Developing a Health Plan Equity Report Harvard Pilgrim s approach Kathryn Coltin Harvard Pilgrim Health Care Diversity Rx Session 1570 October 20, 2010
Harvard Pilgrim Health Care Not-for-profit health plan, based in Wellesley, MA 1100 employees across 7 locations Over one million members in MA, NH, ME, RI Mission to improve the health of the people we serve and the health of society Rated #1 health plan in the U.S. in both quality of care and member satisfaction since 2005 But Racial/ethnic disparities in several quality measures
Developing Harvard Pilgrim s Equity Report: The building blocks 8. Trend performance over time 7. Plan and implement interventions 6. Report on comparative measure rates to identify disparities 5. Define an actionable disparity 4. Analyze population distribution and define reporting categories 3. Select measures to be reported 2. Assure availability of relevant data for population served 1. Determine scope of disparities to be reported (e.g. race, ethnicity, language, gender, income, education, etc.)
2. Assure availability of Race/Ethnicity and Language data Time of collection Context Data source/ type of data Start date Use experience Enrollment Eligibility determination Paper form & EDI Language only mid-1990s Only ~10% of subscribers provide Primary Language Web-based R/E and Language October 2003 Subscribers only ~16% of subscribers provide REaL data Postenrollment Clinical Computerized phone outreach (IVR) Race/ethnicity only June 2007 >95% of members asked provide race/ethnicity data Health Risk Appraisal R/E and Language July 2008 <10% of members complete HRA; most provide REaL data Care Mgmt/Disease Mgmt System R/E and Language July 2010 New source; fields are not forced (yet) Non-clinical Secure Member Web Portal Pop-up Survey Race, Hispanic Indicator, detailed Ethnicity, Written & Spoken Lang. December 2008 <25% of members are registered portal users; most use it infrequently
3. Select Measures: Harvard Pilgrim s Evolving Portfolio Annual since 2003 Preventive Screenings Chlamydia screening Cancer screening Breast CA Cervical CA Colorectal CA Chronic Disease Care Asthma meds 5-17 year olds 18-56 year olds Diabetes care HbA1c testing LDL-C testing Retinal screening Nephropathy monitoring CAHPS measures of access & customer service Note: Italicized measures are outcome measures Added in 2006 Chronic Disease Care Cardiovascular disease Persistent use of betablocker after AMI LDL-C testing in CAD LDL-C control in CAD BP control in patients with HTN Monitoring patients on Persistent Medications Diabetes HbA1c >9 (poor control) HbA1c <7 (good control) LDL-C <100 (good control) Rheumatoid Arthritis (DMARDs) Acute Care Inappropriate antibiotic use for adult bronchitis Imaging for low back pain in adults Added in 2007 Preventive Care/Access Well Visits -Infants 0-15 mo. - Children 3-6 yr. - Adolescents 12-21yr. Chronic Care Diabetes BP control Acute Care Strep Tx prior to antibiotic Rx for children w/ Pharyngitis Appropriate antibiotic use for children w/uri Added in 2010 Patients care experiences Medical Home
4. Analyze member distributions: Defining reporting categories Gender Race/ethnicity Use geocoding and surname coding when self-reported data are not available Agreement on performance measure rates at the population level is very high Target interventions at individual level to members with >90% probability of a given race/ethnicity Current categories: Asian/PI, non-hispanic Black, non-hispanic White, Hispanic, Other/Unknown Education High Education: >50% of residents had a 4 yr college degree Low Education: >25% of residents had less than a high school degree Household income <$50K, $50-75K, $75-100K, >$100K Health status Excellent, Very Good, Good, Fair, Poor Within above categories: Finer age groupings as needed (e.g. adolescent well visits by 12-17 & 18-21 yr olds) Contracted provider entities (integrated delivery systems, large medical groups) Product lines (HMO/POS, PPO, deductible products) States, Counties, Zip code groupings
5. Define an actionable disparity: Our definition Harvard Pilgrim currently defines a disparity as a performance rate for a given population group that is >6 percentage points below that of the group with the best performance rate (i.e., the benchmark population) Why? This definition works across all types of disparities that we measure Overall rates for most measures are above the national 90 th percentile rate Comparison with the benchmark population is consistent with our goal of assuring the highest quality care, not just equal care For racial/ethnic disparities, the white non-hispanic population is frequently not the benchmark population The margin of error on many measures is +/- 5% or higher For preventive care measures having very large denominators, very small differences (1-2%) are statistically significant, but not clinically significant For acute illness and chronic disease measures having small denominators, large differences (>6 percentage points) are often not statistically significant, but can be clinically important
6. Report the Measures in an Equity Report Analyzing disparities Measures for current year performance (or two year performance for measures with small Ns) are usually displayed using bars for each reporting category within a measure. Separate graphs are used to display performance for each attribute (race, ethnicity, gender, education, income, etc.). Performance Rate 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% HEDIS Rates for Comprehensive Diabetes Care by Indirectly Estimated Race/Ethnicity HbA1c Rate Eye Exam Rate LDL Tx Rate Black 91.0% 62.8% 92.3% 70.8% Hispanic 88.7% 54.1% 88.4% 58.0% Asian 91.6% 63.1% 92.0% 62.5% White/other 88.5% 60.8% 90.9% 58.9% HEDIS Measure Nephropathy Monitoring Rate Colorectal Cancer Screening Rates by Race/Ethnicity 2003-2009 Measures with data for multiple years are trended on separate line graphs showing each group that had an actionable disparity when compared with the benchmark group Percent Screened 81% 79% 77% 75% 73% 71% 69% 67% 65% 63% 61% 59% 57% 55% 2003 2004 2005 2006 2007 2008 2009 Performance Year Black Hispanic Linear (Black) Linear (Hispanic)
6. Report the Measures in an Equity Report Stratify Measures by Reporting Category 2004 HEDIS Diabetes Care Rates by Indirectly Estimated Race/Ethnicity 100% HEDIS Rate 90% 80% 70% 60% 50% 40% HbA1c Rate Eye Exam Rate LDL Tx Rate Black 91.0% 62.8% 92.3% Hispanic 88.7% 54.1% 88.4% Asian 91.6% 63.1% 92.0% White/other 88.5% 60.8% 90.9% HEDIS Measure
7. Plan and implement interventions: Selecting and targeting disparities Prioritize disparities for intervention Health impact of disparity Size of disparity Persistence of disparity Choose an intervention strategy based on data analysis Community-based intervention Provider intervention Member intervention Some combination Use data to design and target interventions Provider and community interventions and design of member interventions based on Individually targeted interventions based on REaL data or probabilities >90%
Harvard Pilgrim Health Care Examples of interventions to reduce disparities Diabetic Eye Exams Community-based intervention: free eye exams and counseling at local supermarket/pharmacy in two targeted communities Physician-focused intervention: feedback reports on diabetes care performance by race/ethnicity to large multi-site group practice; quality grants to selected groups to focus on improving eye exam rates Member-focused interventions: eliminated requirement for a PCP referral for a diabetic retinal exam; sent coupons to diabetic members residing in targeted communities to waive the visit co-pay fee for an eye exam Colorectal Cancer Screening Developed an Interactive Voice Response script in English and Spanish that queried members as to whether they had been screened, counseled them about CRC screening, requested REaL data and provided culturally tailored messaging re importance of screening Implemented IVR outreach; offered supplemental written information in English or Spanish Well Baby Visits in 1 st 15 months of life Became a sponsor and promoter of Text4Baby in Greater Boston; broad promotion on our website and in member newsletter with targeted mailing to new black and Hispanic moms Published article in provider newsletter re preventive care guidelines, P4P incentives for well baby visits and proper documentation/coding of well visits
Diabetic Eye Exam Distribution of Missed Opportunities among Probable Hispanic Members by Community in 2004 35 80% 30 25 20 15 10 5 Dorchester Hyde Park/Readville Jamaica Plain 0 New Bedford Roxbury Lowell Chelsea Roslindale Lawrence Framingham Marlborough Randolph Revere/Revere Beach Taunton Brockton Brockton East Boston Framingham Lynn West Roxbury Brighton E. Watertown Everett Fall River Malden Medford Nashua Peabody Quincy Somerville Somerville/Winter Hill Boston Cambridge Fitchburg Stoughton West Lynn Salem Community 70% 60% 50% 40% 30% 20% 10% 0% # Missed Opportunities Cumulative % of Missed Opportunities Defect Freq Cum Defect Pct
35 90% 30 80% 70% 25 60% 20 50% 15 40% 10 30% 20% 5 10% 0 0% HVMA Kenmore BIDPO Boston Univ Med Ctr HVMA West Roxbury HVMA Copley MGHPO HVMA Watertown New Bedford Medical Assoc BWHPO Caritas Christi Netwk Svcs HVMA PO Square HVMA Cambridge Lawrence General IPA HVMA Chelmsford So NE Health Alliance BIDPO-FSEN Caritas Good Sam IPA Southcoast Physician Services Charles River Med Assoc Metrowest Framingham HVMA Medford Highland Healthcare Assoc Lowell General PHO HVMA Braintree SE Mass Physician Grp Primacare IPA New Bedford Medical Assoc-Wareham Cambridge Integ. Solutions-Bridgewater Cambridge Integ. Solutions-Goddard Pk UMass Memorial E. Boston Neighborhood Health Ctr St Elizabeth's Health Professionals Dartmouth-Hitchcock-Nashua No Shore Health System (AKA Essex) HVMA Peabody Lahey Burlington-FSEN Health Alliance with Physicians Inc Hawthorn Medical Assoc HVMA Quincy HVMA Wellesley Newell Medical Delivery Org HVMA Somerville Group # Missed Opportunities Cumulative % of Missed Opportunities Diabetic Eye Exam Distribution of Missed Opportunities among Probable Hispanic Members by Physician Group in 2004 Defect Freq Defect Cum Pct
8. Trend performance over time: Colorectal Cancer Screening Rates 2003-2005 80.0% 75.0% Best Rate HEDIS 2006 National 90 th Percentile (63.5%) Percent Screened 70.0% 65.0% 60.0% 55.0% Disparity 50.0% Black Hispanic Asian White/other ALL Members Performance Year 2003 69.4% 60.7% 66.1% 66.9% 66.2% 2004 71.1% 63.1% 65.8% 67.0% 68.3% 2005 71.2% 65.4% 66.2% 68.5% 69.6% Indirectly Estimated Race/Ethnicity *Racial/ethnic group assigned using geocoding and surname analysis Screening rates increased and disparity decreased (<6 percentage points in 2005) 2005 Rate in lowest performing group was above the HEDIS national 90 th percentile. 14
8. Trend performance over time: Consider gap & performance rate related to intervention Percent Screened 83% 79% 75% 71% 67% 63% 59% 55% 69.4% 60.7% Colorectal Cancer Screening Rates by Race/Ethnicity 2003-2009 IVR IVR + Spanish P4P Gap = 8.7 Gap = 3.8 Gap = 8.1 2003 2004 2005 2006 2007 2008 2009 Performance Year Black Hispanic Linear (Black) Linear (Hispanic) 76.4% 68.3%