Week #8

Week #8

by Emily -

  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. Draw a DAG for each option and reference it in your response (you do not have to post this!).

 

SES can be accounted for by approaches taken when building models, how factors are accounted for within models as mediators or confounders, and by contextualizing research in meaningful ways statistically. Merlo discusses using multilevel epidemiology to understand differences in health which relate to place because of the patterns of health variation related to a population's location and possible geographic segregation. The use of clustering in this article clearly describes how to understand differences at different units of analysis (individual, neighborhood, city). Headen's use of generalized estimating equations provided the opportunity to determine risk ratios and control for multiple relevant covariates. Lorch constructed regression models and compared risk after sequentially adding mediating factors. This method allowed the risks to be compared after each addition of factors.

 

  1. 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.

 

Contraception use among women living in resource limited settings is not infrequently studied issue and so multilevel approaches have been taken in the literature to further its understanding. There are a number of individual level factors such as age, educational level, rural or urban home, religion, and communication about family planning with family and partner. Community level factors include mean number of children in those families living in the community, family planning messages available in the community, approval of family planning in the community, and modern infrastructure in the community. These community factors lend themselves well to be analyzed using clustering because of the interconnectedness of many communities that are considered resource limited – both due to space and social factors but also due to economic and political realities.

 

  1. Respond to one other person's post on the forum with a comment or suggestion.

 I'll respond to Amy -

Re: Week #8

by Nicholas Rubashkin -

I think a clustering/MLRA analysis could be really interesting for family planning.  I think may tend to assume a "one size fits all" model for family planning uptake when we use models that regress on individual outcomes  (e.g. unintended pregnancies,  individual uptake of a long-term method).  I wonder if a contextual approach would lead us to consider factors (and therefore, interventions) that don't begin and end with the individual woman's uptake of a family planning method.  Maybe, a MLRA approach would expose variations in community level factors such as level and type of employment (% women employed in wage labor may be related to disposable income for long-term methods?) , labor migration patterns (do men travel great distances for work and are absent for long periods of time, and therefore long-term FP methods are utilized less?).  Trying to think of community level factors that would illuminate the variation in FP metrics rather than thinking there is "one metric", like the ideal length of time for a long-term method.

Re: Week #8

by Christine Dehlendorf -

Thanks Emily! I love Nick's comments about the contextual factors related to economics, etc. that influence family planning. I know you know I am interested in the social communication/messaging piece you bring up. There was literature in the 1990s in developing countries, published by one of the collaborators on the project you are analyzing, that showed that contraceptive use was influenced by the use of contraceptives by other people in your network, measured as a contextual variable. 

For how to model SES, I just want to make sure that you got that the options include as a "confounder"  - although your use of them term "covariate" is more appropriate, effect modifier, and mediator. In addition, contextual SES factors can be considered.