Quality Measurement and Gender Differences in Managed Care Populations with Chronic Diseases Ann F. Chou Carol Weisman Arlene Bierman Sarah Hudson Scholle Background • Institute of Medicine (IOM) outlined 6 attributes of health care: – Health care should be “safe, effective, patientcentered, timely, efficient, equitable.” • There are many reports of disparities in health care and health outcomes (IOM, Unequal Treatment). • Overall health care quality has improved over time but it has not improved for all patient subgroups. 2 © NCQA, 2007 Background: Disparities in Care for Cardiovascular Disease (CVD) • Substantial literature documents gender differences in guideline indicated services and treatment. • Women may need more aggressive risk factor management than men due to differences in risk factors and symptoms presentation. 3 © NCQA, 2007 Managed Care Population • A significant portion of the US population receives care through managed care organizations, where the quality of care may be more uniform. • In particular, there are few studies that examined disparities among commercially managed care enrollees. 4 © NCQA, 2007 Objective • To examine possible gender and racial disparities in meeting quality performance indicators for cholesterol screening and control among commercial health plans and in commercial and Medicare managed care populations of patients with chronic conditions 5 © NCQA, 2007 HEDIS® • Healthcare Effectiveness Data and Information Set (HEDIS®) is a set of standardized performance measures designed to ensure that purchasers and consumers have the information to reliably compare the performance of managed health care plans. 6 © NCQA, 2007 MEASURING QUALITY OF CARE FOR HEART DISEASE HEDIS Measures: – Comprehensive diabetes care • Screening for Cholesterol • Good control of cholesterol (LDL <100 mg/dL) – Cholesterol management after acute cardiovascular event • Cholesterol screening • Good control of cholesterol (LDL <100 mg/dL) 7 © NCQA, 2007 Study Population • Submissions from Commercial plans (2005) – 46 plan level submissions – 31 member level data submissions, including 11,813 patients – Geocoding and surname analysis used to estimate race, ethnicity and socio-economic status (SES) – Participating plans are larger, higher performing • Submissions from Medicare plans (2004) – 96,055 patients from 148 health plans – Race obtained from CMS enrollment data, geocoding used to estimate SES 8 © NCQA, 2007 Analytic Approach • Member Level – Hierarchical Generalized Linear Modeling (HGLM): HEDIS outcomes were modeled as functions of gender, controlling for other socio-demographic characteristics at the first level, and plan’s clustering effect at the second level – Adjusted rates were calculated • Plan Level – Descriptive statistics – Calculation of disparities score (male-female difference) – T-test to determine significance of the gender difference in performance rates 9 © NCQA, 2007 Results Gender and Racial/Ethnic Disparities in Medicare Managed Care Populations with Diabetes Mellitus 10 © NCQA, 2007 Sample Description Characteristic Medicare Population (%) Age: • <65 • 65+ 12.2 87.8 Gender: • Male • Female Race: 50.3 49.7 • White 86.6 • African American 13.4 11 © NCQA, 2007 HGLM Results: Odds Ratios Cholesterol Screening LDL control <100 mg/dL Female /Male 1.06 0.75* African American /White 0.67* 0.69* Enrolled in plans with >20% minority members 0.81 0.93 HEDIS Outcomes * Denotes statistical significance at p≤0.05 12 © NCQA, 2007 Adjusted Rates for Gender and Race 100 90 80 Cholesterol screening AA Female 70 Wht Female AA Men 60 Wht Men LDL Control <100 50 40 30 13 © NCQA, 2007 Results Gender Differences in Quality Measures Among Commercially Insured Patients 14 © NCQA, 2007 Sample Characteristics Characteristics Gender (women) % 44.2 Race/Ethnicity • African American 8.5 • Latino 7.7 • White/other 74.6 Low SES 10.3 Age •Age 65 and older 6.7 •45-64 years 72.6 15 © NCQA, 2007 Unadjusted Rates: Cholesterol Measures Characteristics Diabetes Management Screening LDL Control Cholesterol Management Screening LDL Control % t-test % t-test % t-test % t-test • Female 91.6 2.13* 37.7 6.04* 76.8 2.53* 45.3 6.22* • Male 92.7 Gender 43.3 79.9 55.0 African American • Female 90.6 • Male 87.7 -1.39 29.9 2.03* 36.1 65.7 -0.24 64.2 34.2 1.55 44.8 White • Female 91.5 • Male 93.2 2.84* 39.4 44.9 5.05* 77.6 80.8 2.32* 46.2 5.31* 55.4 * Denotes statistical significance at p≤0.05 16 © NCQA, 2007 HGLM Results: Odds Ratios Diabetes Management Covariate Cholesterol Management Cholesterol screening LDL Control Cholesterol screening LDL Control Female /Male 0.88 0.81* 0.88* 0.72* African American /White 0.69* 0.74* 0.88 0.71* Latino /White 1.13 0.73* 0.79* 0.87 Low SES /High SES 0.85 0.83* 0.77* 0.75* Age 65+ /(<65) 1.10 1.52* 1.03 1.26* * Denotes statistical significance at p≤0.05 17 © NCQA, 2007 WOMEN LESS LIKELY CONTROL CHOLESTEROL; AFRICAN-AMERICAN WOMEN EVEN LESS LIKELY Cholesterol Control <100 mg/dl for Commercial CVD sample 70 Unadjusted Rate (%) 60 White Males 9.2 50 White Females African-American Males 21.2 40 10.6 African-American Females 30 20 18 © NCQA, 2007 Results Gender Differences in Quality Measures Cross a Sample of Health Plans 19 © NCQA, 2007 Plans Characteristics Measures N (%) Medicare Plans (N=148) Commercial Plans (N=46) 53 (35.8) 12 (26.1) Staff/Group Model 63 (42.6) 2 (4.3) IPA/Network 79 (53.4) 25 (54.3) Mixed Model 6 (4.0) 19 (41.3) Midwest 40 (27.0) 13 (28.3) Northeast 38 (25.7) 11 (23.9) South 33 (22.3) 15 (21.7) West 37 (25.0) 7 (15.2) Not for Profit Health Plan Model Region 20 © NCQA, 2007 Overall Performance of Plans Measures Commercial Plans Medicare Plans Male Female Male Female Cholesterol Screeningdiabetes 92.9 91.7 90.9 91.8* Lipid Control – diabetes 44.4 38.8* 44.4 38.0* Cholesterol ScreeningCVD event 84.2 81.6 81.1 79.5* Lipid Control – CVD event 56.4 47.1* 52.1 43.6* 21 © NCQA, 2007 MORE THAN HALF OF HEALTH PLANS SHOW DISPARITY FAVORING MEN ON CHOLESTEROL CONTROL 70 60 COMMERCIAL PLANS (%) Striking disparities for cholesterol control 50 40 30 20 10 0 BP Control Cholesterol Screening (Diabetes) Cholesterol Control (<100) (Diabetes) Favors Women Cholesterol Screening (CVD) Cholesterol Control (<100) (CVD) Favors Men 22 © NCQA, 2007 MORE THAN HALF OF HEALTH PLANS SHOW DISPARITY FAVORING MEN ON CHOLESTEROL CONTROL 80 70 MEDICARE PLANS (%) Striking disparities for cholesterol control 60 50 40 30 20 10 0 BP Control Cholesterol Screening (Diabetes) Cholesterol Control (<100) (Diabetes) Favors Women Cholesterol Screening (CVD) Cholesterol Control (<100) (CVD) Favors Men 23 © NCQA, 2007 Conclusion • Clinically important and statistically significant gender differences and racial/ethnic disparities in LDL control measures in those with chronic conditions • African-American and poor women at greatest disadvantage from additive effect of race, gender, socioeconomic status • More than half of health plans have gender disparity of 5 percentage points or more for LDL control • Differences persist despite fairly equitable access to care. This is concerning given these are populations with high risk due to the presence of co-morbidities 24 © NCQA, 2007 Limitations • Lack of information on patient clinical, behavioral, and attitudinal characteristics. Results were consistent after adjusting for region, SES and plan clustering • No information on patterns of utilization and providers 25 © NCQA, 2007 Recommendations for Quality Improvement • Inform consumers about disparities and encourage active role in demanding high quality care • Educate physicians about the patterns of gender disparities in CVD care and encourage for active management with patients as partners • Engage employers, purchasers and health plans in promoting focus on disparities in quality improvement 26 © NCQA, 2007 Recommendations for Policy and Research • • • • Build consideration of gender, racial/ethnic, and socioeconomic differences into quality improvement interventions, with incentives for health plans and providers to identify, address and monitor potential disparities in CVD care Gain consensus on the methods for collecting, examining, and reporting on key population-wide health measures by gender, racial/ethnic, and socioeconomic characteristics Improve training in the area of women’s health for the health care workforce Fund further research to disentangle influence of physician and patient factors on outcomes 27 © NCQA, 2007 Acknowledgements • Funded by the Agency for Healthcare Research and Quality, American Heart Association, the California Endowment • NCQA staff for support in project and data management 28 © NCQA, 2007