An Analysis of Racial Disparities in Inpatient Mental Health Treatment via Multilevel Models

Recent comprehensive studies have found pervasive evidence of racial disparities in healthcare treatments and outcomes. This paper examines whether there are racial disparities in the total charges associated with U.S. inpatient mental health treatment. We used U.S. inpatient admission data with discharge and hospital information from the 2003 Healthcare Cost and Utilization Project's Nationwide Inpatient Sample, and 2003 county data from the Bureau of Health Professions' Area Resource File. We employed multilevel (three-level) regression models, with discharges nested within hospitals within counties, to examine whether there were racial disparities in inpatient total charges. Multilevel model estimates of regression coe±cients associated with race were compared to those from a single-level model; both models found significant disparities, but the three-level model yielded more precise estimates of all model effects and revealed that single-level model estimates of disparities were in some cases overstated and/or in the opposite direction. We also tested for both omitted hospital and county variables using a statistical test known as the Hausman test, and utilized a variant of this test to examine whether a specific observed model covariate (race) was correlated with omitted variables. Both Hausman test analyses suggested significant omitted factors.