparities explained by SES and the other identified mediators and to described how these differ by race.
    2. The Headen paper provides an example of treating SES as a confounder of race/ethnicity and they control for SES as a covariate within their regression analysis and compared crude vs adjusted regression results.
    3. The Merlo paper addresses SES as a contextual phenomenon – suggesting that individuals with similar characteristics may have different health outcomes according to different contexts including geographic and SES clustering. Merlo proposes that rather treat contextual factors as confounders to be controlled, contextual factors should be viewed as an important causative factors to be explained and understood. The paper proposes using multilevel regression analysis and measuring variation within and between clustered groups (i.e. neighborhoods or SES classes) to best understand these contextual factors.

                                                                                        

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.

While not exactly a health disparities focus, my research examines variation in prescribing of diabetes medications which involves a lot of thinking about multilevel influences. Our current work is looking at the impact of changes made to outpatient diabetes mediations when older adults are hospitalized for unrelated conditions (i.e. pneumonia). The conceptual framework is that physicians focus on inpatient blood sugar control/numbers that do not represent outpatient/long-term disease control and this may lead to overtreatment and subsequent adverse events. Our primary outcome is dichotomous, intensification of diabetes medications after hospitalization. The levels of influences on this outcome include patient characteristics (inpatient diabetes control, outpatient control, comorbidities, etc), provider characteristics (medical training, trainee status, age, gender) and health system characteristics (academic hospital, hospital size, geographic region). We plan to use Poisson regressions to obtain direct estimates of relative risks including the above influences as predictors within our regression analysis.

Re: Week 8 HW

by Shabnam Peyvandi -

I think a multi-level approach to your research question makes sense particularly for the health system characteristics that can influence the outcome. For example, the analysis can be done by clustering hospitals in different geographical areas. 

Re: Week 8 HW

by Christine Dehlendorf -

Thanks for the reply! In terms of SES as a confounder, as we talked about in class, this is complicated because the direction of the causal arrow from SES to race/ethnicity does not make sense - so it is really more of a "third variable" whose treatment depends on what your research question is (e.g. do you want to know what the race/ethnicity association is independent of certain SES variables, acknowledging they are usually imperfect measures, or do you want to know the association of an outcome with race/ethnicity without controlling for the mediating effect of SES). Of course, most people just call it a confounder!

Also important to think about effect modification as a way of modeling SES.

I agree with the other comment that multilevel analysis, treating providers, clinics, or hospitals, is a common way to analyze health services issues like the one you are studying. As I talked about in the health care disparities lecture, this can help elucidate causes of disparities in health care - i.e. how much is intra- vs. inter- facility.