Medicare & Medicare Prescription Drugs Call for Papers Medicare & Medicare Prescription Drugs: Expense or Investment? Chair: Len Nichols, Ph.D. Sunday, June 6 • 9:30 a.m.-11:00 a.m. • Identifying High Risk Medicare Enrollees, Improved Identification and Payment Possible? Jonathan Weiner, Dr. P.H., Chad Abrams, M.A., Richard Lieberman Presented by: Chad Abrams, M.A., Systems Analyst/Project Manager, Health Policy and Management, Johns Hopkins University, 624 N. Broadway, #600, Baltimore, MD 20215; Tel: 410.614.3957; Fax: 410.955.0470; E-mail: cabrams@jhsph.edu Research Objective: The objective of this project was the development of a prospective payment model for Medicare based upon the ACG Case-Mix System, an alternative to the prevailing disease-specific approaches. The intent was to provide adequate compensation for treating sick individuals while at the same time minimizing incentives for targeting healthier than average enrollees. Study Design: This project augments previous work aimed at developing ACG-based payment models for Medicare with the inclusion of new indicators. The project tested a revised and expanded “hosdom” marker, which indicates a high probability of a future admission, and introduced a new diagnosis-based measure, “frail symptoms,” which reflects diagnoses associated with high risk or frail beneficiaries. The “hosdom” marker was expanded by incorporating additional years of data and underwent substantial clinical review for completeness and face-validity. These two variables, along with Johns Hopkins Expanded Diagnosis Clusters (EDCs), a taxonomy focusing on common diseases/conditions which have significant impact on resource use which were used to juxtapose the impact of incorporating disease-specific measures, were combined in various means to assess their impact on explanatory power and predictive ratios of potential prospective payment models. Population Studied: CMS provided data extracted from the 1996 and 1997 five percent Medicare Standard Analytical Files consisting of nearly 1.4 million aged and disabled Medicare beneficiaries meeting selected enrollment criteria. Diagnoses streams from 1996 were used to assign risk assessment variables which were then used to predict actual expenditures recorded in calendar year 1997. Principal Findings: The hosdom and frail symptoms variables were found to be extremely powerful markers for identifying individuals at risk for future high expenditures. When incorporated in a prospective payment model, these variables led to increased r-squared while simultaneously reducing the sum of errors and generating predictive ratios closer to one for all cost quintiles. In contrast, the impact of introducing disease-specific conditions was mixed. While the r-squared improved, there was a simultaneous increase in total dollars mis-allocated and predictive ratios based on year-1 cost quintiles moved farther away from (rather than closer to) 1.0 for the lower quintiles. Conclusions: The two new indicators show promise as predictors for high future resource use. The frail marker may also serve as a potential proxy for ADL-type measures typically drawn from survey data. Higher r-square values do not necessarily equate with more accurate payment models. Models with both high explanatory power and predictive ratios close to 1.0 across the year-1 cost quintiles may be achieved when disease markers are combined with more comprehensive measures. Implications for Policy, Delivery or Practice: More accurate payment models can be developed if attention is given to more comprehensive risk assessment variables as an adjunct to the current linearly additive disease specific approach, especially for frail and vulnerable sub-groups of the population. Primary Funding Source: CMS • Estimating the Value of Investment: Medicare and the Overall U.S. Health Care Service Bryan Luce, Ph.D., M.B.A., Frank Sloan, Ph.D., Josephine Mauskopf, Ph.D., Clark Paramore, M.S.P.H., Manishi Prasad, M.P.H. Presented by: Bryan Luce, Ph.D., M.B.A., Senior Research Leader; Chairman of Board, MEDTAP International, 7101 Wisconsin Avenue, Bethesda, MD 20814; Tel: 301.664.7274; Fax: 301.654.9864; E-mail: luce@medtap.com Research Objective: To estimate the economic value of investment in health care. Study Design: Three methods were used to estimate this value: 1) Incremental adjusted total national health care costs were compared with national health indices (1980-2000); 2) Changes in Medicare expenditures for acute myocardial infarction (AMI), stroke, diabetes and breast cancer were compared to health benefits (1984-2000), using Medicare claims data and National Long-Term Care Survey data; 3) An exhaustive list of key treatment innovations (1970-2000) for the above four conditions were identified, as were all costeffectiveness (cost/life year; cost/QALY) studies pertaining to those innovations. Mortality health benefits derived from all three methods were converted into dollars using published estimates of the value of a statistical life (estimated from established health economic literature to be $100,000/undiscounted life year; $173,000/discounted life year). Disability gains were recorded but were not valued. Population Studied: Method 1: the entire U.S. population; Method 2: hospitalized Medicare beneficiaries with ICD-9-CM codes for the four listed conditions; Method 3: patients suffering from one of the four conditions studied and treated with one of the key interventions. Principal Findings: From 1980-2000, although annual, adjusted health costs increased by $2,254 (102%), each dollar spent on health care in the U.S. is estimated to produce health gains valued between $2.40 to $3.00. Without the health improvements and associated investments, the U.S. would have spent and estimated $634 billion less on health care but would have experienced an estimated 470,000 more deaths, 2.3 million more people with disabilities, and 206 million days in the hospital. Every additional dollar Medicare spent on the indicated conditions produced the following health gains: AMI = $1.10; diabetes = $1.49; stroke = $1.55; breast cancer = $4.80. In addition, the cost-effectiveness literature considered eligible for analysis showed that key innovations for the four diseases are estimated to return positive economic value. For example, each dollar spent on health care produced health gains valued as follows: beta-blockers = $6.50-$35.00 and Mobile CCUs = $10.00 in the treatment of AMI; screening and treatment for retinopathy = $36.00 and intensive glycemic control = $36.00 in patients suffering from diabetes; rtTPA = $9.00 in the immediate treatment of stroke; adjuvant chemotherapy = $2.40-$7.30 in women suffering from breast cancer. Conclusions: Expressed in dollar terms, the value of improved health in the U.S. population and in Medicare beneficiaries, as well as benefits derived from key interventions for the four indicated conditions studied appear to significantly outweigh the costs expended. Implications for Policy, Delivery or Practice: The cost of health care is a continuing national, state, local and organizational concern. Charted by itself, the policy response is often to seek solutions to contain those costs without regard, necessarily, to benefits potentially forgone. Results from this study suggest that, even in the face of doubtless numerous inefficiencies in the U.S. health care system (e.g. medical errors), overall health care expenditures in the U.S. appears to be a good buy. Primary Funding Source: American Hospital Association, Federation of American Hospitals, Advanced Medical Technology Association, American College of Cardiology, Healthcare Leadership Council, National Pharmaceutical Council, Pharmaceutical Research and Manufacturers of America • Drug Expenditures in Canada: A Population-Based Analysis of Trends and Causes Steve Morgan, Ph.D., Charlyn Black, M.D., Ph.D., Morris Barer, Ph.D., M.B.A., Rob Reid, M.D., Ph.D., Robert Evans, Ph.D., Jonathan Agnew, Ph.D. Presented by: Steve Morgan, Ph.D., Assistant Professor, Centre for Health Services and Policy Research, University of British Columbia, 429 - 2194 Health Sciences Mall, Vancouver, British Columbia V6T 1Z3; Tel: 604.822.7012; Fax: 604.822.5690; E-mail: morgan@chspr.ubc.ca Research Objective: From 1991 to 2001, Canadian expenditures on pharmaceuticals doubled. Current forecasts suggest they could double again by 2007, significantly affecting both the necessity and the affordability of other health care services that might be provided to Canadians. The purpose of this study is to quantify the relative and absolute impact of potential determinants of drug expenditures during a recent period of rapid cost growth. Study Design: We build a population-based pharmaceutical dataset describing the type, quantity, and cost of every prescription drug purchased by residents of British Columbia over the period of 1996 to 2002. We further develop a methodological framework for classifying and measuring cost determinants. We then quantify the relative and absolute impact of six potential determinants of prescription drug expenditures that fall into four distinct categories: population characteristics, utilization volume, therapeutic choices, and price factors. Population Studied: This project studies the prescription drug use of all residents of British Columbia over the period of 1996 to 2002. Individual drug use and cost histories are constructed for each of the 4.1 million residents of this Canadian province. The model stratified the population into groupings by age, sex, and overall level of clinical complexity (ACG scores). Principal Findings: Per capita drug expenditures increased by 59 percent over the study period. A majority of prescription drug spending growth in BC was concentrated among 5 therapeutic classes. Demographic changes had a minimal (2 percent) impact on drug expenditures over the period of analysis. After adjusting for generic drug use, direct price factors actually created a cost-decreasing effect of 4 percent. Increased utilization of prescription drugs explained two third of the total increase in per capita spending; these dynamics were sufficient to increase per capita costs by 48 percent. Changes in therapeutic choice contributed a 17 percent cost increase. Cost dynamics differed by age category: ‘babyboomers’ created the largest impact on utilization volume; therapeutic choices were a more significant determinant of seniors cost dynamics. Conclusions: Despite the Canadian systems of price regulation and negotiation, expenditures in British Columbia increased dramatically over the study period. Whereas demographic change is often cited determinant of spending, it was not a significant cost driver. Drug utilization volume and broad therapeutic choices were the most significant determinants of drug spending in this Canadian province. Implications for Policy, Delivery or Practice: Determining the causes and appropriateness of changes in the level and patter of drug utilization is a policy priority. Further research is now underway to determine the health and health system impact of observed utilization trends. Primary Funding Source: Canadian Institutes of Health Research • The Impact of Prescription Coverage on Drug and Nondrug Spending under Medicare Bruce Stuart, Ph.D., Jalpa Doshi, Ph.D., Becky Briesacher, Ph.D., Marian Wrobel, Ph.D. Presented by: Bruce Stuart, Ph.D., Professor, Pharmaceutical Health Services Research, University of Maryland Baltimore, 515 W. Lombard Street, Room 157, Baltimore, MD 21201; Tel: 410.707.5389; Fax: 410.706.1488; E-mail: bstuart@rx.umaryland.edu Research Objective: To determine whether prescription drug coverage for Medicare beneficiaries produces cost offsets in Medicare Part A and B spending. Study Design: We created a 2-year panel of respondents (1999-2000) to the Medicare Current Beneficiary Survey (MCBS). We estimated 4 cross-sectional regression models for spending in 2000 for (1) prescription drugs measured in average wholesale prices (the AWP measure standardizes differences in drug spending for persons with and without coverage and avoids the problem of comparing retail prices for the former with discounted prices for the latter), (2) hospital services,(3) physician services, and (4) all Medicare covered services combined. Explanatory variables included a binary measure of drug coverage and lagged values (1999) for an extensive list of demographic and health status characteristics including predicted Medicare spending derived from the Diagnostic Cost Group/Hierarchical Coexisting Condition (DCG/HCC) risk adjuster developed by CMS. The regressions were estimated using generalized linear models with a gamma distribution and log link. Tests for endogeneity and omitted variable bias were conducted. The analysis was replicated using propensity scoring (PS) techniques in which subjects observations were weighted based on their propensity to have drug coverage derived from predicted values in a logistic regression containing the same variables as in the GLM regressions. Population Studied: Community-dwelling MCBS respondents enrolled in fee-for-service Medicare throughout 1999 and 2000. Sample (N=3,365) was restricted to persons with Medicare supplementation that included drug coverage or not for both years of the study. Principal Findings: Drug coverage was associated with a statistically significant increase in AWP-priced drug spending of between 49% (GLM) and 66% (PS). Drug coverage had an insignificant effect on all of the Medicare spending measures. Coefficient values varied from +7% (total Medicare spending in the GLM model) to -3% (hospital spending using PS methodology). Conclusions: Drug coverage induces significant additional spending on prescribed medications by Medicare beneficiaries. Higher spending on drugs among those with coverage appears to have little aggregate impact on spending for hospital and physician services. This does not mean that drug therapy cannot substitute or complement other therapies, but rather that neither effect predominates across the Medicare population as a whole. Additional research is necessary to identify particular disease-therapy combinations for which cost offsets are likely to occur. Implications for Policy, Delivery or Practice: The new Medicare drug benefit is likely to result in a large increase in spending on prescribed medications for those who currently have no coverage, particularly for low-income beneficiaries who qualify for the most comprehensive benefits. The impact of the drug benefit on the rest of the Medicare budget remains to be seen. If the management of drug therapy by private health plans under the law follows the current practices of third party payers, our findings suggest that the law will have a neutral effect on non-drug spending. Primary Funding Source: CMS • How Much Would a Medicare Prescription Drug Benefit Cost? Offsets in Medicare Part A Cost by Increased Drug Use Zhou Yang, Ph.D., Edward Norton, Ph.D. Presented by: Zhou Yang, Ph.D., Assistant Professor, Medicine, Michigan State University, B412 Clinical Center, East Lansing, MI 48824; Tel: 517.432.8653; Fax: 517.432.9471; Email: zhou.yang@ht.msu.edu Research Objective: To investigate the effect of outpatient prescription drugs utilization on demand for inpatient care and Medicare Part A cost and quantify the offsets in Part A cost from more drug use among elderly Medicare beneficiaries. Study Design: This study estimates a set of dynamic simultaneous demand equations of outpatient prescription drugs and other Medicare covered health services based on health capital and health production function theory to quantify the effects of outpatient prescription drugs use on subsequent demand for inpatient care and Medicare Part A cost. The longitudinal Medicare Current Beneficiary Survey data from 1992 to 1998 is used to estimate the empirical model. Demographic features, insurance status, functional status and health shocks are controlled as independent variables in all the demand equations. Discrete random effect is used to control for unobserved individual level factors that may lead to biased estimation of the relationship between drug use and other medical care cost. Population Studied: Medicare beneficiaries age 65 or more, from 1992 to 1998. Principal Findings: Increase in outpatient prescription drugs use leads to a minor, but significant offsets in subsequent Medicare Part A cost at 4.4 percent on average for the entire sample over a year. The magnitudes of the offsets differ across different subgroups. The offset rates of Medicare Part A cost from more drug use are higher among people without disability (offset rate 4.9%), with chronic diseases (diabetes, offset rate 10.9%), in lower income (less than $15,000 a year, offset rate 7.3%) than people with disability, without chronic diseases and in higher income level. Conclusions: More outpatient prescription drugs use may help the elderly Medicare beneficiaries to maintain better health, and reduce their subsequent demand for inpatient care as well as Medicare Part A cost. Increase in prescription drugs utilization could be more influential among people in low income or high demand for drugs, like chronically ill patients to help them rely less on inpatient care. Implications for Policy, Delivery or Practice: The future cost of Medicare drug benefit may not be as much as the predictions from previous research that did not consider possible offsets in Medicare Part A cost from the increased outpatient prescription drugs consumption stimulated by the drug benefit. However, the magnitudes of the offsets are not as optimistic as predicted by other existing research that concluded the savings in hospital care cost could even exceed the increase in drug cost. Further research consider both the immediate effect of drug use on demand for inpatient care and long term effect of drug use on general health, disability, mortality and total demand for inpatient care at the population level will inform the Medicare drug policy debate even better. Primary Funding Source: NIA Call for Papers Plan & Beneficiary Decisions in the Medicare+Choice Program Chair: Adam Atherly, Ph.D., M.A. Monday, June 7 • 10:30 a.m.-12:00 p.m. • Transitions in Health Plan Choice: Changes in FEHBP Plan Selection When Beneficiaries Begin Medicare Coverage Curtis Florence, Ph.D., Adam Atherly, Ph.D., Kenneth Thorpe, Ph.D. Presented by: Curtis Florence, Ph.D., Assistant Professor, Health Policy and Management, Emory University, Rollins School of Public Health, 1518 Clifton Road, N.E., Atlanta, GA 30322; Tel: 404.727.2818; Fax: 404.727.9198; E-mail: cfloren@sph.emory.edu Research Objective: The Federal Employees Health Benefits Program (FEHBP) has been suggested as a model for Medicare reform. In the FEHBP, enrollees choose from a menu of competing national Preferred-Provider Organizations (PPO) and local Health Maintenance Organizations (HMO), with the government paying 75 percent of the premium, up to a predetermined cap. In general, the HMOs offer low premium plans with low out-of-pocket expenses, but with closed networks. The PPO plans tend to have higher premiums and greater costs sharing for hospitalization and physician office visits, but have the option of open provider networks. The relative benefit differences in the plans change when an FEHBP beneficiary is also enrolled in Medicare, while the premiums stay the same. Most of the plans, both PPO and HMO, waive out-of-pocket payments for hospital and physicians office visits for Medicare beneficiaries. Since the HMO plans tend to have relatively low co-payments for these services in the absence of Medicare, this tends to cause a greater relative increase in the benefits of the PPO plans. Such a change in benefits without a change in premiums may induce enrollees to leave the HMO plans and enroll in the PPO plans. If this occurs, the incentive for enrollees to choose lower costs plans in the FEHBP is compromised. Study Design: We estimate a multinomial logit model where the outcomes are whether the enrollee picked the Blue Cross/Blue Shield (BCBS) Standard Option PPO (the most popular plan in the program), one of the other PPO plans, or an HMO plan. This choice is modeled as a function of age, income, gender. We use administrative data from the Office of Personnel Management that records the plan choice of each federal employee and retiree, and also contains basic demographic information. Population Studied: A time series of plan choices by federal retirees who were age 64 in 1999 (and therefore not Medicare eligible), through 2003. Principal Findings: We find that a significant share of federal retirees change their health plan choice as they transition into Medicare coverage. However, this change is not primarily a change from HMOs to the national PPO plans. Instead, most of the change in enrollment is into BCBS from the other national PPO plans. In the initial year of our panel, 42 percent of enrollees in our panel are in BCBS, 31 percent are in other PPOs and 27 percent are in HMOs. By 2003, 49 percent are in BCBS, 25 percent are in other PPOs and 26 percent are in HMOs. Our regression results show that these trends still hold when controlling for age, income, gender and attrition from the sample. Conclusions: The differences in plan benefits when enrolled in Medicare does induce shifts in plan enrollment, but most of the changes are among the national PPO plans, not from HMOs to PPO plans. Implications for Policy, Delivery or Practice: Medicare reform proposals that incorporate plan choice usually envision managed care competing with traditional Medicare. However, we find most movement of enrollees among national PPO plans. This suggests that reform proposals should consider choices among open network plans as well. Primary Funding Source: RWJF • Medicare+Choice Plan Decisions, 1999-2001 Rachel Halpern, M.P.H. Presented by: Rachel Halpern, M.P.H., Ph.D. Candidate, Division of Health Services Research and Policy, University of Minnesota, MMC 729, 420 Delaware Street, S.E., Minneapolis, MN 55455; Tel: 612.625.5906; Fax: 612.624.2196; E-mail: halp0011@umn.edu Research Objective: To model Medicare+Choice (M+C) plans' decisions to leave counties, change premiums, and change benefits from 1999-2000 and 2000-2001. Study Design: This is a retrospective study that uses secondary data principally from CMS (MedicareCompare databases, State/Plan/County penetration files, geographic service area files, monthly managed care reports), the Area Resource File, and InterStudy. In the first part of the analysis, the decision to drop a county from an M+C plans' service area is estimated with a reduced-form logit equation. The second part of the analysis - currently in progress - models the decisions to change premiums and benefits. In order to look at changes in premiums and benefits, I use the low-price product an M+C plan offered in each county it served. Approaches for modeling these changes include a fullinformation maximum likelihood function or a multinomial probit equation that reflects choices between combinations of premium and benefit changes (e.g., no change in premium/decrease in benefits, increase premium/decrease benefits, etc.) Population Studied: All M+C plans between 1999-2001 Principal Findings: The number of counties with served by any M+C plan decreased significantly between 1999-2001, and M+C plans were far more likely to drop counties than to add them. Preliminary results for the service area analysis indicate that low payment was a highly significant factor in M+C plans' decisions to leave a county between 1999 and 2000, as were low M+C plan enrollment in the county, high costs as measured by the demographic factor-adjusted FFS per-capita expenditures, and the presence of a VA hospital in the county. There also were limited national firm effects. Payment appears not to have been a significant factor in plans' decisions to drop counties between 2000 and 2001 for most models. Low M+C plan enrollment remained significant during this time period and national firm effects increased in number and significance. It also appears that economies of scope, as measured by total M+C enrollment across the M+C plan's entire service area and the HMO's commercial enrollment, were significant during this period. Preliminary descriptive results for the premium and benefit decisions show that, in general, the level of supplemental benefits tended to decrease between 1999-2001. There was a marked increase in the degree of intra-county product differentiation between 1999 and 2000. This tendency to offer multiple M+C products within a county resulted in M+C plans offering $0 premium products as well as products with positive premiums. Thus, although the average M+C product premium rose between 1999-2000, it is important to account for the spectrum of products offered within counties during this period. Conclusions: This study provides theory-based empirical models for M+C plan decisions regarding service areas, premiums, and benefits. The changes in the M+C program between 1999-2001 were complex. While the number of counties served by any M+C plan decreased dramatically, many M+C beneficiaries experienced an increase in M+C product options. There does not appear to be one, monolithic set of determinants of M+C plans' decisions over time; instead it appears that M+C plans react to changes in government payment rates differently in different years. Implications for Policy, Delivery or Practice: If the M+C program is to survive (and thrive), CMS needs a strategy that will retain plan participation and that will be financially feasible. The success of the M+C program, then, depends on understanding the determinants of M+C plans' decisions, as well as the impact of those decisions on Medicare beneficiaries. Results from studies like this one should provide information that will inform CMS in its negotiations with health plans. Primary Funding Source: CMS • Impact of Medicare+Choice Lock-In Provisions: Who Would Be Affected? Mary Laschober, Ph.D., Gabrielena Alcala-Levy, M.A., Erika Alexandra Melman, M.P.P. Presented by: Mary Laschober, Ph.D., Senior Manager, Health Practice, BearingPoint, 1676 International Drive, McLean, VA 22102; Tel: 703.747.6458; Fax: 703.747.8750; Email: mlaschober@bearingpoint.net Research Objective: BBA97 “lock-in” provisions will limit the number and timing of allowed health plan changes for Medicare beneficiaries beginning in 2006. Assumptions underlying lock-in’s goals include that a significant number of beneficiaries switch plans multiple times a year or mid-year to “game” the system. This study investigates these assumptions, analyzing the law’s potential impact by examining recent health plan switching behavior. Study Design: Cross-sectional/longitudinal design using the complete Medicare history for individuals enrolled in a Medicare Group Health Organization (GHO) any time during 1999-2001. Employs multivariate logistic regression, using Medicare’s Enrollment Database linked to M+C plan and market characteristics, to examine relationships between beneficiary, plan, and market characteristics and the likelihood of making multiple plan elections; a health plan change; or a “prohibited election.” Population Studied: 1999-2001 Medicare GHO enrollees (n=6.7 to 7.5 million beneficiaries). Key outcomes: “multiple switchers” (made two or more health plan elections in a calendar year); “prohibited elections” (a health plan election disallowed under the lock-in rules based on the number, timing, or circumstances of plan change). Principal Findings: Multiple switchers averaged 1.7 percent of GHO enrollees (125,000), declining from 1999-2001; nearly 90 percent made only two health plan elections. Variables associated with multiple switching include: African Americans/Hispanics; Medicaid eligibility; under-age-65disabled; first time GHO enrollee (but not new to Medicare); previous enrollment in the same GHO; M+CO availability; and M+CO withdrawal announcement. Prohibited elections constituted less than 10 percent of all enrollment periods, affecting 11 percent of the GHO population, but accounted for 60 percent of health plan elections. Prohibited elections were increasingly to Medicare FFS. About 85 percent of prohibited elections were due to timing and not a multiple plan election, and 70 percent occurred in the second and third quarters. Beneficiary choices displayed the strongest associations with prohibited elections: GHO plan election made from another GHO; previous enrollment in the same GHO during the year; first time ever GHO enrollment; and markets with a withdrawing GHO. Three-fourths of “snowbirds” (n=2,700) made a prohibited plan change. Conclusions: The 1999-2001 experience strongly indicates that the lock-in provisions would directly affect a very small fraction of the Medicare population. Findings show little health plan switching, although they suggest some multiple switching might be to “game” the system. Multiple switches were more likely for some traditionally vulnerable beneficiaries and first-time ever GHO enrollees. The lock-in rules would most impact the timing of plan changes, many of which may be difficult to accelerate or delay. Few beneficiary, plan, or market characteristics showed any systematic association with the likelihood of a prohibited election, primarily being affected by beneficiary choice. Implications for Policy, Delivery or Practice: Direct lock-in impacts would be small, but could negatively impact choice for traditionally vulnerable Medicare populations and mid-year plan switchers because mid-year disenrollment decisions are more often motivated by perceived problems with accessing needed care, desired providers, or plan information. Current lock-in provisions permitting disenrollment due to quality or access issues need to be adequate to protect mid-year and multiple choices. Expected indirect lock-in consequences (e.g., decline in aggregate M+C enrollment) cannot be modeled until the rules take effect. Primary Funding Source: CMS • Medicare Drug Benefits and Selection Bias in HMO Enrollment and Mortality in Diabetes Matthew Maciejewski, Ph.D., Paul Hebert, Ph.D., Marshall McBean, M.D., Bryan Dowd, Ph.D. Presented by: Matthew Maciejewski, Ph.D., Investigator/Assistant Professor, HSR&D / Health Services, VA / University of Washington, VA Puget Sound Health Care System, Stop 152, 1660 S. Columbian Way, Seattle, WA 98108; E-mail: mlmaciej@u.washington.edu Research Objective: Two aspects of Medicare HMOs may be appealing to chronically ill beneficiaries: a focus on preventive care and the provision of drug benefits. We examine whether selection bias by beneficiaries with diabetes occurred between 1995-1998 and whether Medicare HMOs affected mortality rates of these high-risk beneficiaries. Study Design: Data are drawn from the 1994-1998 Medicare Part A, Part B, and Denominator files from the nationallyrepresentative National Medicare Diabetes Cohort. This data are merged with 2000 zipcode-level Census and county-level Medicare AAPCC and HMO benefit files from 1994-1998. Beneficiary risk was proxied by 1994 total Medicare reimbursement and four indicators for diabetes complications of the eye, heart, kidney and lower extremity. A repeated measures logit model was run to estimate discrete-time survival until first HMO enrollment, and repeated measures logit mortality model was also run, controlling for risk proxies, age, gender, race, income education, and county-level variables including AAPCC, out-of-pocket premiums, drug coverage, and eye and ear benefits. Models included time effects and clustering for repeated measures. Initial results will be compared with similar analyses based on a sample matched on age, gender, race, survival time and baseline total costs to reduce selection bias. Population Studied: Our final sample included 51,222 FFS beneficiaries and 9,333 HMO enrollees that were observed for a total of 2,107,766 person-months. Principal Findings: Beneficiaries with diabetes complications or higher total costs at baseline were less likely to enroll in Medicare HMOs between 1995 and 1998. Greater availability of drug benefits led to greater odds of enrollment. HMO enrollees were less likely to die than FFS beneficiaries. Beneficiaries with diabetes complications and higher baseline costs, regardless of HMO status, were more likely to die. Conclusions: Selection bias into Medicare HMOs is evident in a cohort of beneficiaries with diabetes and the availability of drug benefits is an important determinant of enrollment. Preliminary results indicate that HMO enrollment appears to be protective against mortality, even when time-varying and time-invariant risk proxies are controlled. Implications for Policy, Delivery or Practice: The new Medicare drug benefit may be an important innovation for Medicare beneficiaries with diabetes who rely upon pharmacotherapy to manage their condition. Primary Funding Source: AHRQ • Quality of Evidence and CMS Review Times for Medicare National Coverage Decisions, 1998-2003 Peter Neumann, Sc.D., Nomita Divi, M.Sc., Molly Beinfeld, M.P.H., Batsheva Levine, M.D., Patricia Keenan, Scott Gazelle, M.D., Ph.D. Presented by: Peter Neumann, ScD., Associate Professor of Policy and Decision Sciences, Health Policy and Management, Harvard School of Public Health, 718 Huntington Avenue, Boston, MA 02115; Tel: 617.432.1312; Fax: 617.432.0190; E-mail: pneumann@hsph.harvard.edu Research Objective: In 1998, Medicare amended its procedures for making national coverage decisions for new technologies in an attempt to make the process more consistent, transparent, and evidence-based. We examined the quality of evidence available to Medicare decision makers for each technology reviewed since then, as well as CMS review times and factors influencing the time to final CMS decision. Study Design: We reviewed all complete Medicare national coverage decisions from 1998 through 2003 (n=72), based on publicly available decision memoranda posted on the CMS web site. Each memorandum was scrutinized independently by two trained reviewers, who recorded detailed information on approximately 30 variables, including the type of technology assessed, the quality of evidence available (e.g., whether from randomized clinical trials), CMS review times, and the final decision rendered. Population Studied: Medicare Beneficiaries Principal Findings: National Medicare coverage decisions have pertained mostly to medical devices (41%) and surgical procedures (23%). CMS has referred about one quarter of these decisions to the Medicare Coverage Advisory Committee (MCAC). Only 16% of technologies were supported by what CMS considered “good” quality evidence; in contrast, 42% had “fair” and 32% “poor” evidence. For most technologies, CMS decided to “cover with conditions,” rather than reject or accept a new technology. CMS reviews have averaged 8.9 months. MCAC involvement added about 8 months to review compared to decisions not requiring MCAC involvement (p<0.0001). Reviews during 2001-2003 took on average 5 months longer than reviews during 19992000 (p<0.01). 46 % of cases have exceeded CMS’ own guidelines for timely review. Conclusions: The Medicare national coverage process has become more transparent in recent years. Despite the call for evidence-based decisions, the quality of evidence available to CMS for the vast majority of technologies has been suboptimal. CMS frequently fails to meet its standards for timely reviews. Implications for Policy, Delivery or Practice: Coverage of new medical technologies has long presented challenges for public and private payers. CMS’ call for a more open and evidence-based process is commendable. It has become much easier to track CMS decisions and the rationale underlying them. However, this study demonstrates that good clinical evidence is often lacking for new technologies. It also underscores the tradeoffs between requiring rigorous review and timely decision-making. Primary Funding Source: RWJF Related Posters Poster Session A Sunday, June 6 • 6:45 p.m.-8:00 p.m. • The Impact of the Medicare Prescription Drug Legislation’s Payment Increases for Medicare Managed Care Plans on Health Plan Participation, Premiums, and Benefits in 2004 Lori Achman, M.P.P., Marsha Gold, Sc.D, M.P.H. Presented by: Lori Achman, M.P.P., Health Analyst, Mathematica Policy Research, Inc., 600 Maryland Avenue, Suite 550, Washington, DC 20024; Tel: 202.264.3464; Fax: 202.863.1763; E-mail: lachman@mathematica-mpr.com Research Objective: The Medicare Prescription Drug, Improvement, and Modernization Act (DIMA) increased payment rates for Medicare managed care plans in an effort to stabilize the program as a prelude to expanding private plans options in 2006. Because the increases are based on a complex set of changes, the amounts will vary by county but policy analysts expect them to be substantial. (Updated rates will be released on January 16, 2004, with any changes in participation or benefits effective March 2004.) This paper will build on ongoing work to monitor plan participation and benefit package design in the program to assess the immediate effects of the payment increases in this area, including the extent to which any changes in benefits and premiums alter the course of trends in the Medicare+Choice program, which has been towards higher premiums and less generous benefit packages since 1999. Study Design: Analysis of a research database we have created from CMS’s Medicare Compare, which provides information on plan benefit packages and services areas for all Medicare Advantage (formerly Medicare+Choice) contracts. This information is merged with enrollment and payment rate data from CMS as well. The database is current as of January 2004 and we will be adding March information when that becomes available. Population Studied: Medicare+Choice coordinated care plans from 1999 through 2004. Principal Findings: Pre-DIMA plan filings for 2004 show a continuation in historical trends toward somewhat fewer plans available to beneficiaries, more limited pharmacy benefits, greater cost sharing and higher premiums, although the decline is less dramatic than in recent years. We will reexamine these data in light of March 2004 filings to assess (1) the extent to which increased payments generated immediate changes in benefits or premiums versus increased provider payments or use of stabilization funds; (2) the magnitude and kinds of changes that resulted from the enactment; (3) the particular kinds of counties or plans in which changes were made; and (4) how these changes affected existing trends in these areas. Conclusions: Between 1999 and 2003, the Medicare+Choice program experienced significant health plan withdrawals, declining enrollment, increasing premiums, and constricting benefit packages. These findings will provide early evidence on the extent to which the increased payments authorized by DIMA are beginning to stabilize the market and the immediate implications for beneficiaries of the hefty increases. Implications for Policy, Delivery or Practice: As plan premiums have risen and benefits declined, few health plans have been actively recruiting new enrollees to their Medicare+Choice plans. However, should plans put their payment increases into premium reductions and/or benefit expansions, plans may begin actively marketing to Medicare enrollees once again, signaling a renewed interest in the Medicare market. The manner in which health plans use their payment increases from DIMA will likely signal plan strategies going into 2006 and ultimately the potential of Medicare Advantage under the reformed Medicare system. Primary Funding Source: CWF • Out-of-Pocket Prescription Drug Costs and Use among the Elderly: Analysis of 1998 and 2000 MEPS Data Jonathan Agnew, Ph.D., Adams Dudley, M.D., M.B.A., Steve Morgan, Ph.D., Aman Bhandari, M.Sc., Helene Lipton, Ph.D. Presented by: Jonathan Agnew, Ph.D., Post-Doctoral Fellow, Centre for Health Services and Policy Research, University of British Columbia, 429-2194 Health Sciences Mall, Vancouver, BC, V6T 1Z3; Tel: 604.822.4731; Fax: 604.822.5690; E-mail: jagnew@chspr.ubc.ca Research Objective: Out-of-pocket prescription drug (OOPPD) costs are rising significantly for all age groups, particularly the elderly, who are the highest users of prescription drugs. Federal and state policy makers, private insurers and employers, and advocates for the elderly need to understand the determinants of OOP-PD costs to develop policies that control expenditures while ensuring access to needed drugs. However, little is known about trends in OOP-PD costs and predictors of such costs for the elderly. Previous studies have tended to rely on data sources that underestimate OOP-PD costs, do not address how such costs vary among vulnerable and growing populations (e.g., Hispanics), have excluded some key demographic and social variables, and are cross sectional. By using the latest available data, our proposed project addresses these limitations in the literature by examining the following: 1) how much the elderly pay out-of-pocket for prescription drugs; 2) how these payments vary among vulnerable elderly sub-groups (e.g., low-income, “oldest-old,” minorities, those with multiple chronic conditions); and 3) how these payments have changed over time. Study Design: Secondary data analysis of the 1998 and 2000 MEPS datasets. The primary outcome variables were total number of prescriptions per year and OOP-PD. We use a twopart model (logisitic regression to separate those with at least one prescription from those with none, followed by OLS regression) to predict total OOP-PD costs and number of prescriptions. Independent variables were selected according to the Andersen-Newman behavioral model of health services utilization. Population Studied: All Medicare beneficiaries age 66 or older in 1998 (N= 2562) and 2000 (N=2697). Principal Findings: Average OOP-PD costs increased 22% over the two-year period 1998-2000, from $463.69 (95% CI 426.52-500.87) to $564.15 (526.17-602.14). Average number of prescriptions increased 3.6% over the same period, from 20.4 (19.07-21.74) to 21.14 (19.86-22.42). Increases in OOP-PD costs were greater for Hispanic (38%) and African-American (33%) Medicare-eligible elderly than for whites (21%). Those living in the Northeast had greater increases in OOP-PD costs and prescription drug use than their counterparts from other regions. Marital status, household size, education level, gender, income, insurance status, race, health status, and ethnicity were significant (p<0.10) predictors of OOP-PD costs and use. Conclusions: These findings suggest that OOP-PD costs and prescription drug use relate to important, yet understudied, socio-economic and demographic predictors such as region of residence, rural/urban residence, oldest-old, and race/ethnicity. Changes in OOP-PD costs and use over time and across elderly sub-populations suggest that some vulnerable populations may bear a greater burden of cost increases. How the Medicare prescription drug benefit influences such effects will be of prime concern for policy makers. Implications for Policy, Delivery or Practice: Results from this study can be used to aid policy makers when designing, monitoring and potentially modifying the Medicare prescription drug benefit. Quantitative measures of changes over time can provide benchmarks for achieving policy objectives with respect to OOP-PD costs for vulnerable populations. Future research using multiple years of data to analyze trends over time and addressing un-studied or understudied populations is needed. Additional research should pay particular attention to the clinical effects (e.g., curtailment of needed medications) of increased OOP-PD costs. Primary Funding Source: , Jonathan Agnew is supported in part by a post-doctoral fellowship from the Western Regional Training Centre for Health Services Research. • Medication Costs: The Role Physicians Play with Their Senior Patients Mary Sue Beran, M.D., M.P.H., Marianne Laouri, Ph.D., Marika Suttorp, M.S., Robert Brook, M.D., Sc.D. Presented by: Mary Sue Beran, M.D., M.P.H., Clinician Investigator, Internal Medicine, Park Nicollet Health Services, 3800 Park Nicollet Boulevard, Minneapolis, MN 55416; Tel: 952.993.4528; E-mail: beranm@parknicollet.com Research Objective: To examine how often physicians discuss out-of-pocket medication costs with their senior patients, and the cost reducing strategies they employ when these discussions occur. Study Design: Cross-sectional design, using a mailed, selfadministered questionnaire. Population Studied: Cross-sectional, random sample of 1200 internal medicine and family practice physicians in California, selected from the American Medical Association Masterfile. Principal Findings: We obtained completed surveys from 678 of 1098 (62%) eligible physicians. 68% of physicians reported medication cost as "somewhat important" or "very important" when prescribing to seniors, and 43% reported discussing medication costs with more than half of their senior patients in the last 30 days. 40% reported that there was at least one time in the last 30 days when they did not discuss cost but wish they had. The most common reason given was "I ran out of time" (14%). When cost was discussed, 65% of physicians reported that patients initiated the discussion. Predictors of discussing medication cost with half or more of senior patients were group practice setting (vs. other practice settings) and physician rating of cost as of high importance when prescribing a medication for a senior patient (OR 1.5, CI 1.07-2.24). Physician gender, race, specialty, drug cost knowledge score, years in practice, patient volume, and percentage of senior patients seen were not associated with frequency of discussing medication costs. Finally, the most common cost reducing strategy used by physicians was generic substitution (33%) followed by offering samples (25%). Conclusions: More than two-thirds of physicians believe that out-of-pocket medication costs are important, and a surprisingly large percentage of physicians are spending time discussing cost with their senior patients. Implications for Policy, Delivery or Practice: Physicians can play a critical role in helping to address the high out-of-pocket medication costs many seniors face. However, because of time constraints in practice, a tool for accurate and efficient identification of seniors for whom medication costs represent a substantial financial burden is needed. Primary Funding Source: RWJF • Do Hospitals Restrict Utilization of High Cost Pharmaceuticals When the Medicare Payment Rate is Cut? Mary Jo Braid-Forbes, M.P.H., Kevin Forbes, Ph.D. Presented by: Mary Jo Braid-Forbes, M.P.H., Principal, The Moran Company, 1655 N. Fort Myer, Arlington, VA 22209; Tel: 703.841.8402; Fax: 703.465.9969; E-mail: mjbraid@themorancompany.com Research Objective: Under the Medicare Hospital Outpatient Prospective Payment System hospitals could receive passthrough payments for pharmaceuticals, calculated at 95 percent of Average Wholesale Price (AWP). Because the statutory limit on pass-through payments was projected to be exceeded, the Centers for Medicare & Medicaid Services (CMS) instituted an across the board pro-rata reduction to these payments in 2002. The following year payment rates fell further as the products came off pass-through status. The objective of this research is to examine whether the reduction in payment for pass-through drugs that were in effect April 1, 2002 through December 31, 2002 reduced utilization of high cost therapies whose reimbursement fell below acquisition cost. Further, we will investigate whether for-profit hospitals exhibited a more sensitive response to the payment change. Study Design: Data on actual hospital acquisition costs of pharmaceuticals products separately reimbursed under the Outpatient Prospective Payments System (OPPS) were obtained from hospital pharmacy departments who volunteered their data. We analyzed these data relative to the payment rate effective after the pro-rata reductions went into effect. Utilization trends were calculated from the 2001 and 2002 Medicare Outpatient PPS claims files. An econometric model of the relationship between the ratio of payment rate to acquisition cost and utilization trends was estimated. The effect of the hospitals for-profit status on the model was also estimated. Population Studied: Hospital acquisition costs for separately reimbursed pharmaceutical products were obtained from two large hospitals. Quarterly utilization trends were obtained from the Medicare Hospital Outpatient PPS claims files for 2001 and 2002. Principal Findings: Because the final 2002 dataset was not available from CMS until late December 2003, preliminary findings are not yet completed. Conclusions: Implications for Policy, Delivery or Practice: The hospitals’ response to reductions in reimbursement for specific therapies has implications for beneficiary access to care. The hospitals’ response to the pro-rata reductions may foreshadow their response to the larger cuts that occurred in 2003. Primary Funding Source: Corporate • Variation in Charges and Reimbursement for Colorectal Cancer Screening Procedures Jeffrey Burkhardt, Ph.D., Maria Pisu, Ph.D., Xinzhi Zhang, M.D., Ph.D., Mohamad Eloubeidi, M.D., Gerald Glandon, Ph.D., Norman Weissman, Ph.D. Presented by: Jeffrey Burkhardt, Ph.D., Associate Professor, Health Services Administration, University of Alabama at Birmingham, 1675 University Boulevard, Birmingham, AL 35294; Tel: 205.934.1670; E-mail: jburkhar@uab.edu Research Objective: We examine charges and reimbursements for colonoscopy and flexible sigmoidoscopy, two of the primary colon cancer screening procedures, and characterize their variation. Primary factors under consideration as causes of variation are region of the country, specialty of the provider, practice setting, health insurance plan, and type of personnel performing the procedure. We look at market factors that might also contribute to the variation, and examine the variation in the difference between the level of charges and reimbursement. Study Design: We analyzed submitted charges and reimbursement for physician services only through a series of Ordinary Least Square (OLS) regressions. These analyses identify independent predictors of the average charge or reimbursement and the difference between reimbursement and charges. Population Studied: We used data from the Medicare Physician/Supplier Procedure Summary files for the years 1997 to 2001, the Military Health System’s Management Analysis and Reporting Tool (M2) for 1999 to 2001, and MarketScan commercial claims data from the Medstat Group for 1999 to 2001. Principal Findings: There were differences in reimbursement and submitted charges, not only by insurer (Medicare, M2, or MarketScan), but also by region and specialty of provider. The average reimbursement from Medicare and M2 was very similar. For colonoscopy, the average payment from both Medicare and M2 was $232, while reimbursement for flexible sigmoidoscopy was $69 and $61, respectively. The commercial insurance payments from the MarketScan data were almost three times greater for both types of procedures ($764 for colonoscopy and $161 for flexible sigmoidoscopy). Under Medicare, reimbursement was approximately 30% of submitted charges for both procedures, while MarketScan commercial claims paid nearly 70% of submitted charges. When comparisons are made of the four regions (Northeast, South, Midwest, and West), submitted charges and reimbursement were generally highest in the Northeast for colonoscopy and flexible sigmoidoscopy across all payers. Using regression analysis and controlling for externalities, the South frequently had submitted charges that were greater than the West or Midwest, but reimbursement was generally lower. The exception was in the MarketScan data, which showed that charges and reimbursement for colonoscopy were highest in the South. When the procedure was performed by a specialist rather than a generalist, specialists were generally paid significantly more than non-specialists. Conclusions: Generally, procedures performed in the Northeast or by specialists were reimbursed at a higher rate than other regions or by generalists. The commercial claims of the MarketScan data reimbursed a much larger proportion of a much larger amount for both flexible sigmoidoscopy and colonoscopy. Payments to commercial providers were nearly three times greater than to Medicare or M2 providers. Implications for Policy, Delivery or Practice: While Medicare has made payment for colorectal screening procedures part of its benefit package, the low reimbursement paid for performing the procedure may be a deterrent to Medicare providers. Many commercial insurers are reimbursing at a much higher rate. This may be incentive for physicians to perform the lower cost colorectal cancer screening procedures on Medicare recipients, while performing the better paying colonoscopies on patients with commercial insurance. Primary Funding Source: CDC • Efficacy and Toxicity of Simvastatin Compared to Therapeutic Diet in the Treatment of Hyperlipidemia: A Clinical Study Fernando Del Fiol, Ph.D., Lara Cristina Casadei Ubeda, Prof. M.S. Presented by: Fernando Del Fiol, Ph.D., School of Pharmacy, University of Sorocaba, Rua Eng. Urbano P Araujo, 134 ap. 42, Cerquilho, 18520000; Tel: 55.1533843397; E-mail: fernando.fiol@uniso.br Research Objective: Hyperlipidemia occurs because of the increase in the plasmatic concentration of total cholesterol and triglycerides. These increase leads to the development of arteriosclerosis and consequently to coronary heart disease. One of the most used drugs to control the levels of cholesterol and triglycerides is simvastatin, but its use has been related to a great many side effects, involving muscular, heart and hepatic alterations. The objective of our study was to evaluate in humans the efficacy and safety of simvastatin in patients with hyperlipidemia. Study Design: All the patients were submitted to a controlled and standardized diet with low fat levels, composed of 55% vegetables, 30% carbohydrates and 15% proteins. The patients were then divided into two groups. The treated group received simvastatin (10mg/day) plus the recommended diet and the control group received a placebo plus the recommended diet. The groups were submitted to the treatment for 60 days. To assess the efficacy and the toxicity of the drug, blood samples were collected weekly from the patients and the following were assessed; LIPID PROFILE: total cholesterol (TC), high density lipoprotein cholesterol (HDL), low density lipoprotein (LDL), very low density lipoprotein (VLDL) and triglycerides, HEART PROFILE: creatine kinase (CK), lactic dehydrogenase (LDH) and HEPATIC PROFILE: alkaline phosphatase (ALP), aspartate aminotransferase (AST), alanine aminotransferase (ALT), bilirubin and gammaglutamyltransferae(GGT). Population Studied: To assess the efficacy and toxicity of simvastatin use, a double blind study was carried out on 25 patients with diagnosed hyperlipidemia. Principal Findings: Nine weeks after starting the treatment, a reduction in the total cholesterol of 55% was observed for the treated group and of 33% for the control group. For LDL the reduction in the treated group was 64% and for the control group it was 40%. The reduction in the triglycerides levels was 69% for the treated group and 60% for the control group. The HDL levels increased by about 43% for both groups. In the hepatic and muscle toxicity assessment of the drug, increases were observed of 63% in ALP, 158% in AST, 198% in ALT, 102% in bilirubin, 147% in GGT and 102% in CK for the group that received simvastatin. No alteration was detected in patients in the control group Conclusions: With the data from the present study, it was concluded that although the drug was more effective than the diet in decreasing the cholesterol levels, the side effects presented suggest that in some cases only the isolated therapeutic diet can be effective in hyperlipidemia treatment • Generosity of Drug Coverage and Use of Effective Cardiovascular Medications among Aged Medicare Beneficiaries with Heart Disease Jalpa Doshi, Ph.D., Bruce Stuart, Ph.D. Presented by: Jalpa Doshi, Ph.D., Health Services Research Scientist, School of Medicine, University of Pennsylvania, Blockley Hall, Room 1214, Philadelphia, PA 19104-6021; Tel: 215.898.7989; Fax: 215.898.0611; E-mail: jdoshi@mail.med.upenn.edu Research Objective: Under the new Medicare drug benefit legislation, the proportion of drug costs paid out-of-pocket by beneficiaries will vary directly with their total annual drug spending levels. However, the relationship between generosity of drug coverage and essential medication use has not been formally assessed till date. The objective of this study is to examine the impact of generosity of drug coverage on the use of four cardiovascular medications selected from wellestablished quality indicators: 1) warfarin in patients with atrial fibrillation(AF); 2) beta-blockers for patients post-acute myocardial infarction(MI); 3) statins in patients with hyperlipidemia and coronary artery disease(CAD); and 4) ACEinhibitors in patients with diabetes and a cardiovascular risk factor. Study Design: The study used the 1997-2000 Medicare Current Beneficiary Survey, a nationally-representative survey of the Medicare population linked with Medicare claims. Disease conditions were identified using ICD-9-CM codes in the claims. We defined drug use as any report of the study medication filled during the year. Generosity of drug coverage was defined as the percentage of the beneficiary’s annual drug expenditure paid by insurance. All beneficiaries were classified into one of four categories, namely, 0%, 1-50%, 51-75%, and 76-100%. Logistic regressions estimated the likelihood of receiving the study medication while controlling for beneficiary- and disease-related confounders. Population Studied: Fee-for-service community-dwelling Medicare beneficiaries aged 65 years or older with: 1)AF [n=2,915], 2)MI [n=1,522], 3)hyperlipidemia and CAD [n=3,331], and 4)diabetes with a concomitant cardiovascular risk factor [n=3,515] Principal Findings: Overall, only about half to two-thirds of the Medicare beneficiaries with heart disease used cardiovascular medicines recommended for their conditions (warfarin 49.8%, beta-blockers 52.7%, statins 67.1%, ACEinhibitors 52.1% in 2000). No clear association was observed between generosity of drug coverage and use of warfarin or beta-blockers. In contrast, statin and ACE-inhibitor use increased with increasing generosity of drug coverage. Beneficiaries with more than three-quarters of their annual drug bills paid by insurance had over twice the odds of receiving a statin than those with no third-party payments [OR 2.23; 95% CI(1.81-2.74)]. Beneficiaries with >50% of their annual drug bills paid by insurance had higher odds of receiving an ACE-inhibitor than those with no third-party payments or those with 1-50% of their expenditures paid by insurance. Conclusions: Substantial under-use of highly effective cardiovascular medications exists among community-dwelling aged Medicare beneficiaries. Under-use of warfarin and betablockers, which are relatively inexpensive due to the availability of generic substitutes, is not explained by generosity of drug coverage. On the other hand, a strong association was observed for the expensive statins and ACE-inhibitors. Not all beneficiaries who had some drug coverage were at an advantage; use of these expensive medications was correlated with more generous coverage. Implications for Policy, Delivery or Practice: With the passage of the Medicare drug benefit legislation, policymakers have begun to evaluate how the new law will affect drug use among Medicare beneficiaries. This study suggests that the law will fail to provide access to needed but expensive drug therapies for those beneficiaries whose out-of-pocket expenses are not sufficiently reduced under the drug benefit design, but would probably not affect beneficiary use of relatively cheaper medications. • The Impact of the Interim Payment System on Ohio Home Healthcare Agencies Abigail Gerding, B.S.N., M.S., Ph.D., Pamela Salsberry, BSN, M.S.N., Ph.D. Presented by: Abigail Gerding, B.S.N., M.S., Ph.D., Assistant Professor, Family and Community Nursing, East Tennessee State University College of Nursing, Box 70676, Johnson City, TN 37614-1709; Tel: 423.439.5623; Fax: 423.439.4050; E-mail: gerding@etsu.edu Research Objective: The overall objective of the research was to determine the changes that occurred in the home health delivery system with the implementation of the Interim Payment System (IPS) in the state of Ohio. There were three primary issues that were of interest-structural changes in the delivery system, caseload changes, and utilization changes. Study Design: This was a retrospective, explanatory study to examine the impact of the implementation of the IPS on Ohio Certified Home Health Agencies using the Andersen-Aday Model to guide the research. Population Studied: Four hundred and two Ohio certified home healthcare agencies from 1996 and 326 Ohio home healthcare agencies from 1999 were included in the study. Agencies with less than 6 months of service were excluded from the study. Principal Findings: There were 402 agencies in 1996, 326 agencies 1999, and 277 agencies functioning in the market in both 1996 and 1999. 125 agencies left, while 49 agencies entered the Ohio market between 1996 and 1999. Distribution of agencies by agency control demonstrated significant differences (prob<0.0002). Proprietary agencies showed the greatest fluidity with respect to market entry and exit. Rural agencies were less likely to go out of business than urban agencies. With agencies that remained in the market, there was some growth in services offered and most notably an increase in pediatric care. DME, pharmaceuticals, psychiatric care, and respiratory services were dropped by many agencies with significant differences between 1996 and 1999. There was a significant shift downward in the age of the patients cared for by home health agencies in 1999. There was a significant decline in Medicare (non-profit and proprietary, p<0.0001, government, p=0.0018). Case mix demonstrated a significant drop from 3.00 in 1996 to 2.14 by 1999 (p<0.0001). A significant decrease was in mean number of visits per patient with 36% decrease by 1999 (p<0.0025) and mean cost per patient with greater than 30% decrease (p<0.0001). Conclusions: These changes and reductions are impressive in the speed in which they were accomplished and in their magnitude. These results clearly demonstrate the power of reimbursement policy to shape practice as posited by the Andersen-Aday model. While these results show that the IPS was successful in changing practice, it is not entirely clear if the reduced services and fewer visits resulted in limited access for Medicare patients. Implications for Policy, Delivery or Practice: Given the continued economic environment to reduce costs and utilization, home healthcare administrators need to persist in the mechanisms that deliver quality care to those clients who need it while developing strategies to streamline care and productivity. Since home care service seems to be focused on specific needs and more acute care, the need for service to patients with chronic, debilitating disorders must not be overlooked. Reimbursement strategies that allow delivery of care to clients with long term needs should to be investigated, especially those clients who can still manage much of their care with little assistance. Primary Funding Source: Center for Community Based Homecare • Composite Quality Scoring and Pay-for-Performance: The Premier Hospital Quality Incentive Demonstration Project Benjamin Gutierrez, Ph.D., Chris Craver, MA, Stephen Stemkowski, MHA Presented by: Benjamin Gutierrez, Ph.D., Senior Director, Analytics & Research, Premier, Inc., 2320 Cascade Point Boulevard, Charlotte, NC 28208; Tel: 704.733.5447; Fax: 704.733.4666; E-mail: ben_gutierrez@premierinc.com Research Objective: A recent initiative has focused national attention on pay-for performance approaches to reimbursing hospitals. However, researchers and policy-makers currently have limited understanding of effective aggregate performance measurement and incentives for hospitals to more consistently deliver evidence-based clinical interventions. We have refined a composite scoring methodology that incorporates hospital process and outcome measures for three disease categories that will be used in a Medicare demonstration project. The composite scores will be used to measure hospital performance and award bonus payments to top performing hospitals. Study Design: We used patient-level discharge, billing, and JCAHO Core Measure data for the six-month period ending March 31 2003 for 189 hospitals with more than 30 discharges in at least one of three disease categories. Using the quality score, we cross-tabulated the results to explore variation among hospitals based on various hospital characteristics. Population Studied: Our sample of 189 hospitals was drawn from the 278 U. S. acute care hospitals that have agreed to participate in the three year project. Each hospital will submit patient-level discharge, billing and performance measurement data to Premier’s national hospital data repository quarterly. Twenty individual process and outcome indicator values for acute myocardial infarction (AMI), heart failure, and community-acquired pneumonia will be calculated for each hospital and reported semi-annually. Composite scores for each disease category will be calculated to determine top performers, from which incentive payments will be determined. Principal Findings: Our intent was to illustrate preliminary findings after applying the composite score methodology to hospital data and assess the methodology’s capability to discern differences in hospital performance. Preliminary data show considerable variation among sample hospitals in both individual measure values and composite score. For example, aspirin at arrival rate ranged from 84.4% to 99.2% and the beta-blocker at arrival rate ranged from 71.6% to 97.2% for AMI patients. Greater variance was observed for time to thrombolytics, 14.6% to 67.7%, and time to PTCA, 16.7% to 83.3%. Similarly, composite AMI scores ranged from 50.19 to 99.89. The data show comparable variation in other disease categories and across hospital demographic factors. Conclusions: We anticipate that the range of scores for participating hospitals in the demonstration project will decrease significantly over the course of the project and that the median score for the cohort across disease categories will increase as a result of the composite score and the potential for incentive payments to top performing hospitals. Implications for Policy, Delivery or Practice: Introducing payment differentials based on quality may be an effective policy initiative to reduce the variation observed across hospital services. • The Impact of Medicare’s Prescription Drug Law: Empirical Estimates Kevin Hawkins, Ph.D., Ronald Ozminkowski, Ph.D, Jodi Peters, B.S., William Marder, Ph.D., William Crown, Ph.D. Presented by: Kevin Hawkins, Ph.D., Senior Economist, Outcomes Research & Econometrics, The Medstat Group, 777 E. Eisenhower, Ann Arbor, MI 48108; Tel: 734.913.3145; Fax: 734.913.3200; E-mail: kevin.hawkins@medstat.com Research Objective: The new Medicare prescription benefit has sparked debate, but little research with real data has been conducted. This paper utilizes healthcare and pharmaceutical claims from Medicare beneficiaries in supplemental plans to better understand how the prescription law may impact utilization and expenditures. The objectives are: 1) Identify the top 10 most common conditions afflicting Medicare beneficiaries, 2) estimate what percent of expenditures for these conditions are accounted for by pharmaceuticals, 3) illustrate what percent of the population would face the different drug copayment tiers embedded within the new law. Study Design: : The MarketScan Disease Profiler was used to identify the top 10 most common diseases and summarize annual pharmaceutical expenditures for Medicare beneficiaries. The Profiler summarizes inpatient, outpatient, emergency room (ER) and pharmaceutical utilization statistics for over 600 diagnostic categories that include all ICD-9-CM diagnosis codes. The Profiler produces summary statistics noting the amount paid by Medicare and by the patient for each condition. Information for each part of the treatment process (inpatient, outpatient, ER, and pharmaceutical) is produced as well. Population Studied: The MarketScan Medicare database for 2002 was used for this analysis. The data base contains inpatient, outpatient, enrollment and pharmacy claims data from over 960,000 Medicare beneficiaries, from all 50 States, who were covered by supplemental health plans. Principal Findings: The 10 most common conditions for Medicare beneficiaries include: Essential Hypertension, Diseases and Disorders of Skin and Subcutaneous Tissue, Coronary Artery Disease, Cataract, Osteoarthritis, Respiratory System disorders, Disorders of the Gastrointestinal Tract, Nutritional/Immune/Metabolic Disorders, Back Disorders, and Diseases of the Ear/Nose/Throat (ENT). The percent of the population inflicted with each of these conditions varied, with hypertension at 37% and ENT problems in tenth place with 15%. The average annual healthcare expenditure per patient for these conditions is $9,994, with drugs accounting for 29% of the total. After summarizing the Medicare drug spend, 11% of the population would have annual drug spending less then $250, where the copayment rate is 100%. 52% would be in the $251$2,250 range, where the copayment rate is 25%. About 29% would be in the so called “donut hole” ($2,251-$5,100), where the copayment rate is 100%. The remaining 8% would have drug expenditures over $5,100 and have a copayment of 5%. Conclusions: Conclusions The most common health conditions for the Medicare population are chronic and affect a significant portion of the population. The average annual costs of treating these conditions is about $10,000 each, with drugs making up almost a third of the treatment costs. About half of the Medicare patients would face a 25% drug copayment rate, but almost 1/3 would fall into the donut hole and pay 100% for at least some of their drugs. Implications for Policy, Delivery or Practice: Policy In 2006, millions of Medicare beneficiaries will receive pharmaceutical coverage, the effect on these individuals, employers, pharmaceutical manufacturers, and the Government is largely unknown. The impact of the new drug program will vary by condition. The need for additional coverage is not likely to disappear. Primary Funding Source: The Medstat Group • Developing a Reliable Medicare Case Mix Adjuster for Renal Dialysis Facilities Richard Hirth, Ph.D., John Wheeler, Ph.D., Alyssa Pozniak, M.A.E., Marc Turenne, Ph.D., Erik Roys, M.A., Philip Tedeschi, Ph.D. Presented by: Richard Hirth, Ph.D., Associate Professor, Health Management and Policy, University of Michigan School of Public Health, 109 S. Observatory, Ann Arbor, MI 48109-2029; Tel: 734.936.1306; Fax: 734.764.4338; E-mail: rhirth@umich.edu Research Objective: To inform the development of a Medicare case mix adjuster for dialysis facilities by estimating the relationships between case mix and treatment costs using two available data sources to describe each facility's patient population. Study Design: Regression models were estimated explaining costs per dialysis treatment as a function of the dialysis facility's case mix. Dependent variables include total dialysis facility costs, costs for services provided under the existing composite rate for a limited bundle of dialysis-related services, and costs for services that are currently billed separately from the composite rate but are being considered by the Centers for Medicare and Medicaid Services (CMS) for inclusion in an expanded prospective payment system. Cost data are derived from the CMS Cost Report filed annually by each dialysis facility on CMS Form 265-94. All models include a set of basic demographic and clinical indicators for the facility's patient population as well as the CMS area wage index. In addition, the models include alternative measures of patient comorbidities derived from two sources: CMS Form 2728, which measures a baseline history of comorbid conditions at the onset of renal failure, and a measure of the presence of comorbid conditions based on diagnoses reported in Medicare claims for services other than dialysis. Population Studied: All free-standing dialysis facilities in the United States in 2001 linked to information on the characteristics of their patient population. Principal Findings: Models using claims-based comorbidity data performed better than those using baseline comorbidity data from CMS form 2728. These models explain about 11% of the variation across facilities in total and composite rate costs per treatment and about 16% of the variation in separately billable costs. Predicted cost differences are financially significant, with a standard deviation of $12 per treatment, representing over 5% of the average cost per treatment. Using two years of claims data rather than one year of data to identify diagnoses did not improve the model's performance. To reflect the potential for delays in data availability, we estimated claims-based models using different lags between the measurement of comorbidities and costs. Models using claims data with a one year lag performed similarly to those with contemporaneous data in explaining composite rate costs, but somewhat worse in explaining separately billable costs. Performance of all models deteriorated when using claims data with a 2 year lag. Conclusions: Case mix adjustment would result in a significant shift in payments among facilities to reflect the relative costliness of their patient populations. Given the relative homogeneity of the dialysis patients and their treatment, the magnitude of predictable variation in costs is substantial. Claims data out-performed baseline data and are likely to be more reliable and less subject to "gaming" because claims from non-dialysis providers have lower potential for strategic overreporting of comorbidities. Given that new dialysis patients under age 65 generally do not have a Medicare claims history, claims and baseline data could be combined to provide a timely and complete picture of a facility's case mix Both the modelling approach and the combination of baseline and claims data could be useful for Medicare case mix adjustment methods outside of dialysis as all Medicare case mix adjustment systems have to deal with the issue of patients new to the Medicare program. Implications for Policy, Delivery or Practice: Our findings imply that a meaningful renal case-mix adjustment system could be built using existing CMS data. The importance of case mix adjustment would increase further if CMS adds services that are currently billed on a fee-for-service basis, primarily injectible medications, to the prospectively paid service bundle. Further, CMS is expanding the role for case mix adjustment in several areas of the Medicare program. For example, Medicare managed care plans will received capitation rates adjusted for 65 groupings of diagnoses in addition to the basic demographic variables used in the past. This approach is similar to the one we test for dialysis providers. An approach that combines data collected at baseline with claims-based diagnoses that accrue over time could also be a useful template for case mix adjustment in other areas of Medicare policy. Primary Funding Source: CMS • Prescription Drug Coverage and Use among Elderly Medicare Beneficiaries Soonim Huh, M.P.H., Thomas Rice, Ph.D., Susan Ettner, Ph.D. Presented by: Soonim Huh, M.P.H., Ph.D. Student, Health Services, UCLA, 650 S. Young Drive, Los Angeles, CA 900951772; Tel: 310.206.3538; Fax: 310.825.3317; E-mail: soonim@ucla.edu Research Objective: The issue of supplemental prescription drug coverage and drug utilization among Medicare beneficiaries has been a national concern. To date, a majority of prior studies on drug coverage and drug utilization are largely descriptive rather than analytical and, therefore, additional multivariate studies are crucial to investigate the determinants of prescription drug coverage for Medicare beneficiaries and the impact of drug coverage on prescription drug utilization. Furthermore, a few previous multivariate analyses using cross-sectional data have the potential problem of selection bias since the beneficiary’s decision to purchase drug coverage can be correlated with the unobservable characteristics. Thus, the effect of drug coverage on use is a difficult question to study because imprecise estimation that’s the problem with selection bias can create a bias towards results. The objectives of this study are (1) to determine the factors that affect ownership of prescription drug coverage among Medicare beneficiaries aged 65 and older; and (2) to investigate the impact of supplemental drug coverage on prescription drug utilization by Medicare beneficiaries with and without taking selection bias into account. Study Design: We used nationally representative data from the 2000 Medicare Current Beneficiary Survey (MCBS) Cost and Use file. It includes detailed information on the demographic, socioeconomic, health status, and health care experiences including prescription drugs. A probit regression was employed to predict the probability of purchasing drug coverage, which includes proxy measures for the premiums of drug coverage and supply of health insurance as identifying variables. The effect of drug coverage on drug use was estimated by a two-part model, controlling for the observable characteristics. First, any use of drugs was predicted by a probit regression, followed by OLS regressions to predict the level of use, measured as the total number of prescriptions and drug expenditures per year. In order to take self-selection into drug coverage into account, we estimated the two equations simultaneously, allowing the error terms to be correlated. Population Studied: Medicare beneficiaries aged 65 or above, dwelling in the community, enrolled in Part A and Part B. Those with end-stage renal disease and those who died any time during the year were excluded. Principal Findings: Preliminary estimates are expected in April. Conclusions: It's expected in April. Implications for Policy, Delivery or Practice: Findings from this study will address whether Medicare beneficiaries with the need for drug coverage obtain it and to what extent drug coverage has impact on drug utilization after controlling for the potential selection bias. The results will have implications for the outpatient prescription drug benefit and future prescription drug expenditures of Medicare beneficiaries. • Patterns of Prescription Drug Use and Cost among Medicaid-Medicare Dually Eligible Beneficiaries and among Medicaid Beneficiaries Who Resided in Nursing Facilities in 1999: An Analysis of MAX 1999 Files Myoung Kim, Ph.D., Deo Bencio, M.A., Jim Verdier, J.D., Jennifer Schore, M.S. Presented by: Myoung Kim, Ph.D., Health Researcher, Mathematica Policy Research, 600 Alexander Park, Princeton, NJ 08540; Tel: 609.275.2383; Fax: 609.799.0005; E-mail: mkim@mathematica-mpr.com Research Objective: To examine, at the national and the state level, the pattern of prescription drug use and Medicaid expenditures among dually eligible beneficiaries and among beneficiaries residing in nursing facilities in 1999, by beneficiary and by drug characteristic. Characteristics of beneficiaries are defined by basis of Medicaid eligibility, age group, sex, race, and use of nursing facilities during 1999. Characteristics of prescription drugs are defined by brand status (patented brand-name, off-patent brand-name, and generic), therapeutic drug category, and drug group. Measures of drug use and expenditure include the distribution of beneficiaries with respect to the number of prescriptions paid by Medicaid; number of prescriptions and expenditure per beneficiary and per benefit month; expenditure per prescription; pharmacy expenditure as percentage of Medicaid expenditure; and number of prescriptions and cost, by brand status, by therapeutic drug category, and for top 10 drug groups. Study Design: Using Medicaid Analytic Extract (MAX) 1999 files for 47 states, we created two analysis files for our study population: a beneficiary-level file linked beneficiary characteristics to measures of drug use and cost; a drug-level file linked all drug claims associated with the study population to a comprehensive, detailed drug classification system (up to the drug subclass level). We tabulated measures of drug use and expenditure, by beneficiary characteristic and by drug characteristic, at the national level and for each state. We made comparisons among states and examined the nationallevel pattern of use and Medicaid expenditure among dual beneficiaries and nursing facility residents. Use and expenditure measures were compared across states with different pharmacy benefit features to illustrate the potential effect of pharmacy benefit design. Population Studied: (1) Medicaid beneficiaries who were eligible for Medicare and received full Medicaid benefits (including pharmacy benefit) and (2) Medicaid beneficiaries who had full Medicaid benefit and resided in nursing facilities in 1999. Principal Findings: The results cannot be released at this time because they have not yet been cleared by CMS for release. We anticipate, however, the full results to be released in time for the conference. The following is a small sample of outcome measures to be prepared for the presentation: Medicaid programs spent $__ on prescription drugs in 1999. Dually eligible beneficiaries accounted for a large share (__ percent) of this spending; __ percent was for beneficiaries residing in nursing facilities. While __ to __ percent of nondual beneficiaries filled at least one prescription in 1999, the share among dual eligibles and nursing facility residents was as much as __ percent and __ percent, respectively. Dual eligibles filled __ prescriptions per benefit month, __ times as high as that among non-dual eligibles. Nursing facility residents filled an average of __ prescriptions per month. The share of generic prescriptions among dual and non-dual eligibles was comparable (__ percent). This share was higher in states with generic substitution requirements. Antidepressants and antipsychotics combined accounted for as much as __ percent of all Medicaid prescription drug expenditures, __ percent for dual eligibles, and __ percent for nursing facility residents. Implications for Policy, Delivery or Practice: Despite the importance of understanding prescription drug use pattern among dual eligibles—who account for up to 80 percent of Medicaid pharmacy expenditures, detailed data on patterns of drug use and expenditure in this group have not been readily available. In particular, there is little reliable information about the amount and pattern of prescription drug use among beneficiaries who reside in nursing facilities. The new MAX files offers an opportunity to develop in-depth analyses to be used to estimate Medicare prescription drug costs for dual eligibles and nursing facility residents. These estimates will illuminate Medicaid and Medicare prescription drug issues for federal and state policymakers, stakeholders, and researchers with a degree of detail that was not readily accessible before. Primary Funding Source: CMS • New Medicare Beneficiaries’ Understanding of the Medicare Program Mary Laschober, Ph.D., Gabrielena Alcala-Levy, M.A. Presented by: Mary Laschober, Ph.D., Senior Manager, Health Practice, BearingPoint, 1676 International Drive, McLean, VA 22102; Tel: 703.747.6458; Fax: 703.747.8750; E-mail: mlaschober@bearingpoint.net Research Objective: CMS began fielding the monthly Survey of New Medicare Beneficiaries (SNMB) in February 2003 to assess its Initial Enrollment Package, Medicare & You campaign, and other informational materials. This study profiles new eligibles with respect to their basic knowledge of Medicare benefits and supplemental insurance options, awareness of initial decision requirements, sources and usefulness of information, and supplemental insurance holdings. Study Design: The mail-administered SNMB yields approximately 300 completed interviews per month. The cross-sectional sample consists of new Medicare eligibles 65 years old living in the 50 states, DC, and Puerto Rico. It is prestratified by M+CO availability (including PFFS plans), then by resident state and beneficiary characteristics. Frequency and contingency tables were produced for the first six months of data (n=1,901). Statistical significance tests of differences in proportions were conducted to detect variation in subgroup responses. Population Studied: Individuals newly eligible to Medicare recently turning age 65. Sample persons receive their questionnaires within two months after initial Part A enrollment. Principal Findings: A sizeable number of new eligibles are aware of several important Medicare features (e.g., Medicare does not pay all health care expenses (88%), right to decline Medicare Part B (75%), and ability to purchase a Medigap plan (78%)). Many are also familiar with their IEP package (89%) and Medicare Handbook (82%). However, relatively few are aware of the following: Medicare HMO option (56%), right to appeal Medicare decisions (58%), coverage of preventive care services (25%), the Medicare Savings Programs (47%), and the six-month Medigap guarantee (40%). Awareness of Medicare’s toll-free number (50%) and website (40%) is also relatively low. Overall, 86 percent of respondents had supplemental insurance when surveyed. However, there is a significant overlap among Medicare HMO enrollees, and those with employer-sponsored coverage, who had a Medigap policy. Some traditionally vulnerable subgroups (racial/ethnic minority beneficiaries, those in poor health, and low-educated individuals) are substantially less likely to report understanding Medicare features, are less able to effectively use information, and had no supplemental insurance when surveyed. Conclusions: A large percentage of new Medicare enrollees are aware of several important features of their Medicare coverage. However, about one-fourth of respondents said they had very little knowledge about Medicare’s basic features. Additionally, relatively few respondents are aware of some important program aspects, and there is considerable variability in knowledge about specific features. Of particular concern is the comparatively significantly lower proportions (in terms of statistical tests and magnitude) of some traditionally vulnerable beneficiary subgroups who have good Medicare program knowledge. Implications for Policy, Delivery or Practice: There is still room for improving information to a sizeable group new to Medicare. SNMB findings can be used to investigate whether knowledge differences are related to intensity, complexity, or mode of message delivery. Additionally, the findings can be used as benchmarks to assess the early impact of specific or general CMS communication interventions and to aid in intervention design. Differences among beneficiary subgroups signify a continuing need to segment and target Medicare communications, particularly to individuals often in most need of understanding Medicare benefits, rights, and choices. Primary Funding Source: CMS • Racial Difference in Prescription Drug Expenditures among Medicare Only and Dual Eligibles, 1996-2000 Sangho Moon, Ph.D., Alex Sekwat, Ph.D., Jaeun Shin, Ph.D. Presented by: Sangho Moon, Ph.D., Assistant Professor, Institute of Government, Tennessee State University, 330 10th Avenue N., Nashville, TN 37203-3401; Tel: 615.429.7712; Fax: 615.963.7245; E-mail: smoon@tnstate.edu Research Objective: The purpose of this study is to estimate the trend and patterns of the rescription drug expenses among Meicare-Medicaid dual eligibles (e.g., persons having both Medicare and Medicaid) and Medicare only beneficiaries who are either Afro-American or white recipients, focusing on the role of individual-specific medical conditions. Study Design: We used Negative Binomial (NB) regression model to estimate correlations of the prescription drug expenses and daul eligibility associated with diverse ethnic background over the period 1996-2000. The dependent variable is the annual amount of prescription drug expenditure (rxexp). Independent vairables include demographic characteristics, dummies for dual eligibles, race/ethnicity (i.e., asian, hispanic, black, etc.), and interaction of the two binary indicators. Individual-specific chronic health conditions (i.e., diabetes, hypertension, ischemic heart disease, arthritis, depression, etc.) are contolled to obtain unbiased, consistent estimates. Population Studied: The prescription drug expenditure information was from the Medical Expenditure Panel Survey (MEPS) 1996-2000. The source data presented the expenses and frequency of refills for prescription drugs of racial minorities (i.e., Afro-Americans, Hispanics, and Asians) and whites who are either dual eligibles or Medicare only beneficiaries over 1996-2000. Principal Findings: A test of the prescription drug expenditure sample means of dual eligibles and Medicare only beneficiaries showed that dual eligibles' prescription drug expenses are significantly higher than Medicare only beneficiaries, e.g., $1,194.33 versus $822.68; a = .05 and Z calculated = 149.80. NB regression results reveal that dual eligibles are 68% more likely to spend prescription drug expenses than Medicare only recipients. Black dual eligibiles are 38% more likely to spend prescription drug expenses compared with white dual eligibles. The size and statistical significance of the correlation coefficients constantly increased over the years 1996-2000. Conclusions: The proportion of dual eligibles with chronic health conditions was higher by 4.6% than that of the Medicare only beneficiaries and that dual eligibles made more frequent refills of prescription drugs. Both expenses and the number of refills were significantly higher for Afro-American dual eligibles than white dual eligibles. This finding seems to be contributed by relatively high medical needs among dual, and particularly black dual, eligibles due to prevalent multiple chronic diseases. Implications for Policy, Delivery or Practice: It is an important issue in the U.S. public health care finance that prescription drug expenses of Medicare-Medicaid dual eligibles are much higher than those of Medicare only beneficiaries. In order to have cost-effective public health care system facing with increased health care demand among black dual eligibles, the precise assessment of their medical service need is prerequisite. Based on the need assessment, the CMS can provide culturally-sensity coordinated care between Medicare and Medicaid. Additionally, case/disease specific utilization review to monitor physicians’ supply of services is helpful to control the possible overutilization problem among dual eligibles. Induced private competition among MCOs within public health care system is providing an alternative strategy. Primary Funding Source: TSU Alumni Research Fund • Physician Practice Patterns in the Management of Ischemic Heart Disease Among Medicare Beneficiaries Soma Nag, Ph.D., Dennis Shea, Ph.D., Joseph Vasey, Ph.D., Ibrahim Ibrahim, M.D., Ph.D. Presented by: Soma Nag, Ph.D., Health Policy and Administration, The Pennsylvania State University, 1402 Braeburn Terrace, Lansdale, PA 19446; Tel: 215.631.9012; Fax: 215.652.0860; E-mail: soma_nag@merck.com Research Objective: Physician practice patterns significantly impact the cost and quality of care received by patients. The objective of this study is to examine patient, physician, and health system factors associated with the occurrence of physician practice patterns in the treatment of ischemic heart disease - IHD. The research examines the relationship between these factors and the use of consultation, referral, and treatment by the usual physician as practice patterns. Study Design: The data used link together the Medicare Current Beneficiary Survey - MCBS - for 1992 to 1995 including the Part B physician-supplier claims, the 1994 Physician Identification Master Record file, and the 1990 Area Resource File. The unit of analysis is the individual with information on each year split into two half-year observation periods, which are derived based on the claims and total 13,894 for the 4,817 beneficiaries in the sample. A consultation visit was identified using CPT codes related to both office and hospital based consultations with physicians. If a visit was not a consultation, it was considered as treatment by the usual physician if the usual physician and the treating physician were the same, or a referral - which may have been self-referral or a physician initiated referral - if the usual physician differed from the treating physician. The ordered logit method was used to estimate the factors associated with all three practice patterns. Population Studied: The study population included Medicare beneficiaries with evidence of IHD in the MCBS for the years 1992-1995. Principal Findings: Consultation, referral, and treatment by the usual physician account for 9.6%, 54.9, and 35.5% of the total of 13,894 half-year observations. The usual physician’s specialty being in internal medicine; p=0.0026 or cardiology ; p<0.0001 versus in general or family practice reduces the probability of both, consultation by 0.06-2.5% and referral by 3.3-13.1%, and raises the probability of being treated by the usual physician by 3.9-15.6%. Increased IHD severity raises the probability of consultation and referral by 1.6% and 8.4% respectively, while lowering the probability of treatment by the usual physician by 1.0%; p<0.0001. Compared to Medicare only coverage, the presence of supplemental insurance that is obtained either through Medicaid; p=0.0380 or an employer; p<0.0001, or privately purchased; p=0.0001, increases the probability of a consultation by less than 1.0% to about 1.0% and referral by 3.6%-5.4%, while lowering the probability of being treated by the usual physician by 4.3%-6.4%. Conclusions: Practice patterns are primarily influenced by the specialty type of the usual physician and patient need viz. IHD severity. Implications for Policy, Delivery or Practice: Variations in practice patterns have an influence on the process of care and related costs. Therefore, understanding factors that influence these practice patterns which function as the point of access to the subsequent use of resources may be important, especially for chronic conditions such as IHD among Medicare beneficiaries. • Physician Practice Patterns, Interventions, and Costs in the Management of Ischemic Heart Disease Soma Nag, Ph.D., Joseph Vasey, Ph.D., Dennis Shea, Ph.D., Ibrahim Ibrahim, M.D., Ph.D., N. Edward Coulson, Ph.D. Presented by: Soma Nag, Ph.D., Department of Health Policy and Administration, The Pennsylvania State University, 1402 Braeburn Terrace, Lansdale, PA 19446; Tel: 215.631.9012; Fax: 215.652.0860; E-mail: soma_nag@merck.com Research Objective: Physician practice patterns influence the subsequent process of care and costs. This research examined the occurrence of interventions and related costs for ischemic heart disease -IHD - associated with differences in practice patterns representing consultation, referral, and treatment by the usual physician. Study Design: The data used link together the Medicare Current Beneficiary Survey - MCBS - for 1992 to 1995 including the Part B physician-supplier claims, the 1994 Physician Identification Master Record file, and the 1990 Area Resource File. The unit of analysis is the individual where annual data were split into two half-year observation periods - total of 13,894 half-years for 4,817 beneficiaries. A consultation visit was identified using the relevant CPT codes related to both office based and hospital based consultations. The most frequently occurring physician from the claims record with a specialty matching the usual physician specialty indicated in the survey was tagged as the beneficiary’s usual physician. Subsequently, if a visit was not a consultation, it was considered as treatment by the usual physician if the usual physician and the treating physician were the same, or a referral - self-referral or physician referral - if the usual physician differed from that of the treating physician. Interventions examined include use of invasive e.g. CABG and non-invasive procedures viz. catheterization, angioplasty, and drug therapy. The random effects probit method estimated the factors associated with the probability of interventions, while a two-part model estimated the probability of and actual charges generated as a result of these interventions. The models controlled for patient, physician, health system characteristics, and type of practice pattern. Population Studied: The study population included beneficiaries with evidence of IHD in the MCBS. Principal Findings: Consultation or referral increase the probability of receiving any IHD related intervention by 13.3%; p<0.0001 and 7.2%; p<0.0001, respectively, versus treatment by the usual physician. Increased IHD severity and poor / fair versus very good / excellent self-reported health status are associated with an almost 6.0%; p<0.0001 and 2.1%; p=0.0105 increase, respectively, in the occurrence of any intervention. Compared to physicians who are general or family practitioners, physician specialty classified as internal medicine or cardiology increase the probability of any intervention by 3.8%; p<0.0001 and 8.2%; p<0.0001, respectively. Consultation or referral, increase the probability of generating any charges by 13.3%; p<0.0001 and 7.4%; p<0.0001. These practice patterns also increase actual charges generated by 55.8%; p<0.0001 and 11.9%; p<0.0001, where mean charges are $213.38 per half year. Usual physicians who are cardiologists (versus general or family practitioners) generate 32.6%; p<0.0001 higher charges. Conclusions: Consultation contributes to increased use of interventions and charges, compared to referral or treatment by the usual physician. Although the usual physician being a cardiologist increases actual charges incurred by about 32.0%, the magnitude of this impact is lower than that for consultation but higher than that for referral. Implications for Policy, Delivery or Practice: These findings suggest that restricting referrals to specialists for the management of chronic conditions such as IHD may not necessarily be the most effective means of controlling healthcare costs among the Medicare population. • Burden of Illness: Medicare Beneficiaries with Parkinson’s Disease Katia Noyes, Ph.D., M.P.H., Yue Li, M.S., Robert Holloway, M.D., M.P.H., Andrew Dick, Ph.D. Presented by: Katia Noyes, Ph.D., M.P.H., Assistant Professor, Community and Preventive Medicine, University of Rochester Medical Center, 601 Elmwood Avenue, Rochester, NY 14642; Tel: 585.275.8467; Fax: 585-461-4532; E-mail: katia_noyes@urmc.rochester.edu Research Objective: To assess the economic burden of Parkinsons disease (PD) on Medicare beneficiaries, to examine PD population characteristics and to identify factors that contribute to high cost and use of medical services in this population. Study Design: Using the Medicare Current Beneficiary Survey (MCBS), personal characteristics of subscribers were compared based on their PD status. Using data from a randomized clinical trial of PD subjects, a health status question from the survey was mapped into subject’s health utility score (EuroQol). Using log transformed CPI-adjusted dollars, multivariate linear models were developed to explore factors associated with total healthcare costs, Medicarereimbursed total costs, HMO-reimbursed costs, out-of-pocket costs, hospitalization costs, outpatient costs, and prescribed medication costs. Similar negative binomial models were developed for appropriate utilization. Independent variables included age, gender, race, residence area, health status, PD status, co-morbidities, number of activity of daily living (ADL) deficiencies, and year of enrollment. All analyses incorporated complex survey design. Population Studied: 40,881 survey participants representing 1992-1999 Medicare subscribers. Principal Findings: Annual prevalence of PD in the Medicare population was between 1.8 and 2.3%. PD patients were significantly older than the rest of the Medicare population (76.2 vs 71.0 years, p<0.0001), had lower health utility (0.68 vs 0.80, p<0.0001), had more psychiatric problems (19.8% vs 8.8%, p<0.0001) and ADLs (2.8 vs 1.1, p<0.0001). Total annual healthcare expenditures were twice higher for the PD patients than for other Medicare beneficiaries ($21,410 (SE 944) vs $10,686 (SE 137), p<0.0001). Average annual Medicare reimbursement was significantly higher for the beneficiaries with PD ($9,374 (SE 635)) than for others ($5,207 (SE 89), p<0.001). Cost of prescription medications constituted 5.7% of total annual cost ($1,311 SE 73) for the beneficiaries with PD, and 8.1% ($809 SE 8) for the non-PD patients. Beneficiaries paid at least a fraction of the cost for almost every prescription. Medicaid contributed to 25% of prescription payments, while private HMOs paid for 10% of drugs. Based on the results of the multivariate analysis, beneficiary health status was the largest predictor of the health expenditures. Other factors associated with higher costs were age, gender, race, location (metropolitan, North West or East of the U.S), PD status, renal, cardiovascular or psychiatric comorbidities, and ADLs. The use of health services increased significantly every year. Patients who had PD for more than 2 years used more prescribed medications than other beneficiaries, but newly diagnosed PD patients did not. Conclusions: This study provides population-based estimates of the economic burden of PD among the U.S. elderly population. Medicare patients with PD use more healthcare services, cost more, and have lower quality of life and more ADLs than non-PD patients. Implications for Policy, Delivery or Practice: Although roughly 2% of the elderly population has PD, they consume more than 4% of total health care resources. There is an urgent need for disease management programs in the PD population. Primary Funding Source: NIA revenues to supplement the financial contributions of beneficiaries. Conclusions: Based on several new circumstances in 2003, we demonstrate why there was a historic opportunity to add a Medicare prescription drug benefit. We also identify numerous challenges ahead, however, to design an effective policy that meets the actual needs of Medicare beneficiaries. Implications for Policy, Delivery or Practice: Health services researchers, administrators, policy analysts, and policy makers need to understand the political context in order to understand the multiple goals of policy makers, the constraints on policy design, the challenges for policy implementation, and appropriate criteria for policy evaluation. Primary Funding Source: Kaiser Family Foundation • A Political History of Medicare and Prescription Drug Coverage Thomas Oliver, Ph.D., M.H.A., Philip Lee, M.D., Helene Lipton, Ph.D. • Rural/Non-Rural Perceptions of Barriers to Providing Medicare Diabetes Self-Management Education Paige Powell, Ph.D., Saundra Glover, Ph.D., Janice Probst, Ph.D. Presented by: Thomas Oliver, Ph.D., M.H.A., Associate Professor, Health Policy and Management, Johns Hopkins University, 624 N. Broadway, Room 403, Baltimore, MD 21205; Tel: 410.614.5967; Fax: 410.955.6959; E-mail: toliver@jhsph.edu Research Objective: To provide a concise history and analysis of the role of prescription drugs in the evolution of Medicare policy. Study Design: The study applies theories of the policy making process to identify patterns in Medicare policy development, specifically the many missed opportunities to add prescription drugs to the Medicare program prior to 2003. The paper is based on a review of the literature on Medicare politics and policy, political theory, research on prescription drug utilization and costs, and personal experience of one of the authors (Dr. Lee) dating back to 1965. Population Studied: The paper reviews eight critical episodes in the history of Medicare and prescription drug coverage between 1965-2003. Principal Findings: This history illustrates how early policy decisions led to a variety of consequences, some intended and some unintended, that shaped policy options decades later. Most importantly, the omission of outpatient prescription drugs from the initial package of Medicare benefits helped prompt the development of other sources of coverage— employer retirement programs, privately purchased supplemental benefits (“Medigap”), Medicaid, and managed care plans—that generally deterred subsequent efforts to integrate prescription drugs into Medicare. We identify several important patterns in policy making over four decades. First, prescription drug coverage has rarely been considered on its own merits; instead, it has been tied to the fate of broader proposals for Medicare reform. Second, action to add this basic element of modern medicine has been hampered by divided government, federal budget deficits, and ideological conflict between those seeking to expand the traditional Medicare program and those preferring a greater role for private health care companies. Third, contemporary proposals are in part legacies of missed opportunities in the past. In particular, they keep participation voluntary and provide significant subsidies from general governmental Presented by: Paige Powell, Ph.D., Research Assistant Professor, South Carolina Rural Health Research Center, The University of South Carolina, 220 Stoneridge Drive, Suite 204, Columbia, SC 29210; Tel: 803.251.6317; Fax: 803.251.6319; Email: ppowell@gwm.sc.edu Research Objective: To explore the barriers that rural practitioners face in providing diabetes self-management education (DSME) services to Medicare beneficiaries and determine if perceptions of barriers differ between rural and non-rural providers. Study Design: 1,200 American Diabetes Association (ADA) recognized diabetes education centers (DECs) (50.5% of the popultion) were randomly sampled. A one-page survey was sent to the sampled facilities. Respondents were asked to rate the extent to which they agreed or disagreed with 15 possible general barriers to providing DSME. Respondents were then asked if they provided DSME in rural areas and to rate the extent to which they agreed that these 15 barriers were present in rural areas. The same list was used for both general and rural barriers. Thirty-nine surveys were returned invalid and 785 were returned valid, yielding a response rate of 67.6%. Chi-square analyses were conducted between the rural and non-rural barriers and between rural and non-rural providers and their responses to the general and rural barriers. Population Studied: All ADA recognized DECs in the U.S. (n=2,375). Principal Findings: There was more agreement on barriers to DSME in rural areas (six factors with more than 50% agreement) than in urban areas (three factors). Financial factors were the most commonly identified barriers to DSME by all respondents, regardless of service area. Financial barriers included “Too little Medicare reimbursement” in general (73.4%) and in rural areas (78.0%); “patient financial problems”, with higher agreement that this was a barrier in rural areas (72.8%) than in general (59.7%); and “Medicare does not cover enough hours” (55.8% in general, 56.0% in rural). For rural areas, “staffing/financial/institutional support” and “transportation” were identified as barriers by a greater number of respondents than the number of Medicarecovered hours, with 58.2% and 56.8% of respondents agreeing, respectively. The ADA recognition process was perceived to be a barrier to care in rural areas by just over half of all respondents (51.6%). These factors were rated as barriers to DSME in rural areas by more than half of current rural providers: Medicare reimbursement, patient financial problems, Medicare hours, transportation, and institutional support. For eight of the potential barriers, agreement that the factor was a barrier for rural providers was significantly higher among providers who did not serve rural areas. Conclusions: Medicare reimbursement was singled out as a pressing problem for providing diabetes education to Medicare beneficiaries. Patient financial problems and Medicare’s limitation on the number of education hours reimbursed also were prominent problems. There was some variation between rural and non-rural providers’ responses. More barriers were perceived as being present in rural areas than in general. In addition, non-rural providers consistently rated rural barriers more strongly than rural providers, meaning that non-rural providers may have misperceptions about the severity of problems in rural areas. Implications for Policy, Delivery or Practice: The widespread agreement concerning inadequate Medicare reimbursement could indicate a need for change in the current Medicare reimbursement policies and/or amounts. Policy makers should revisit the reimbursement requirements and levels. Primary Funding Source: HRSA, Office of Rural Health Policy • Effect of Changes in Medication Cost-Sharing on Oral Hypoglycemic Use among Diabetes Patients Douglas Roblin, Ph.D., Richard Platt, M.D., M.Sc., Michael Goodman, Ph.D., John Hsu, M.D., M.B.A., Pharm.D., David Smith, Ph.D., Susan Andrade, Sc.D., Stephen Soumerai, Sc.D. Presented by: Douglas Roblin, Ph.D., Research Scientist, Research Department, Kaiser Permanente Georgia, 3495 Piedmont Road, N.E., Building 9, Atlanta, GA 30305; Tel: 404.364.4805; Fax: 404.364.4798; E-mail: Douglas.Roblin@KP.Org Research Objective: In recent years, health insurance plans have added pharmacy benefit options that substantially increase patient cost-sharing. The new Medicare prescription drug benefit has complex cost-sharing provisions. There are few current studies on how high levels of cost-sharing might affect medication use by the chronically ill. We compared change in oral hypoglycemic (OH) use among adults with diabetes (DM) who experienced increases in medication costsharing compared with those who experienced constant medication cost-sharing over a 12-month time period. Study Design: Quasi-experimental, time series with comparison group design. Adults with DM in 5 managed care organizations (MCOs) who experienced an increase in medication cost-sharing (intervention) were matched with those who had no increase. All patients had sustained OH use in a 6-month pre-intervention period and a pharmacy benefit for the 12-month observation period. Patients were matched on pre-intervention benefit, type of OH use, and intervention year. The dependent variable was average daily dose (ADD) standardized to each patient’s mean ADD in the 6-month pre-intervention period. The principal independent variable was change in patient cost per 30-days OH supply between the 6-month pre- and post-intervention periods. Effect of change in cost-sharing on standardized ADD was estimated using both time series regression models and hierarchical linear models (to assess effects of patient covariates). Population Studied: 13,407 adults with DM and a medication cost-sharing increase during the period 1997-2000; 13,407 who had no increase during this period. Principal Findings: At 6-months following an increase in cost-sharing, patients with >$10 increase ($20 median increase) had 20.0% lower mean standardized ADD (p<0.05) than that predicted from their pre-intervention trend in standardized ADD. Patients either with no change in costsharing or with a $1-$5 ($3 median) or a $6-$10 ($6 median) increase did not experience a significant change between the pre- and post-intervention periods in the expected trend of increasing OH dose. 11.5% of those with >$10 increase discontinued their pre-intervention OH medication compared with 6.0% of those with no increase (p<0.05). Postintervention mean standardized ADD among those with >$10 increase remained significantly (p<0.05) lower than that among those with no cost-sharing change after adjusting for patient age, gender, concurrent insulin use, and imputed household income. Conclusions: Large increases in medication cost-sharing (>$10 increase per 30-days supply) were associated with significantly reduced OH use; more modest increases ($1-$10) were not. Whether these reductions in OH use are clinically important (e.g. leading to worse glycemic control) remains to be investigated. Implications for Policy, Delivery or Practice: For patients with a chronic disease, increased cost-sharing for medications may lead to unintended consequences, such as reduction in medication supply or discontinuation of essential medications. Chronic diseases, such as DM, are prevalent among Medicare enrollees. Design of a Medicare prescription drug benefit needs to account for the therapeutic implications of high levels of cost-sharing. Primary Funding Source: AHRQ • Medicare First Dollar Coverage of ACE-Inhibitors for Beneficiaries with Diabetes Saves Money and Lives Allison Rosen, M.D., M.P.H., Mary Beth Hamel, M.D., David Cutler, Ph.D., Milt Weinstein, Ph.D., Sandeep Vijan, M.D. Presented by: Allison Rosen, M.D., M.P.H., AHRQ Health Services Research Fellow, Health Policy and Management, Harvard School of Public Health, 330 Brookline Avenue, BIDMC, Rose 130, Brookline, MA 02215; Tel: 617.667.1342; Fax: 617.667.2854; E-mail: arosen@hsph.harvard.edu Research Objective: Diabetic nephropathy is the leading cause of end stage renal disease (ESRD) in the United States and is associated with marked morbidity, mortality, and costs. ACE-Inhibitor (ACE) use in diabetics slows progression of renal disease and also reduces cardiac morbidity and mortality. Six percent of the Medicare budget is spent annually on the care of the 0.6% of the Medicare population with ESRD. The objective of this study was to assess the health outcomes and budgetary impact to Medicare of first dollar (i.e. no cost-sharing) coverage of ACE for elderly beneficiaries with diabetes. Study Design: Clinical events, survival, and Medicare costs for 65 yo diabetic patients, with and without ACE, were assessed using a Markov cohort model. We assumed that `no coverage´ resulted in 40% ACE use (NHANES 4) at no cost to Medicare, and `Medicare coverage´ increased ACE use by 20% (to 60% overall) based on price elasticity data from the literature. Modeled outcomes included progression of renal disease, cardiovascular events, life expectancy, qualityadjusted life years (QALYs), lifetime costs (2003 US $), and incremental cost-effectiveness ratios. Costs and benefits were discounted at 3%; all analyses took a Medicare perspective. One-way and multi-way sensitivity analyses were performed on uncertain model parameters. Population Studied: Base case estimates are for a typical 65yo Medicare beneficiary. National aggregate estimates were based on the age distribution of the current 6.2 million Medicare beneficiaries with diabetes. Principal Findings: Standard Medicare coverage results in a discounted lifetime cost of $80,129 and quality-adjusted life expectancy of 10.44 QALYs. The addition of first dollar ACE coverage decreased lifetime costs to $79,005 and increased benefits to 10.78 QALYs. Thus, Medicare ACE coverage saves both lives and money. Results were robust to a wide range of renal and cardiac risk reductions, costs, utilities, and discount rates. Results were most sensitive to the cost of ACE (threshold at which no longer cost-saving, drug price >1.8 times the annual average wholesale price) and the impact of coverage on utilization rates (threshold at which no longer cost-saving, increase in ACE use of <7% ). Applying our model to the 6.2 million current Medicare beneficiaries with diabetes, first dollar coverage of ACE would result in a total of 1.3 million QALYs gained and 4.88 billion dollars saved over this cohort's life. Conclusions: Medicare first dollar coverage of ACE-Inhibitors extends life and reduces Medicare program costs. Implications for Policy, Delivery or Practice: A reduction in program costs from one cost-saving intervention will result in more money to spend on other health care needs of the elderly. • The Pent-Up Demand for Medical Services in the Transition from Uninsurance to Medicare at Age 65 Jody Schimmel, M.A. Presented by: Jody Schimmel, M.A., Ph.D. Candidate, Economics, Institute for Social Research, University of Michigan, 426 Thomspon Street, Room 3050, Ann Arbor, MI 48109; Tel: 734.761.3441; E-mail: jodys@umich.edu Research Objective: To analyze the effects of uninsurance and intermittent insurance coverage in the years prior to Medicare eligibility on utilization and expenditures once enrolled in the Medicare program at age 65. This study attempts to disentangle the simple price effects of switching to the Medicare system from the effects of pent-up demand, which would lead to initial increased utilization by those who were not previously insured. Study Design: Difference-in-difference regression models are estimated to analyze the change in service utilization before and after Medicare eligibility for previously uninsured and insured. Changes in utilization and out-of-pocket expenditures are considered for hospital services, doctor visits, outpatient surgery and other clinical services. Regression models include standard demographic controls as well as controls for baseline health status and chronic health conditions. The use of the entire HRS panel from 1992-2002 allows observation of longer windows of insurance patterns and utilization adjustments. Population Studied: Respondents from the Health and Retirement Study (HRS) born between 1931-1941 and their spouses. The HRS is a nationally representative panel of the non-institutionalized United States population over age 50. Specifically, individuals who first reported being age-eligible for Medicare in interview years 1994, 1996, 1998 and 2000 are studied. Principal Findings: Preliminary results indicate that previously uninsured Medicare recipients were more likely than those who were previously insured to increase service utilization after Medicare enrollment. In particular, 6 percent more of the previously uninsured saw a doctor two years after Medicare enrollment (compared to two years prior to enrollment), compared to only about a 2 percent increase in visiting a doctor for those who had been insured prior to Medicare. Doctor visits were higher by about one doctor visit per year for those who had been previously uninsured in the two years after Medicare eligibility. The percentage change in hospital visits between the two years prior to and the two years after enrollment is roughly 4.5 percent higher for those who were previously uninsured. The number of visits to the hospital as well as the number of hospitalized nights are also higher for the previously uninsured once covered by Medicare. Similar qualitative patterns are observed for other measures of utilization. Conclusions: Medicare recipients who were uninsured prior to age 65 increase their service utilization after Medicare enrollment by a larger amount than individuals who were previously insured. More extensive analyses will be performed to look at various definitions of insurance coverage prior to Medicare as well as longer term utilization patterns after Medicare enrollment. Implications for Policy, Delivery or Practice: Reforms to Medicare include the option of buy-in beginning at age 62 or increasing the eligibility age to 67. The relationship between prior insurance status and later Medicare utilization is important as elements of adverse selection may appear in the former plan and moral hazard in the latter, both leading to increased federal expenditures on the Medicare system. Higher Medicare utilization by the previously uninsured and intermittently uninsured after age 65 also suggests the need for changes in the private insurance market to provide coverage for the near elderly. Primary Funding Source: NIA • The Effect of the Availability of Spousal Coverage on the Tradeoff Between Wages and Health Insurance Benefits Jody Schimmel, M.A. Presented by: Jody Schimmel, M.A., Ph.D. Candidate, Economics, Institute for Social Research, University of Michigan, 426 Thompson Street, Room 3050, Ann Arbor, MI 48109; Tel: 734.761.3441; E-mail: jodys@umich.edu Research Objective: To identify the presence of a tradeoff between wages and health insurance in compensation packages offered to workers by their employers. Although economists believe that such a tradeoff must be present in theory, the existing empirical literature has generally not been able to observe this relationship. This paper seeks to identify the wage differential resulting from the offer of employer- sponsored health insurance by utilizing variation in the availability of spousal insurance coverage among married individuals. Study Design: The basic ordinary least squares (OLS) regression of wages on health insurance is inadequate for studying the tradeoff between wages and health insurance since unobserved factors are correlated with both wages and benefits, so that “better” jobs have higher wages and better benefits and thus there is an observed positive relationship. The wage reduction for those without an alternative source of coverage should be larger than for those who have other coverage, since those without other coverage presumably value their coverage more. Here regression models which utilize variation in the availability of spousal coverage is used to identify the wage differential due to the offer of health insurance through one’s own employer. Population Studied: Respondents from the 2002 Health and Retirement Study (HRS), which is a nationally representative sample of the non-institutionalized United States population over age 50. Older workers have higher health care costs than younger workers and thus the reduction in wages due to health insurance that they face are presumably higher and therefore potentially more observable. Principal Findings: Exploiting the variation in the availability of spousal coverage reduces point estimates of the effect of health insurance on wages by thirty to eighty percent as compared to the simple regression on wages on health insurance. Although this identification strategy removes some of the upward bias, the point estimates are still positive and/or statistically insignificant. These results suggest that variation that a priori could be thought to be exogenous does not fully correct for the positive relationship between wages and health insurance. Conclusions: Results here were wrong-signed and/or insignificant, as has been found by other empirical studies. The main problem plaguing most of these studies is that it is quite difficult to find truly exogenous variation that would allow for the removal of the upward bias of the effect of health insurance due to unobserved factors. In order to identify the health insurance-wage differential, different empirical strategies which do not rely on seemingly exogenous differences in health insurance availability are needed in future work. Implications for Policy, Delivery or Practice: In a time of insurance reform, if policy prescriptions to widen the reach of public programs or expand employer-sponsored coverage to more workers are seriously debated, policymakers must consider the reaction of firms and workers to these changes. The lack of conclusive evidence on the magnitude of the tradeoff between wages and health insurance in the existing literature suggests that further research is needed before large-scale changes to employer-sponsored health insurance are enacted. Primary Funding Source: NIA • Prescription Medicine Utilization Trends among Noninstitutionalized Medicare Beneficiaries, 1992 and 2000 Ravi Sharma, Ph.D. Presented by: Ravi Sharma, Ph.D., Analyst, MCBS Survey Operations, RP5022, Westat, 1650 Research Boulevard, Rockville, MD 20850; Tel: 301.738.3589; E-mail: RaviSharma@Westat.com Research Objective: To evaluate the rising trends in the usage of and expenditures on prescription medications among noninstitutionalized Medicare beneficiaries in terms of prescription drug coverage and health status. Study Design: We tracked prescription medication (PM) usage by noninstitutionalized respondents to the Medicare Current Beneficiary Survey (MCBS) in 1992 or in 2000. Using bivariate methods, we compared users with nonusers of PMs by demographic characteristics, health insurance, prescription drug coverage, and health status. We estimated logit models of the likelihood of using PMs. For users, we estimated the average number of PMs and total PM expenditure. Population Studied: All noninstitutionalized, non-ESRD Medicare beneficiaries responding to the MCBS in 1992 or in 2000. Principal Findings: Whereas 85% of beneficiaries used prescription medications in 1992, 91% did in 2000, a trend coincident with expanding PM coverage and a rise in the average number of chronic conditions. Irrespective of year, on average, nonusers of PMs are healthier (as measured by fewer chronic conditions, functional limitations and fair/poor selfreported health status) and are less apt to use major healthcare services. After accounting for beneficiary demographics, health insurance status, PM coverage, and health status, during both 1992 and 2000, the likelihood of PM usage climbed steadily as health status declined. Within each category of health status, the likelihood grew over time, with healthy beneficiaries showing the largest percentage increments. Since 1996 the likelihood of usage steadily rose for beneficiaries with PM coverage as well as those without, albeit a sharper rise occured for the former. Between 1992 and 2000, PM users exhibited rapid growth in PM usage and PM spending. Beneficiaries with poorer health status consumed successively greater amounts of PMs and accrued greater expense over time relative to their healthier counterparts. Average usage increased for each stratum of health status during this period, with unhealthy individuals experiencing the largest increases. A similar relationship is evident between health status and total PM expense. Conclusions: Compared with users, nonusers of PMs are healthier and consume less healthcare resources. After accounting for critical beneficiary characteristics, the temporal rise in the likelihood of using PMs is due to greater PM coverage, the higher incidence (or treatment) of chronic diseases and functional limitations, and other trends (such as direct-to-consumer advertising of PMs) unaccounted for here. Greater severity of illness seems to outweigh the impact of PM coverage on likelihood. The user rate of PMs, and, among users, average PM usage and total PM expenditures, have all shifted up for both healthy and unhealthy beneficiaries, and for both those with and without PM coverage. Implications for Policy, Delivery or Practice: As PM coverage is expanded under the new Medicare law, more intensive drug utilization management may be helpful to curtail unnecessary PM usage. Since unhealthy beneficiaries consume greater PM (and healthcare) resources, greater emphasis on the prevention and limiting the progression of chronic illness and disability among the Medicare population (as well as younger cohorts) is called for. Primary Funding Source: CMS • Prescription Drug Expenditures among Elderly Medicare Beneficiaries: A Comparison of Medical Expenditure Panel Survey and Medicare Current Beneficiary Survey Usha Sambamoorthi, Ph.D., Dennis Shea, Ph.D., Wenhui Wei, MA; M.S., Dennis Shea, Ph.D., Stephen Crystal, Ph.D. Presented by: Dennis Shea, Ph.D., Professor of Health Policy and Administration, Department of Health Policy and Administration, Pennsylvania State University, 116 Henderson Building, University Park, PA 16802; Tel: 814.863.2901; Fax: 814.863.2905; E-mail: dgs4@psu.edu Research Objective: This study evaluates prescription drug expenditures among elderly Medicare beneficiaries from two different data sources Study Design: We used detailed self-reports on prescription drug expenditures from the Medicare Current Beneficiary Survey (MCBS) and Medical Expenditures Panel Survey (MEPS) for the year 1999. Population Studied: Medicare beneficiaries aged 65 and over living in the community. Principal Findings: Overall, average prescription drug expenditures did not differ by data source. However, differences in the average expenditures were noted in certain subgroups of the elderly and the predictors of prescription drug expenditures. Among subgroups of elderly with significant differences in expenditures across the two surveys, we found lower annual prescription expenditures among respondents from the MCBS. With respect to out-of-pocket prescription drug expenditures, MCBS figures were consistently lower than those derived from the MEPS. While gender, health status, functional status, number of chronic conditions remained a predictor of prescription drug expenditures in both surveys, other determinants of prescription drug expenditures varied by data source. Poverty level, health status, and number of chronic conditions are the only common predictors of high burden of drug prescription out-of-pocket expenditure, as a proportion of income. Conclusions: For estimating over all drug expenditures among elderly and examining the associations between gender, health status and expenditures, choice of survey does not seem to matter. However, due to differences in measurement of income, burden of out-of-pocket prescription drug expenditures was not comparable across the two surveys. Implications for Policy, Delivery or Practice: Studies that use magnitude of out-of-pocket expenditures for prediction and studies that use subgroup differences to suggest policy need to be cautious about the source of data. Primary Funding Source: AHRQ Building, University Park, PA 16802; Tel: 814.863.2901; Fax: 814.863.2905; E-mail: dgs4@psu.edu Research Objective: The objective of this research is to identify who among the population of Medicare beneficiaries would benefit from the reduction in or removal of the doughnut hole. Study Design: Demographic and economic data are used to create a 2006 baseline dataset of the Medicare population, and then take-up decisions and drug spending are microsimulated under scenarios of incremental decreases in the size of the doughnut hole in the Medicare drug benefit. The effects of these reductions on increased costs to the Medicare program and decreases in out-of-pocket costs among different groups of Medicare beneficiaries are assessed. Population Studied: The population studied is the community residing Medicare population as of 2000, with a weighting adjustment to reflect the 2006 Medicare population. Principal Findings: As the doughnut hole is reduced, Medicare spending rises dramatically, increasing by approximately 15% for every $1,000 reduction in the stop-loss theshold. More than half of the increased spending goes toward individuals with incomes above 200 percent of poverty, who have significant drug costs, but also significant amounts of income and prior drug coverage. Comparative analysis of reducing the doughnut hole to expanding drug coverage subsidies for near-poor Medicare beneficiaries suggests the latter may target a population in greater need of assistance. Conclusions: The gap in coverage between the spending cap and the stop-loss amount in the Medicare prescription drug benefit has attracted much attention. This is not a common feature of many insurance plans and its implications for beneficiaries and the program are unclear. This paper identifies one possible argument that can be made in favor of this feature. Individuals most likely to reach those spending levels tend to have higher incomes and other potential sources of drug coverage. Thus, reductions in the doughnut hole tend to provide greater benefits to higher income individuals. Implications for Policy, Delivery or Practice: As a Medicare drug benefit moves towards implementation, there are likely to be efforts to modify the program prior to 2006. One feature that may be targeted is a reduction in the gap known as the “doughnut hole”. While such a change may be appropriate for several reasons, the benefits of that reduction may fall mainly to well-off Medicare beneficiaries. Alternative changes in the benefit package may offer more relief to Medicare beneficiaries in greater need. Primary Funding Source: CWF • Eating the Doughnut Hole: Who Benefits? Dennis Shea, Ph.D., Bruce Stuart, Ph.D., Becky Briesacher, Ph.D., Joseph Vasey, Ph.D., Chandra Ganesh, B.S. • Formulary Variability in California Health Plans: A Challenge to Prescribing Physicians William Shrank, M.D., Susan Ettner, Ph.D., Peter Glassman, M.D., Steven Asch, M.D. M.P.H. Presented by: Dennis Shea, Ph.D., Professor, Health Policy and Administration, Pennsylvania State, 116 Henderson Presented by: William Shrank, M.D., Health Services Research and Development Fellow, General Internal Medicine/Health Services Research, UCLA/West LA VA Hospital, 843 11th Street, #8, Santa Monica, CA 90403; Tel: 310.478.3711 Ext. 40920; Fax: 310.268.4933; E-mail: shrankw@aol.com Research Objective: Multi-tiered formularies and patient cost-sharing increasingly have been utilized as mechanisms to control the rising costs of prescription drugs. These developments in the marketplace have forced physicians, who often manage patients from multiple insurers, to become aware of multiple formularies. In this study variability in the formularies of major health plans in California was evaluated to assess the challenge that an individual physician faces when attempting to prescribe a patient the least expensive among similarly effective medications from that patient's formulary. Study Design: We compared the formulary variability within and across 6 major insurers in California from 2000 to 2002. To track this variability, we evaluated 4 of the 5 highest volume drug classes in terms of expenditures: Proton Pump Inhibitors, HMG CoA Reductase Inhibitors (Statins), Calcium Channel Blockers and ACE Inhibitors. Population Studied: Formularies of 6 of the 7 largest health plans in California were compared. Principal Findings: When formulary variability across plans was assessed, no more than 5% of the branded drugs evaluated were “preferred” on either all or none of the formularies. 60% of drugs evaluated in 2000 and 65% in 2002 were “preferred” on 2-4 out of the 6 formularies studied. Variability of formulary coverage within plans between 2000 and 2002 was seen in 11-37% of branded drugs, by drug class. Conclusions: In the classes evaluated over a two-year period, appreciable variability was seen between formularies offered by California health plans, with comparatively less variability within plans. This variability poses a challenge to prescribing physicians and may affect patient compliance with prescribed medications. Systems improvements are discussed, especially in the setting of the Medicare prescription drug benefit which will likely subject increasing numbers of elderly Americans to this formulary variability. Implications for Policy, Delivery or Practice: The complexity in todays system of providing prescription drugs to privately insured patients might be overwhelming for elderly patients who are enrolled in private health plans. As we institute a ne Medicare prescription drug benefit, care must be taken to ensure that beneficiaries are receiving medications that require the least co-payment among medications that are safe and effective. Primary Funding Source: VA • Modeling the 2004/2005 Medicare Transitional Assistance in Oregon’s Medically Needy Program Jeanene Smith, M.D., M.P.H., Judy Zerzan, M.D., M.P.H., Tina Edlund, M.S. Presented by: Jeanene Smith, M.D., M.P.H., Deputy Administrator, Office for Oregon Health Policy and Research, 800 N.E. Oregon Street, Room 607, Portland, OR 97232; Tel: 503.731.3005 Ext. 652; E-mail: Jeanene.Smith@state.or.us Research Objective: In January 2003, people covered by Oregon’s Medically Needy (MN) program lost their benefits due to financial shortfalls in the state budget. The MN program is a federally-matched optional Medicaid program in which states may chose to provide Medicaid coverage and/or Medicare premium assistance to certain groups not otherwise eligible for Medicaid with significant health care needs. This population is not well characterized either nationally or locally. In Oregon, the MN program mainly provided a prescription drug benefit. Recently the Medicare Prescription Drug, Improvement, and Modernization Act (DIMA) was signed into law, ending in 2006 all state Medically Needy programs. During the 2004-2005 period, DIMA provides transitional assistance to Medicare beneficiaries through a prescription drug discount card program for Medicare enrollees and $600 per year toward prescription drug expenses for individuals under 135% of the federal poverty level (FLP). Study Design: A 49-question survey instrument was created to collect information about this population including demographic information, health insurance coverage, health conditions, access to care, prescription drug use and utilization of health care. A vendor telephoned participants with an overall completion rate of 58%. Medication use during the MN program was obtained from administrative data for the past year and current medication use was obtained from the interview. Costs of medications were calculated using average wholesale price plus the current Oregon dispensing fee of $3.50. Population Studied: A statewide, random sample of 1,269 people who were enrolled in the MN program. Principal Findings: The 439 people surveyed included 36% men and 64 % women from age 21-91 with 74% at less than 135% of the FLP. Ninety-two percent have Medicare coverage and only 7.5% currently have prescription coverage. In the six months since the MN program ended, 61% have skipped doses of a medication and 64% have gone without filling a prescription. In order to pay for medications, 60% have cut back on their food budget, 47% have borrowed money and 49% have skipped paying other bills. Yearly prescription costs were calculated in order to model the financial effects of the new Medicare bill for this group. Preliminary results for those under 135% of the FPL show average yearly costs of $2,996. Eighty-seven percent had costs exceeding $600 per year, the amount of the drug discount card for this population. The range of yearly prescription costs is $0-$16,712. The first quartile is $1,304 and the third quartile is $4,025. Further analysis will demonstrate the impact for the entire sampled MN population. Conclusions: The MN program provided coverage for a lowincome, chronically ill population. Since its termination, there have been significant changes in prescription drug use and financial impact in their daily lives. Implications for Policy, Delivery or Practice: In states where prescription coverage options are not offered to this vulnerable population, the 2004/2005 transitional assistance provided in DIMA is inadequate. Low-income, chronically ill individuals who are not eligible for Medicaid or state-only coverage will have significant unmet prescription drugs costs potentially resulting in adverse health effects and continued personal financial impact. Primary Funding Source: Oregon Health Policy and Research through a RWJ State Coverage Initiatives Grant • Does Ownership Form Affect Hospital Responses to Revenue Shocks? Evidence from the Balanced Budget Act of 1997 David Song, B.S., Kevin Volpp, M.D., Ph.D. Presented by: David Song, BS, MD/PhD Candidate, Health Care Systems, The Wharton School of Business, 3641 Locust Walk, Philadelphia, PA 19104-6218; Tel: (215)561-0895; Fax: (215)573-7278; E-mail: dasong@wharton.upenn.edu Research Objective: The Balanced Budget Act of 1997 (BBA) was enacted to save the Medicare program about $116 billion in reimbursements from 1998-2002. It is unknown whether these revenue reductions affected quality of care and whether patient quality declined more in hospitals with larger rate cuts and larger declines in net Medicare patient revenue. We examine the impact of BBA on patient outcomes in California hospitals immediately following the implementation of the policy, and address whether the magnitude of any changes in quality differed by hospital ownership form. Study Design: We use audited hospital financial data, which contain Medicare patient revenue data, and patient outcomes data provided by California’s Office of Statewide Health Planning and Development (OSHPD). We construct a measure simulating the impact of the BBA on hospital revenues through 1996 financial data from OSHPD and preBBA cost report data from the Centers for Medicare and Medicaid Services. Alternative instruments of BBA’s revenue impact are employed as well. Using a dataset of 150,972 Medicare acute myocardial infarction (AMI) admissions in California general acute hospitals between 1996 and 2000, we construct hospital-level mortality rates and risk-adjusted quality measures. We use hospital-level fixed-effects models relating AMI quality measures to the policy impact variable and other covariates. Population Studied: Medicare AMI patients in California admitted between 1996 and 2000. Principal Findings: California hospitals on average experienced a decline in net Medicare patient revenues per admission by 5.0% ($582) from 1997 to 1999, versus a projected decline by 6.9%, according to our simulation. We demonstrate that the degree of decline in net revenue varied with Medicare exposure and resident intensity for California hospitals. Although similar prior to the BBA, Medicare revenues per admission in low Medicare exposure hospitals and high Medicare exposure hospitals diverged post-BBA, as actual Medicare revenues per admission in low Medicare hospitals were $1000-$1500 greater than those in high Medicare hospitals. More important, a 1% simulated decrease in a hospital’s Medicare revenues corresponds to an increase in 6 deaths per 1000 AMI admissions (p=0.020) for that hospital, where the baseline Medicare AMI mortality is 13%. The effect sizes were no different in non-profit hospitals and for-profit hospitals for Medicare AMI mortality (p=0.94); length of stay (p>0.2); and nurse-bed ratios (RNs and LPNs combined, p>0.2). However, non-profits cut charity care to a greater degree than for-profits, even after adjusting for levels of charity care (p=0.03). Mean NP and FP levels of charity care in 1996-97 were $226 and $68 per admission, respectively. Conclusions: Our results suggest that hospitals alter their levels of quality during periods of revenue windfall or shortfall, but that behavioral responses by non-profits and for-profits were similar. Future research would investigate the role of non-profit competition during the BBA, as well as differential NP-FP responses in caring for uninsured patients. Implications for Policy, Delivery or Practice: Policies that are designed to cut costs to reduce a budget's exposure to Medicare expenditures may lead to changes in quality that have important consequences for social welfare. Furthermore, because of its unique mix of non-profit, for-profit, and government-operated firms, studying quality in the hospital industry sheds light on economic theories of firm ownership form. Primary Funding Source: National Bureau of Economic Research • Impact of a Generic-Only Drug Benefit on Medication Use by Medicare Beneficiaries Chien-Wen Tseng, M.D., M.P.H., Robert Brook, M.D., ScD, Emmett Keeler, Ph.D., G. Caleb Alexander, M.D., M.S., Beth Waitzfelder, M.A., Carol Mangione, M.D., MSHS Presented by: Chien-Wen Tseng, M.D., M.P.H., Assistant Professor, Department of Family Practice and Community Health, University of Hawaii/Pacific Health Research Institute, S. Hotel Street, Suite 303, Honolulu, HI 96813; Tel: 808.524.4411; Fax: (808) 524-5559; E-mail: cwtseng@hawaii.edu Research Objective: Rapidly rising prescription costs have led many Medicare+Choice plans to drop brand-name drug coverage and to offer generic-only drug benefits. How generic-only drug benefits affect Medicare beneficiaries’ medication use and financial burden is unknown. Study Design: A cross-sectional telephone/written survey was conducted March-June 2002. We asked patients if they adopted any of six behaviors to lower their drug costs either in 2001, when they had brand-name drug coverage, and/or in 2002, after drug benefits changed to generic-only coverage. These six behaviors included: using current medications less often than prescribed (e.g. skip or decrease a dose),, stopping medications altogether, not starting new medications, switching medications, using free samples, and/or using others’ medications. Participants also reported any financial burden from prescription costs in 2001 and 2002. Population Studied: We sampled participants from a major Medicare+Choice plan in one state whose pharmacy benefits changed from brand-name drug coverage in 2001 to genericonly coverage in 2002. We used the plan’s claims data to identify Medicare beneficiaries age 65 and older who had moderately high annual drug expenditures in 2001 of $2,548 on average. A total of 611 of 973 randomly sampled eligible participants completed the survey resulting in a 63% response rate. Principal Findings: Participants were significantly more likely to adopt 4 of 6 cost-lowering behaviors after brand-name drug coverage changed to generic-only coverage : use less 9%,15%, p<.0001, stop medications 8%,15%, p<.0001, not start new medications 5%, 10%, p<.0001, and/or switch medications 8%, 27%, p<.0001. Rates of using free samples 26%, and taking others’ medications 1%-3% did not differ significantly before/after brand-name coverage ended. Of the 3 behaviors leading to a decrease in medication use, use less, stop, not fill a new prescription, 28% of participants adopted at least one behavior after coverage was limited to genericonly drugs versus 17% when coverage included brand-name drugs, p < .0001. Younger age and lower income were significant independent predictors of decreasing medication use with a generic-only drug benefit, in multivariate analyses examining the independent influences of age, gender, education, ethnicity, health status and income. A greater percentage of participants also reported difficulty paying for prescriptions once the generic-only drug benefit was in place 37%, 65%, p<.0001 and said that drug costs affected enjoyable activities 34%, 44%, p<.0001, caused difficulty paying rent/bills 18%, 24%, p<.0001, and led to not getting medical care 13%,15%,p=.004. Conclusions: In this study of Medicare beneficiaries with moderately high annual drug expenditures, switching from brand-name coverage to generic-only drug coverage resulted in many participants decreasing medication use because of cost and to report financial burden from drug costs. Implications for Policy, Delivery or Practice: It remains to be determined the extent to which appropriate changes to generic medications was possible, and whether careful consideration by physicians/patients on how best to take advantage of generic drugs could have prevented or reduced the financial burden of drug costs after switching to a genericonly drug benefit. Primary Funding Source: RWJF, American Academy of Family Physicians • The Effect of Retiree Health Insurance on Early Retirement Kaan Tunceli, Ph.D., Pamela Farley Short, Ph.D., William Greene, Ph.D. Presented by: Kaan Tunceli, Ph.D., Health Services Researcher, Center for Health Services Research, Henry Ford Health System, One Ford Place, Suite 3A, Detroit, MI 48202; Tel: 313.874.5485; Fax: 313.874.7137; E-mail: ktuncel1@hfhs.org Research Objective: (1) To estimate the effect of planning to retire early on access to employer-provided retiree health insurance (ERHI); and (2) to make a consistent estimate of the effect of access to ERHI on the probability of early retirement, allowing for the possibility that access to retiree insurance is an endogenous variable. Study Design: We utilized a panel study design with three sequential observations at different points in time. The retirement model is specified with three sequential equations predicting retirement expectations in 1992, access to retiree health insurance in 1996, and retirement between 1996 and 2000. Demographic and socioeconomic characteristics of spouses are included in all three equations. In contrast to most previous studies, we used multi-equation estimation techniques (bivariate and trivariate probit) to account for simultaneities involving the three dependent variables, in addition to a conventional univariate probit technique. Population Studied: The empirical analysis uses data from the first five waves of the Health and Retirement Study (HRS), which were conducted every other year from 1992 to 2000. The analysis focuses on men who were less than 65 in 2000 and were full-time workers with insurance as active workers in 1996. The resulting analytic sample includes 1,039 respondents. Principal Findings: The results vary significantly by estimation technique. Planning to retire early increases access to ERHI in the bivariate and trivariate models, while it does not in the univariate model. According to the bivariate model, men who planned to retire early had access to retiree insurance at a rate 20 percentage points higher than their counterparts who did not plan to retire early. The effect of access to ERHI on early retirement is positive and significant in all three models. However, the effect is larger for bivariate and trivariate models compared to the univariate model. According to the bivariate model, full-time employed men who had access to retiree health insurance in Wave 3 were 19 percentage points more likely to retire than their counterparts without access to retiree insurance. The effect of access in the univariate probit was only 5.1 percentage points and weakly significant. All three correlation coefficients between error terms in the structural model were significant (p < 0.01). Conclusions: These findings suggest that the effect of access to retiree health insurance on the probability of retirement might be quite large. In particular, estimation techniques that account for endogeneity and the correlations between the error terms in the structural model produced larger effects of retiree health insurance on early retirement than in most previous studies. This implies that previous research using mostly univariate models might have underestimated the effect of retiree health insurance on retirement. Implications for Policy, Delivery or Practice: Policy initiatives such as a Medicare buy-in or health insurance tax credits that increase access to retiree health insurance are likely to have larger effects on retirement decisions than previous research has suggested. In other words, the unintended effects on labor force participation of policies that increase access to retiree health insurance might be greater than anticipated. • Trends in Medical Spending by Age: 1963-2000 Ellen Meara, Ph.D., Chapin White, M.P.P., Ph.D., David Cutler, Ph.D. Presented by: Chapin White, M.P.P., Ph.D., Post-doctoral Fellow, National Bureau of Economic Research, 14 Suffolk Street, #2, Cambridge, MA 02139; Tel: 617.642.2073; Fax: 617.868.2742; E-mail: chapin_white@post.harvard.edu Research Objective: To measure trends in medical spending from 1963 through 2000, by age group and by type of spending (e.g. hospital care). Study Design: We combine the Medical Expenditure Panel Survey and earlier similar household surveys with the National Health Expenditures (NHE) data on aggregate medical spending. We use the household surveys to estimate the fraction of spending accounted for by each age group, and allocate total spending accordingly. We shift some of the NHE spending across categories to make it more consistent with the household surveys. Because institutionalized individuals are excluded from the household surveys, we use National Nursing Home Survey data to allocate nursing home spending across age groups. Spending is inflated to 2002 dollars using the Consumer Price Index. Population Studied: The U.S. resident population. Principal Findings: From 1963 to 2000, the elderly as a share of the population grew from 9.4% to 12.6%, and the fraction of medical spending accounted for by the elderly grew from 19.7% to 39.2%. Over this period, the annual growth rate in real per capita medical spending was 3.5% among the nonelderly, and 5.3% among the elderly, with per capita spending among the elderly reaching $12,322 in 2000. Among the nonelderly, growth was fairly constant over this period. Among the elderly, however, growth slowed dramatically, with the real annual growth rate falling from 6.9% in 1963-1987, to 2.8% in 1987-1996, to only 0.9% in 1996-2000. Spending on pharmaceuticals accelerated throughout the period for both the elderly and non-elderly, with the real annual growth rate in 1996-2000 reaching 12.9% among the non-elderly and 9.4% among the elderly. Conclusions: Although the elderly share of the population has been growing, population aging has played a very small role in the overall growth in medical spending. The main driver of medical spending growth has been the increase in spending per capita within all age groups. The recent sharp slowdown in spending growth among the elderly likely reflects new Medicare payment policies, which affected acute care hospitals, physicians and, more recently, post-acute care providers. Spending on pharmaceuticals has grown very rapidly among the elderly in recent years, despite the lack of a universal drug benefit. Implications for Policy, Delivery or Practice: Over the long term, spending per capita has grown more rapidly among the elderly than among the non-elderly. More recently, however, Medicare payment policies appear to have slowed the growth in spending among the elderly. These findings suggest that projections of aggregate medical spending should reflect differential trends in spending among the elderly versus nonelderly, and differential trends by spending category. The new Medicare drug bill has two features that make it difficult to project spending growth: it does not include substantial cost containment measures, and it expands benefits for a spending category (i.e. pharmaceuticals) in which the recent growth rate has been very high. Primary Funding Source: NIA Invited Papers New Capitated Alternatives in Medicare Chair: Melvin Ingber, Ph.D. Monday, June 7 • 8:30 a.m.-10:00 a.m. • Panelists: John Kautter and Gregory Pope, both from RTI International; John Robst, Centers for Medicare and Medicaid Services (no abstracts provided) Invited Papers Medicare Beneficiaries & Prescription Drugs Chair: Brigid Goody, Sc.D. Monday, June 7 • 2:00 p.m.-3:30 p.m. • Panelists: Daniel Gilden, JEN Associates; Melvin Ingber, Centers for Medicare and Medicaid Services; Cindy Thomas, Brandeis University; Marian Wrobel, Abt Associates, Inc. (no abstracts provided) Invited Papers Medicare Drug Benefits Chair: Roger Feldman, Ph.D. Tuesday, June 8 • 9:15 a.m.-10:45 a.m. • Panelists: Marisa Elena Domino, University of North Carolina, Chapel Hill; Geoffrey Joyce, RAND; Steven Pizer, Boston University; Dennis Shea, Pennsylvania State University; Robert Town, University of Minnesota (no abstracts provided)