Medicare Prescription Drug Program and Using Part D Data for Research Becky Briesacher, PhD Associate Professor, Medicine Division of Geriatric Medicine 1

Funding and COI Supported by grants R01AG and R01AG from the National Institute on Aging (NIA), and the Harvard Pilgrim Health Care Foundation. Dr Briesacher is also supported by a Research Scientist Development Award from the NIA (K01AG I declare no conflict of interest. 2

3 "75% of older Americans lack decent, dependable, private-sector coverage of prescription drugs. That’s three out of every four seniors. To those who think prescription drug coverage isn’t a problem for most Medicare beneficiaries, I say, think again." President Bill Clinton July 1999

Career in Part D Research 1) identified high predictability of prescription drug expenditures from one year to the next 2 ; 2) demonstrated selection into prescription drug coverage is also predictable, based almost exclusively on observable health status 3 ; 3) created elasticity estimates that reducing medication costs by 10% would increase drug utilization by 5.4%- 6.6%; 3 and 4) established Part D effects differ in nursing home setting.

Overview of Presentation Brief overview of Medicare Part D Brief overview of Part D data Description of Medicare Part D evaluation 5

6 History of Drug Coverage in Medicare Medicare is health care insurance for elderly and disabled Americans since 1965 Medicare had 2 notable coverage gaps: LTC and prescription drugs –Alternatives include employer/retiree benefits, self-purchased Medigap plans, Medicare HMO option, Medicaid for the poor, Military & VA Medicare Catastrophic Coverage Act of 1988 – first prescription drug program in Medicare, repealed before implemented. Part D passed in 2003 and implemented in 2006 ~65%-80% of population had drug coverage in 2005

7 Medicare Part D Drug Benefit Voluntary enrollment unless in Medicaid –Choose from dozens (~40) of private Rx coverage plans and Medicare Advantage organizations –premiums are heavily subsidized, late penalty for late enrollment Auto-enrolled into Part D if in Medicaid

NOTE: Percentages do not sum to 100% due to rounding. 1 Includes Veterans Affairs, retiree coverage without RDS, Indian Health Service, state pharmacy assistance programs, employer plans for active workers, Medigap, multiple sources, and other sources. 2 Includes Retiree Drug Subsidy (RDS) coverage and FEHBP and TRICARE retiree coverage. SOURCE: Centers for Medicare & Medicaid Services, 2009 Enrollment Information (as of February 1, 2009). Total Number of Medicare Beneficiaries = 45.2 Million Total in Part D Plans: 26.7 Million (59%) Medicare Advantage Drug Plan Retiree Drug Coverage 2 No Drug Coverage Other Drug Coverage million 10% 7.9 million 18% 6.2 million 14% Prescription Drug Coverage Among Medicare Beneficiaries, million 20% Stand-Alone Prescription Drug Plan 17.5 million 39%

Medicare’s “Standard” Drug Benefit in 2009 … Most plans do not offer the “standard” benefit, and coverage varies across most dimensions, including: –Monthly premiums –Deductibles –The “doughnut hole” –Covered drugs and utilization management restrictions –Cost sharing for covered drugs $295 Deductible $2,700 in Total Drug Costs $3,454 Coverage Gap (“Doughnut Hole”) Plan Pays 75% Plan Pays 15%; Medicare Pays 80% $6,154 in Total Drug Costs ($4,350 out-of-pocket) Enrollee Pays 25% 5%

Part D data for research Since June 2008, Part D data are available to researchers. Chronic Condition Data Warehouse (CCW) contains 100% Part D data and is official data source. –CCW offers chronic disease indicators (21 conditions) to ID disease cohorts but has data on all Part D enrollees Researchers may request random 10% or 20% sample Part D data are linkable to other Medicare data 10

Part D data Part D data include information about drug (NDC), cost, pharmacy, provider and benefit information Data are de-identified to protect identity of beneficiary, prescriber, pharmacy and plan. Researchers must select each variable and provide variable-level justification Assistance offered through ResDAC 11

12

Part D data Constructed variables “may not exactly represent the beneficiary experience at the time of the prescription fill.” Part D data differ from pharmacy claims –Contains only final status record Will not include drugs excluded from Part coverage or filled through 3 rd party, or not filed as claim (e.g., 100% cash). 2-year lag in availability E.g., OPTIMIZING CHRONIC DISEASE PREVENTION AND MANAGEMENT IN ADVANCED DEMENTIA R21HS : PI Tjia –$20,000, 9 month lag = Part D data linked to Part A, MDS, and OSCAR on 200,000 Medicare enrollees with end-stage dementia in NHs. 13

NOTE: Percentages do not sum to 100% due to rounding. 1 Includes Veterans Affairs, retiree coverage without RDS, Indian Health Service, state pharmacy assistance programs, employer plans for active workers, Medigap, multiple sources, and other sources. 2 Includes Retiree Drug Subsidy (RDS) coverage and FEHBP and TRICARE retiree coverage. SOURCE: Centers for Medicare & Medicaid Services, 2009 Enrollment Information (as of February 1, 2009). Total Number of Medicare Beneficiaries = 45.2 Million, 2009 Total in Part D Plans: 26.7 Million (59%) Medicare Advantage Drug Plan Retiree Drug Coverage 2 No Drug Coverage Other Drug Coverage million 10% 7.9 million 18% 6.2 million 14% Part D Data available only on Part D enrollees 9.2 million 20% Stand-Alone Prescription Drug Plan 17.5 million 39%

15 Medicare Part D and Changes in Prescription Drug Use and Cost Burden: National Estimates for the Medicare Population, Becky A. Briesacher, PhD (2) Yanfang Zhao, MA (4) Jeanne Madden, PhD (1) Fang Zhang, PhD (1) Alyce S. Adams, PhD (3) Jennifer Tjia, MD, MSCE (2) Dennis Ross-Degnan, ScD (1) Jerry H. Gurwitz, MD (2) Stephen B. Soumerai, ScD (1) (1) Harvard Medical School and Harvard Pilgrim Health Care Institute; (2) Div of Geriatric Medicine, Meyers Primary Care Institute and UMass Medical School; (3) Kaiser Permanente Division of Research; (4) Duke Translational Nursing Institute, Duke University School of Nursing

Evaluations of Medicare Part D Previous evaluations used –Pharmacy chain data –Regional Medicare Advantage plan data Important gap in research on Part D to date: –Lack of nationally-representative evaluations 16

Study Design Overall Methods Examined study outcomes over 8 years ( ) Compared actual post-Part D ( ) outcomes to projected values (using data) Projections based on time-series regressions and parametric bootstrapping techniques and 10,000 simulations of post-Part D outcomes. Standardized estimates by fixing population characteristics Subgroup analyses -- by demographic & health status 3-year continuous cohort for sensitivity analyses 17

18 Study Design Data from MCBS CMS conducts the Medicare Current Beneficiary Survey –Since 1991, continuous face-to-face panel survey –3-year rotating panels –>12,000 community-dwelling and institutionalized, elderly and disabled per year Rich variety of measures: –Demographics, income, assets, living arrangements, family supports, health status and functioning –Health insurance and drug coverage –Utilization of services – self-reported and FFS claims –Costs, copayments, deductibles –Access to medical care, satisfaction –Medications self-reported and Part D data, beginning in 2006

Study Design Sample and Variables Community-dwelling Medicare enrollees, –~11.5K persons/year; total unique individuals, n = 38,777 Study Variables: –Prescription fills and OOP costs from MCBS Cost and Use file self-reported fills (disregarding Part D claims) –Prescription drug coverage categorized by coverage status between 2005 and 2007 –no drug coverage in all three years –gained drug coverage after Part D (in 2006 or 2007) –had drug coverage in all three years 19

20 Annual Drug Use and Out-of-Pocket Drug Costs unadjusted means for overall population, Prescription Drug FillsOOP Drug Costs Medical Care. 49(9):834-41, 2011 Sep.

21 Changes in Annual Rx Fills and OOP Drug Costs Following Part D Implementation (overall population) Means (95% Confidence Intervals) Difference between Standardized Observed and Predicted Outcome measure N Observed Actual Observed Mean Standardized Observed* Mean Standardized Predicted Mean P value 2006Rx fills34, (30.3 to 31.7) 1.8 (1.1 to 2.5) <.001 OOP drug costs 34, (578.1 to 657.4) ( to ) < Rx fills38, (30.9 to 32.3) 3.4 (2.7 to 4.1) <.001 OOP drug costs 38, (590.0 to 657.1) ( to ) <.001 Standardization to hold 2000 population characteristics stable Standardization to hold 2000 population characteristics stable Predictions used autoregressive models based on historical trends Predictions used autoregressive models based on historical trends Confidence intervals were constructed by creating 10,000 simulated outcomes using bootstrap method Confidence intervals were constructed by creating 10,000 simulated outcomes using bootstrap method Medical Care. 49(9):834-41, 2011 Sep.

22 Subgroup Analyses: Annual Prescription Drug Fills absolute differences between observed and predicted means Medical Care. 49(9):834-41, 2011 Sep.

23 Subgroup Analyses: Annual OOP Drug Costs absolute differences between observed and predicted means Medical Care. 49(9):834-41, 2011 Sep.

24 Changes in Annual Fills and OOP Rx Costs among 3-year Cohorts, by Insurance Status, Gained drug coverage after Part D Always had drug coverage Mean Rx FillsMean Rx OOP Costs Medical Care. 49(9):834-41, 2011 Sep.

My Current Part D Research NIH-National Institute on Aging(R01AG028745) (Co-Investigator) Changes in Cardiovascular Care and Outcomes in Eight Years after Medicare Part D   Developing R01 to conduct clinical trial on Part D enrollment protocols in nursing home setting 25