Medicine
Structural Equation Modeling in Health Disparities Research
Quick fact
SEM can simultaneously test multiple pathways—such as how education, stress, and healthcare access together produce health disparities—revealing which intermediate steps are most influential and could be targeted by interventions.
Why this is interesting
You've probably seen headlines like 'Income affects lifespan.' But how do we actually test the pathways through which money gets under the skin to influence health?
Read the full explanation
Understanding Structural Equation Modeling in Health Disparities Research
Imagine you want to understand why people with lower income tend to have worse heart health. It's not simply about buying medicine. Many factors are at play: stress from financial strain, limited access to healthy food, and living in high-crime neighborhoods. These are all related, and they influence each other. Structural equation modeling (SEM) is a statistical tool that lets researchers draw a map of these relationships and test if the data match that map. Think of it like a flow chart: income (the starting point) has arrows pointing to stress, diet, and healthcare, and those arrows then point to heart health. SEM quantifies the strength of each arrow, so we can see which pathway matters most. It also allows researchers to use 'latent variables'—concepts like 'socioeconomic status' that aren't measured directly but are inferred from multiple indicators like income, education, and occupation. This is powerful because it gives a more complete and accurate picture than looking at one factor at a time.
A deeper explanation
At its core, SEM combines factor analysis and path analysis. The key mechanism is that it estimates a set of linear equations that represent the hypothesized network of causal relationships. Each equation describes how a variable (or latent factor) depends on others, and the model simultaneously solves all equations to find the best-fitting parameters. Crucially, SEM separates the measurement model (how latent variables relate to their indicators) from the structural model (how latent variables relate to each other). This separation allows SEM to account for measurement error, meaning that the relationships between latent constructs are not attenuated by noise from imperfect measurement. In health disparities research, this is vital because concepts like 'stress' or 'social support' are hard to measure precisely. By modeling them as latent variables with multiple indicators, researchers can estimate more accurate effects. SEM also enables mediation analysis: it can decompose the total effect of a predictor (like SES) on an outcome (like mortality) into direct and indirect effects through mediators (like access to care). This helps identify which mediating pathways are most modifiable and thus where interventions might be most effective. However, SEM is not a magic wand for causal inference. The model's validity hinges on correct specification of the model and the assumption that there are no unmeasured confounders. Researchers must ground their models in strong theory and test alternative models to build confidence in the causal story.