Year: 2017
Type: dissertation
Abstract: Introduction: Health disparities are a major public health concern. Researchers often seek to explain whether these disparities are attributable to modifiable downstream risk factors that are also predictive of outcomes. Traditionally-used mediation approaches may not be appropriate to explain such disparities, due to the complex nature of socio-demographic factors such as race/ethnicity. Methods: Aim 1: We carried out a methodological review of the literature. We searched for studies that investigated a racial/ethnic disparity in a health outcome and adjusted for a potential modifiable mediator of the disparity. Out of the 969 studies meeting inclusion criteria, a simple random sample of fifty studies were drawn to undergo thorough data abstraction. Data were abstracted to determine what mediation approaches were employed, if causal language was used to describe these estimates, and if relevant assumptions were considered in justifying this causal language. Aim 2: We also carried out an analysis of National Survey of Family Growth (NSFG) data representing 103,919,318 live singleton births to mothers in the United States. This analysis employed traditional (Difference and Product methods) and general mediation methods (Inverse Probability-Weighted Marginal Structural Models and Structural Transformation) to result in different estimates of how payment method for delivery explained the racial/ethnic disparity in low birth weight. Results: All studies included in the target review employed traditional mediation approaches, the majority used language that could be interpreted as causal (86%), and none of the studies explicitly addressed assumptions or conditions for interpreting estimates causally. In the analysis of NSFG data, for every 100 births, there were an additional 5.27 low birth weight infants born to non-Hispanic Black mothers, compared to births to non-Hispanic White mothers. Estimates of the counterfactual disparity measure ranged from 1.62 using the product method to 4.66 using the structural transformation method. Conclusion: Traditional mediation approaches are commonly used to explain racial/ethnic disparities. The use of such traditional approaches to explain disparities along the lines of complex socio-demographic variables may be inappropriate due to underlying assumptions that are likely broken. Considerations of these underlying assumptions and use of appropriate mediation approaches are crucial to understanding the important public health issue of health disparities.
Author: Loren J. Schleiden
Institution: University of Pittsburgh
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2738543665
