The Effect of Quality Improvement on Racial Disparities in Diabetes Care Thomas D. Sequist, MD MPH Alyce S. Adams, PhD Fang Zhang, MS Dennis RossRoss-Degnan, ScD John Z. Ayanian, MD MPP Background Gaps between evidence and quality exist for diabetes care Racial disparities in quality well documented Generic quality improvement is a potential solution to reduce disparities Division of General Medicine, Brigham and Women’s Hospital Department of Health Care Policy, Harvard Medical School Department of Ambulatory Care and Prevention, Harvard Medical School School Study Goals Assess baseline racial differences in diabetes care within a large multispecialty group practice Analyze impact of generic quality improvement efforts on existing racial disparities Methods - Study Population Adult patients ≥ 18 years with 24 months continuous enrollment in Harvard Pilgrim Health Care Harvard Vanguard Medical Associates Integrated multispecialty group practice – 14 health centers in Boston area – 250,000 adult patients Implemented electronic health record Computerized reminders to physicians Disease registries/ centralized outreach to patients Methods - Quality Measures Collected from electronic medical record Cholesterol management – Annual lipid testing – LDL control (< 130 mg/dL mg/dL)) – Statin dispensing (pharmacy claims) Glycemia management – Annual HbA1c testing – HbA1c control (< 7.0%) Rolling annual cohort – 1997 to 2001 – Diagnosis of diabetes for entire calendar year Generic QI efforts during 1997 to 2001 – – – Diabetes diagnosis – ≥ 1 inpatient diagnosis diabetes mellitus, or – ≥ 2 outpatient diagnoses diabetes mellitus, or – Dispensing of diabetes drug (insulin, oral agent) Methods – Study Site Annual retinopathy screening 1 Methods - Analysis Patient Characteristics Baseline (1997) racial differences in care – Multivariate logistic regression – GEE to account for clustering of patients – Adjusted for age, gender Black (n = 1,987) p value 60.2 53.8 <0.001 Male, % 51 41 <0.001 Long Term Enrollment*, % 74 73 0.44 Mean age, years Longitudinal changes in disparities White (n = 5,101) – Similar to baseline models – Data included for 1997 to 2001 – Race*year interaction term * Enrolled for at least 3 out of the 5 study years 100 LDL Cholesterol Control 100 Adjusted p<0.001 (race*year interaction) 90 80 63 70 60 50 40 43 30 20 10 46 51 65 62 53 35 White Black 40 29 % Acheiving LDL < 130 mg/dL % Receiving Annual LDL Testing Annual LDL Cholesterol Monitoring Adjusted p<0.001 (race*year interaction) 90 80 70 60 50 40 29 30 20 10 1997 1998 1999 2000 19 70 60 50 40 22 26 20 15 18 1997 1998 30 23 35 39 27 30 2000 2001 White Black % Receiving Annual HbA1c % Prescribed Statin 1999 2000 2001 100 Adjusted p=0.23 (race*year interaction) 80 10 1998 Annual HbA1c Monitoring 100 30 13 1997 2001 Statin Use 90 30 9 0 0 39 21 18 White Black 45 40 90 80 79 80 77 76 77 70 76 75 74 75 76 60 White Black 50 40 30 20 10 Adjusted p=0.11 (race*year interaction) 0 0 1999 1997 1998 1999 2000 2001 2 HbA1c Control Dilated Eye Exams 100 % Acheiving HbA1c < 7.0% 90 80 % Receiving Annual Eye Exam 100 Adjusted p=0.47 (race*year interaction) 70 60 50 40 34 35 37 24 26 28 34 36 27 26 White Black 30 20 10 90 80 75 71 68 69 71 66 63 65 66 70 60 71 50 40 White Black 30 20 10 Adjusted p=0.77 (race*year interaction) 0 0 1997 1998 1999 2000 1997 2001 Single multispecialty group practice with advanced EMR Unmeasured confounding No measures of patient experience with care 1999 2000 2001 Discussion Limitations 1998 Baseline disparities in diabetes care – Substantial disparity in low performing measures – No disparity in high performing measures Cholesterol management quality improvement – Reduction in process measure disparity – Less marked reduction in outcome measure disparity – Disparity in statin use persisted Glycemia management – – – No disparity in process measure No quality improvement in outcome measure Disparity in outcome measure persisted Implications Health care organizations can and should measure disparities in care Generic quality improvement may represent an effective tool to diminish disparities But…. Important to monitor outcomes measures and patterns of treatment Persistent disparities may require specific focus on minority health 3 Centralized Patient Mailings Electronic Reminders Changes in Cholesterol Management by Gender Annual LDL Testing by Center* 1997 2001 White Black White Black Δ in Disparity 1 36 20 63 60 13 2 39 28 68 63 6 3 42 30 68 65 9 4 50 38 68 63 6 5 38 30 64 60 4 HVMA Center * Among centers with at least 50 black patients LDL Control by Center* 1997 HVMA Center White 2001 Black White Black Δ in Disparity 1 15 6 41 42 10 2 13 12 52 43 (8) 3 19 9 50 41 1 4 19 6 46 33 0 5 14 10 44 41 3 * Among centers with at least 50 black patients 4