Application of the nonlinear Blinder-Oaxaca decomposition to study racial/ethnic disparities in antiobesity medication use in the United States

Background: The nonlinear Blinder-Oaxaca (BO) decomposition method is gaining popularity in health services research because of its ability to explain disparity issuesThe present study demonstrates the use of this method for categorical variables by addressing antiobesity medication use disparity Objective: To examine racial/ethnic disparity in antiobesity medication use and to quantify the observed factor contribution behind the disparity using the nonlinear BO decomposition Methods: Medical Expenditure Panel Survey data, 2002-2007, were used in this retrospective cross-sectional studyAdults with body mass index (BMI) >30, or BMI ≥27 and comorbidities such as hypertension, cardiovascular diseases, diabetes, or hyperlipidemia were included in the cohort (N=65,886,625)Multivariable logistic regression was performed to examine racial/ethnic disparity in antiobesity medication use controlling for predisposing, enabling, and need factorsThe nonlinear BO decomposition was used to identify the contribution of each predisposing, enabling, and need factors in explaining the racial/ethnic disparity and to estimate the residual unexplained disparity Results: Non-Hispanic Blacks were 46% (odds ratio [OR]: 0.54; 95% confidence interval [CI]: 0.35-0.83) less likely to use antiobesity drugs compared with non-Hispanic Whites, whereas no difference was observed between Hispanics and non-Hispanic WhitesA 0.22 percentage point of disparity existed between non-Hispanic Whites and BlacksThe nonlinear BO decomposition estimated a decomposition coefficient of -0.0013 indicating that the observed disparity would have been 58% higher (-0.0013/0.0022) if non-Hispanic Blacks had similar observed characteristics as non-Hispanic WhitesAge, gender, marital status, region, and BMI were significant factors in the decomposition model; only marital status explained the racial/ethnic disparity among all observed characteristics Conclusions: The study revealed that differences in the predisposing, enabling, and need characteristics (except marital status) did not successfully explain the racial/ethnic disparity in antiobesity medication useFurther studies examining racial/ethnic differences in individual beliefs, behavioral patterns, and provider prescription patterns are vital to understand these disparities