Weng W, Gray B, Lipner RS. — American Board of Internal Medicine
Presented: American Public Health Association Conference, November 2009
Objective: Using patient-level medical records, we examined differences in quality-of-care received by African-American/Caucasian (non-Hispanic) patients. We also measured the share of within versus across physician variation of these differences.
Methodology: Our data is drawn from a 13-state sample of 236 general internists. These data included 22,526 medical record audits on seven chronic care conditions and six preventive services measures. Our quality-of-care measures had three intermediate-outcome measures, 13 chronic processes-of-care and six prevention measures.
We measured within versus across physician variation in quality by comparing results from a linear-probability versus fixed-effects model, adjusting standard error for clustering of patients within physicians. We also accounted for patient characteristics (age, gender, comorbidity) and ZIP code level SES (median income, education).
Results: Controlling for patient characteristics, African-American patients had significant poorer controls in two out of three intermediate-outcomes measures (p <.05) - blood pressure control (<140/90) for cardiovascular disease, diabetes, and hypertension patients (-11%) and hemoglobin A1c control (< 7.0%) for diabetes patients (-11%). LDL control (<130) for CAD and diabetes patients had no significant racial differences. Only one of 13 processes measures – warfarin therapy for atrial fibrillation patients (-18%) – and two of six prevention measures – pneumococcal vaccination (-8%) and osteoporosis screening for eligible (-20%) – showed significant disparities.
Our fixed-effects model estimates indicated that most of the African-American/Caucasian differences for outcome measures were the result of within versus across physician variation in quality-of-care (for blood pressure control: -7% within versus -4% across, for hemoglobin A1c control almost all was within variation). For prevention measures, the within variation of pneumococcal vaccination was only half of the across (-3%/-6%) and the within variation of osteoporosis screening was two-thirds of the across (-8%/-12%).
Conclusion: Overall, our findings suggest that, both within and across physicians, African-American patients received similar quality of care as measured by processes-of-care and preventive services. The disparities that existed in a few measures were mainly due to a bigger proportion of African-American patients receiving care from physicians providing lower quality-of-care overall. However, African-American patients had significantly worse outcomes than Caucasian in two of three intermediate-outcome measures. For example, our estimates indicate there would be 32% increase in the likelihood if African-American patients had the same blood pressure control as otherwise similar whites. These differences were largely due to variation within physicians rather than across. This suggests that all physicians have the opportunity to improve outcomes among African-American patients.
For more information about this presentation, please contact Research@abim.org.