Evaluating context's influence on health disparities: An illustrative example of multilevel structural equation models

Evidence shows that influences on health disparities occur across individual and contextual levels. Understanding the contextual aspects of health disparities requires contextual measures. However, rarely does existing data map directly onto the contexts of specific studies. Researchers can use individuals' responses to questions about their context to develop contextual measures, but this is fraught with methodological hurdles. Multilevel structural equation models (ML-SEM) offer an innovative method for addressing these. However, ML-SEM have seen little exposition and confusion exists over their use. In this presentation, I address this.

Methods: Using data came from the 2008 Ohio Family Health Survey, a large, representative survey of adults in Ohio (n = 50,944), I take an applied approach and illustrate the use of ML-SEM. I use individuals' responses to questions about their access to care to simultaneously examine: which variables predict disparities in an individual's access to health care and which predict county-level disparities in the “health care access environment.”

Results: ML-SEM indicated that both individual- and contextual-level disparities in health care access occur. At the individual level, minority status, income, and education predicted disparities. Contextually, the percent of minorities in a county and median county income predicted contextual disparities among other variables.

Conclusions: Results show that using a ML-SEM approach allows one to partition variance in disparities across levels and also more accurately estimate disparities by partialing out random measurement error. Efforts to eradicate disparities will proceed more successfully when incorporating ML-SEM-based studies that more fully recognize and identify disparities at multiple levels.