3051 — Methods for measuring racial differences in hospital outcomes attributable to disparities in use of high-quality hospital care

Hebert PL, VA Puget Sound Health Care System; Wong ES, VA Puget Sound Health Care System; Hernandez SE, VA Puget Sound Health Care System; Rinne ST, VA Connecticut Healthcare System; Sulc CA, VA Puget Sound Health Care System; Neely EL, VA Puget Sound Health Care System; Liu CF, VA Puget Sound Health Care System;

Objectives:
To test a new approach to measuring racial/ethnic disparities in the use of high-quality hospitals.

Methods:
To study disparities in the use of high-quality hospital care, researchers typically identify hospitals that are "minority-serving" and report differences in risk-adjusted outcomes between minority-serving and non-minority-serving hospitals. There are problems with this approach. First, it requires making ad hoc decisions on the definition of a minority-serving hospital. Second, disparities in outcomes between minority-serving and non-minority-serving hospitals are meaningful statistics only if a large percentage of the minority population receives care from minority-serving facilities. Third, these methods estimate disparities due to "between-hospital" differences in care quality; disparities in outcomes can also materialize if minority patients received poorer quality of care than whites at all hospitals (a "within-hospital" disparity). We proposed a new approach that estimated the reduction in outcomes if minority patients switched hospitals with white patients. Through simulations, we compared the new approach to the commonly-used "minority-serving" approach. We evaluated each method's ability to detect and measure a between-hospital disparity when the simulation imposed one. We also conducted an empirical demonstration of readmissions for heart failure in the VA or Medicare, 2006-2009.

Results:
In simulations, the commonly-used and new approaches correctly identified between-hospital disparities when they existed, and rejected them when racial differences in readmissions were caused by within-hospital disparities. The new approach estimated the magnitude of the disparity accurately; the commonly-used approach underestimated it by 36%. In the empirical analysis of 32,956 Veterans with an index hospitalization for heart failure at 125 VA hospitals, there were 186.2 (95% CI, 107.1 to 265.4) excess readmissions for African American veterans. Using the new approach we estimated that only 31 (95% CI, -3 to 65) of these could be attributed to between-hospital disparities in quality of care.

Implications:
The new methods required fewer ad hoc assumptions, gave results that were pertinent to all minorities in a community, not just those at "minority-serving" facilities, and gave more accurate estimates of the magnitude of between-hospital disparities.

Impacts:
Researchers should consider using the new approach for measuring disparities in use of high-quality care by vulnerable populations.