Weng W, Gray B, Lipner RS. — American Board of Internal Medicine
Presented: AcademyHealth Annual Research Meeting, June 2010
Objective: Using patient-level chart audits, we examined disparities in quality of diabetes care received by minority (African American, Hispanic) versus Caucasian patients from board certified internists. We parsed these differences between physicians and within physicians' patient panel.
Study Design: We defined disparities as differences in quality of care measures after controlling for patients' age and gender. We measured within versus between-physician disparity by comparing results from linear-probability versus fixed-effects models, adjusting standard error for clustering of patients within physicians. We then utilized Blinder-Oaxaca decomposition to identify proportion of between-physician disparity explained by physicians' cognitive skills (equated score on first attempt of internal medicine certification examination; certification in endocrinology subspecialty), a practice-infrastructure measure [NCQA's Physician Practice Connections (PPC)], and insurance revenue source (% of revenues from Medicare and Medicaid patients). Quality measures included: intermediate-outcomes (A1C, blood pressure and LDL good and poor controls), diabetes care process (A1C test, lipid profile, eye and foot exam, nephropathy assessment) and preventive care (influenza and pneumococcal vaccination, nutrition and physical activity plan).
Population Studied: In years 2008-2009, 774 internists completed American Board of Internal Medicine's practice characteristics survey and the diabetes practice improvement module with 17,896 chart audits and 774 physician surveys.
Principal Findings: Overall, disparities in quality of care across racial/ethnicity groups as measured by process of care were small (<1%) except for African American patients' pneumococcal vaccination (difference=-3.3%; SE=2.1%). However, a much different pattern emerged among intermediate-outcome measures for African American patients in that they had worse intermediate-outcome measures comparing to Caucasians with the largest disparity in blood pressure control (<130/80) (difference=-10.4%; SE=1.3%). Hispanic patients had worse outcomes in only the A1C measures relative to Caucasians (e.g., A1C>9%: difference=7.1%; SE=1.3%).
Most of the disparities in intermediate-outcome measures were due to variation within physicians' panel of patients (mean within variation/total variation=75.8%) rather than between physicians. For example, African American patients' LDL <100: within difference=-6.4% (SE=1.2%) versus between difference=-2.3% (SE=0.8%). On average, our physician level factors explained 74.4% of between variation; insurance revenue source explained 55.9%, cognitive skills explained 24.4% while practice-infrastructure explained only 0.2% of this between variation.
Conclusions: There were no significant disparities observed in process measures between or within physician patient panels. However, African Americans had significantly worse intermediate-outcome measures than Caucasians, while Hispanics had worse outcomes in A1C measures. Disparities in intermediate-outcome measures were largely due to variation within physicians' patient panel rather than between. For intermediate-outcomes, the dominant physician-level factor was insurance revenue source; practice-infrastructure explained little between-physician variation.
Implications for Policy, Practice or Delivery: Our findings suggest that interventions targeted at improving intermediate-outcome measures rather than processes of care are more likely to address racial/ethnic disparity. The fact that between-physician variation was small indicates that addressing disparity may require interventions targeted at meeting different needs of minority patients. Improving cognitive skills for board certified internists treating minority patients might somewhat reduce the between-physician variations that do exist. Conversely, our results indicate that policies designed to improve practice infrastructure as measured by the PPC would not materially affect racial/ethnic disparities.
For more information about this presentation, please contact Research@abim.org.