Effect of CostCost-Sharing on Screening Mammography in Medicare Managed Care Plans Background Amal Trivedi, MD, MPH William Rakowski, PhD John Ayanian, MD, MPP CostCost-sharing affects nearly every American with health insurance Copayments reduce moral hazard to “overconsume” overconsume” health care with full insurance May also reduce use of appropriate preventive care Several wellwell-known studies of costcost-sharing have excluded the elderly 2007 AcademyHealth Annual Research Meeting Objectives To determine the prevalence of mammography costcost-sharing in Medicare health plans To understand the impact of costcost-sharing on the appropriate use of mammography To assess effects of mammography costcost-sharing on vulnerable population groups Sources of Data 20012001-4 Medicare HEDIS data 20012001-4 Medicare enrollment file 20012001-4 CMS health plan benefit data 2004 Interstudy Competitive Edge database Study Population Included women age 6565-69 who had been assessed for the HEDIS breast cancer screening indicator Excluded women who died during the measurement year Final study population: 366,475 women in 174 health plans Variables Independent variable: enrollment in a plan with >$10 or >10% coinsurance for screening mammography Dependent variable: receipt of a biennial screening mammogram Covariates: IndividualIndividual-level: race (black, white, other), zip code level income and education, buybuy-in eligibility, year PlanPlan-level: census region, model type, plan size, plan age, taxtax- status 1 Analyses - Main Sample Analyses - Subsample Assessed characteristics of enrollees and breast cancer screening rates in costcost-sharing and fullfullcoverage plans Constructed multivariate regression models adjusting for individual and plan covariates and clustering by plan Included interactions of costcost-sharing with income, education, race and buybuy-in eligibility Identified seven plans that changed from full coverage to costcost-sharing Performed a differencedifference-inin-difference analysis by comparing longitudinal changes in screening compared to 14 matched control plans Plans matched based on: Trends in Mammography CostCost-sharing Year # of plans (N=174) % of women in costcost-sharing plans 2001 3 0.5 2002 10 4.3 2003 11 4.8 2004 21 11.4 Census region ProfitProfit-status Model type Characteristics of Enrollees in CostCost-Sharing and Full Coverage Plans Age (y) Black (%) Below Poverty (%) College Attendance (%) BuyBuy-in (%) CostCost-Sharing 67.1 23 11 32 10 Full Coverage 67.1 10 9 35 7 Median copayment $20 (Range $13-$35) 5 plans charged 20% coinsurance Screening Rate (%) Breast Cancer Screening Rates in CostCostSharing and FullFull-coverage Plans 100 90 80 70 60 50 40 30 20 10 0 76.7 65.7 2002 Effect of CostCostsharing 78.2 77.5 70 69.7 2003 Cost-sharing Adjusted Impact of CostCost-sharing on Screening Mammography Full coverage 2004 Unadjusted -8.3% Adjusted for SES, plan characteristics, and clustering by plan -7.2% 95% CI/p -9.7%, -4.6% P<0.001 * Cost-sharing had the largest effect on mammography of any of the plan covariates in the model 2 Mammography Rates in plans that instituted costcostsharing compared to matched controls Adjusted Effect of CostCost-sharing by Income and Education 0.0% 2002 2004 Change ∆-∆ -2.0% Adjusted ∆-∆ -4.0% -4.7% -6.0% -8.0% -8.3% -5.5% Low Medium High -11.8% Education 74.8% Maintained full coverage 71.9% -7.5% -7.7% -10.0% -12.0% Added costcostsharing Income P<0.001 for trends Mammography Rates in 7 Plans that Instituted CostCostsharing in 2003 Compared to 14 Matched Control Plans 2002 2004 Change Added costcostsharing 74.8% 69.3% -5.5% Maintained full coverage 71.9% 75.3% +3.4% Mammography Rates in 7 Plans that Instituted CostCostsharing in 2003 Compared to 14 Matched Control Plans 2002 2004 Change ∆-∆ Added costcostsharing 74.8% 69.3% -5.5% Maintained full coverage 71.9% 75.3% +3.4% Mammography Rates in 7 Plans that Instituted CostCostsharing in 2003 Compared to 14 Matched Control Plans 2002 2004 Change ∆-∆ 74.8% 69.3% -5.5% -8.9% -8.8% Adj. ∆-∆ Maintained full coverage (-4.0, -13.6%) P=0.002 71.9% 75.3% +3.4% -8.9% Limitations Added costcostsharing Adj. ∆-∆ Women not randomly assigned to costcost-sharing plans Unable to analyze differential impacts of specific copayment amounts Lacked information on rescreening ZipZip-code proxies for income and education Limited to Medicare managed care 3 Conclusions Copayments of >$10 or coinsurance of >10% associated with lower rates of breast cancer screening CostCost-sharing disproportionately affects vulnerable populations Prevalence of costcost-sharing is dramatically increasing in Medicare managed care Implications CostCost-sharing should be tailored to the underlying value of the health service Eliminating copayments may increase adherence to appropriate preventive care Important implications for Medicare FFS, where enrollees without supplemental coverage face 20% coinsurance Moral Hazard “If your office gives you and your coco-workers all the free Pepsi you want— want—if your employer, in effect, offers universal Pepsi insurance— insurance—you’ you’ll drink more Pepsi than you would have otherwise.” otherwise.” - Malcolm Gladwell “The Myth of Moral Hazard” Hazard”, in The New Yorker, August 29, 2005 Study Population Health Plan Characteristics of Enrollees in CostCost-Sharing and Full Coverage Plans 24,468,528 observations in HEDIS data from 2001-4 Matched to Enrollment File (97% match rate) 23,656,038 observations Excluded males, enrollees not between 65-69, persons who died in measurement year, and observations from 2001 2,189,983 observations from 178 plans from 2002-4 Linked to Interstudy database; excluded four plans where Interstudy data not available 2,143,556 observations from 174 plans CostCost-sharing Full coverage South 36% 16% West 21% 41% Staff/Group Model ForFor-profit 7% 18% 72% 56% Excluded enrollees who were not assessed for HEDIS mammography measure 550,082 observations (366,475 enrollees) 4 Adjusted Negative Effect of CostCost-sharing by Race and Medicaid Eligibility 0.0% -2.0% Yes No White Black -3.7% -4.0% -6.0% -6.8% -7.8% -8.0% -10.0% -12.0% -14.0% -13.4% Medicaid Race P<0.001 for trends 5