(now ISPOR)
Second Annual International Meeting
April 23-26, 1997, Philadelphia, PA


Invited Presentations

Decision-making for Coverage and Adoption of Medical Technologies Abstracts FS1, FS2
Decision-making in Health Care: Theory Abstract SS1, SS2
Decision-making in Health Care: Bayesian Applications of Preventive Medicine Abstracts TS1, TS2, TS3
Future Directions of Economic Evaluations in Healthcare Abstracts LS1, LS2, LS3

Scientific Presentations - Original Research

Economic and Outcomes Issues in Mental Health Abstracts MH1-6
Disease State Management: Episodes, Outcomes, and Cost of Care Abstracts CC1-6
Economic and Outcome: Issues in Pharmaceutical Care Abstracts PC1-6
Patient Preferences & Analytical Methods Abstracts AM1-6
Comparative Outcomes and Cost-effectiveness Analysis Abstracts EA1-6
Hypertension: Identification of Patients and Costs Abstracts HY1-6
Costs, Resources, and Utilities in Cardiovascular Disease Abstract CD1-6
Economic and Outcomes Issues in Managed Care, Pharmacy Benefit Management, and Government
Abstracts E 1-25
Economic and Outcomes Issues in Clinical Trials, Chart Reviews, Surveys, and Long-term Care Abstracts LC1-25

Applications Workshops

A.P.O.R. Consensus Development Committee’s GAPP (Generally accepted Pharmacoeconomic Principles): Practical issues Abstracts AP1, AP2
Pharmacoeconomic Terminology: Generating a Common Language Abstracts CL1, CL2
Use of Pharmacoeconomics in Decision-making: Practical Issues Abstracts PI 1-5
Pharmacoeconomic Methods: Practical Issues Abstracts PM1-10

Decision-making for Coverage and Adoption of Medical Technologies

Abstract #: FS2

Decision Making for Health Care Technology: The Real World

William T. McGivney, Ph.D

Technology Assessment increasingly has become the foundation for a significant amount of decision making in health care. Technology assessments have become the clinical policy guiding the utilization of drugs, devices, procedures and techniques. Technology assessments are translated directly into coverage policy and to patient selection criteria for use in precertification or utilization management programs in managed care. Technology assessments constitute the building blocks of clinical guidelines for the overall management of disease states.
Many definitions of technology assessment exist with the most common one being:
"Technology assessment is the evaluation of the safety, effectiveness, and cost-effectiveness of the drugs, devices, procedures, and techniques used in the practice of medicine and of the settings under which such care is developed." This presentation will examine the application of the evaluation of safety and effectiveness and of the evaluation of costs and cost-effectiveness in decision-making in the health care system, with particular emphasis on managed care.
Clearly, the focus of coverage decision-making in managed care companies for devices, procedures, techniques and even for drugs has been on the evaluation of safety and effectiveness. Consideration of costs and cost-effectiveness in what even indirectly might be construed as coverage policy has been limited to the development of drug formularies in HMOs. In stating this, it is important to recognize that managed care is not synonymous with the term HMO, but rather encompasses products from managed indemnity to preferred provider organizations to point of service to health maintenance organizations. In national managed care companies, coverage policy tends to be established based on the evaluation of safety and effectiveness and extended consistently across all products.
Trends in methods of reimbursement increasingly will shift the responsibility for establishing clinical and coverage policies away from managed care companies back to the provider communities. Assumption of reimbursement through capitation and global case rates forces providers to better understand the cost implications of the utilization of individual technologies for the overall management of a disease.
The presentation will examine reasons for the limited practical application of cost-effectiveness analysis heretofore in the managed care world and explore the potential for the increasing use of cost analysis and cost-effectiveness analysis under more global methods of reimbursement. Among the issues that will be explored in this context are:
mechanisms for decision-making in managed care companies (i.e. coverage vs. medical necessity);
locus of decision-making authority in the United States vs. other countries;
importance of data derived from controlled trials vs. General practice;
acceptance by society of CEA in decision-making;
CEA analysis in prevention; CEA is okay when we are not really sick;
the complexity of and uncertainty in CEA;
CEA as phase II in the era of accountability in medicine.

William T. McGivney, Ph.D. Biosketch

Bill McGivney recently accepted the position of Chief Executive Officer of the National Comprehensive Cancer Network in Philadelphia. Prior to joining the National Comprehensive Cancer Network, he was Vice President of clinical and Coverage Policy, Aetna U.S. Healthcare where he was responsible for coverage policy across all products (drugs, devices, procedures, and techniques, technology assessment activities, and direct clinical decision support development)
Prior to joining Aetna in June of 1991, Bill spent 10 years with the American Medical Association, most recently serving as Director of the Division of Health Care Technology.
Bill received his Ph.D. in Pharmacology from the University of North Carolina Medical School and then completed a post-doctoral fellowship in the Department of Psychiatry at Harvard Medical School. Bill is a nationally recognized expert in the area of drug and device regulation and coverage and reimbursement policy. In 1989, he was recognized for this expertise and his contributions to drug and device policy development with the FDA Commissioner's Medal of Appreciation. Bill has served on numerous national committees including The Board of Directors of the United Network for Organ Transplantation, the nation's transplant policy board.

Abstract #: FS2

Decision-Making in a Managed Care Formulary
Robert Martin, R.Ph
Foundation Health Corporation and Integrated Pharmaceutical Services adopted a policy in January 1997 to require all pharmaceutical manufacturers and others who wish products to be considered for Formulary acceptance to meet evidentiary and analytical standards in their submission documentation. This presentation will introduce the details of these standards and describe the methodological basis of the guideline evaluation. Submissions will be required to take a systems impact perspective. This will be important in predicting patterns of drug substitution and the cost and outcomes consequences of shifts in therapy. This approach is quite unique in its focus on evidentiary and analytical terms in developing clinical and economic models of a managed health care system, compared to standards required by other health systems in other countries and economic evaluations of pharmaceuticals by expert groups in the United States.

Robert Martin, R.Ph. Biosketch

Robert Martin received his Bachelor of Science in Pharmacy from Purdue University in 1973. He continued his pharmacy training at Purdue and received a Master of Science in Clinical Pharmacy in 1975, including a one year residency in Internal Medicine at Indiana University Hospital in Indianapolis, Indiana. He served as Clinical Instructor and Assistant Professor at the University of Arizona College of Pharmacy for five years. Bob then held Operations Director responsibilities with a large national chain for nine years, and moved into the managed care environment in the last five years with Foundation Health and Integrated Pharmaceutical Services in Rancho Cordova, California, where he was Vice President of Pharmacy Operations. His responsibilities included all aspects of pharmacy operations, as well as Clinical Pharmacy operations and Disease Management Program development. These programs were used to evaluate, educate, and improve the quality and cost-effectiveness of pharmaceutical care management, and promote targeted changes in prescribing behavior impacting long term medical outcomes. Currently, he is Vice President, West Coast Operations, for ProMedex, Inc. a health informatics company that provides electronic data interchange tools for Health Risk Assessments and Clinical Outcomes measurement and monitoring. These tools are available for several disease areas through an electronic health booth and Internet format.

Decision-making in Health Care: Theory

Abstract #: SS1

Methodologic Issues in Cost-Effectiveness Analysis of Health Care Technologies
Milton C. Weinstein, Ph.D.

Cost-effectiveness analysis (CEA) has roots in a variety of disciplines, including economics, operations research, statistics, psychology, and ethics. These perspectives have merged with those of clinical medicine, epidemiology, and pharmacy to produce a vibrant field of research and practice. This diversity of backgrounds is both a blessing and a curse for the field of CEA. It is a blessing because CEA is more widely used than it might be if it were the exclusive property of a single profession. It is a curse because each discipline exacts different, and often conflicting, demands on the methodology. Controversies exist regarding acceptable sources of outcome data for use in CEAs, valuation of health outcomes, appropriate costs to include, and many others. The result has been a lack of consistency in the application of CEA, which undermines its usefulness as a tool for comparing the cost-effectiveness of diverse health interventions.
The Panel on Cost-Effectiveness in Health and Medicine was convened by the U.S. Public Health Service to provide some consistency in the practice of CEA. By appealing to theoretical principles in each of the contributing disciplines, the Panel derived a set of recommendations of the practice of CEA that should make it more broadly applicable. Nonetheless, remaining methodologic issues deserve further attention.
The outstanding issues in CEA encompass four categories: issues relating to the legitimate sources of outcome estimates and the roles of clinical data and modeling; issues relating to the measurement and valuation of health-related quality of life; issues relating to the scope of costs included and the role of the analytic perspective; and issues relating to the ethical underpinnings of CEA.
Both primary data from clinical trials and modeling based on secondary data are valid and essential inputs to CEA. Clinical trials alone have limitations of design and purpose that make them incomplete and often inappropriate sources of data for CEA. Models alone are limited by their validity and perceived validity and are subject to manipulation. Both are needed in order to produce valid and useful CEAs.
There is an emerging consensus that health outcomes in CEA should be measured in terms of quality-adjusted life years (QALYs) and that the quality weights should be based on the preferences of affected persons regarding health states. One question is whether these preferences should be those of the general community or those of the patients affected by the condition being evaluated. The Panel on Cost-Effectiveness in Health and Medicine recommends the use of community preferences, but this leaves open a number of methodologic issues. Is it acceptable that different CEAs use different health-status classification systems and different sets of preference weights, provided that the weights are properly scaled? Should the preference weights be obtained from the general community, or from individuals familiar with particular health impairments (providers, relatives of patients)? Should preference information be collected within clinical trials, or in free-standing community surveys? Do simple rating scales provide valid reflections of preferences, or must choice-based stimuli such as time tradeoff or standard gamble be used? It is encouraging that research is underway to develop a variety of preference-weightable health status instruments, both generic and disease-specific. More research is needed to compare these measures, and to assess under what conditions the choice of instrument and weighting method materially affects the results of CEAs.
The varying roles of CEA for different decision makers have led some to suggest that the societal perspective may not be useful for decisions in managed care organizations or other settings. Economic theory suggests that fully informed consumers who wish to maximize health subject to a budget constraint will select health plans which allocate resources cost-effectively, but numerous market imperfections undermine this model. People switch health plans, which dilutes the incentive of their current plan to adopt a lifecycle perspective on either costs or health outcomes. Cost borne by third parties, such as government, are likewise not reflected in the resource allocation decisions of private agents, which also leads to misallocations of resources. The Panel on Cost-Effectiveness in Health and Medicine recommends the use of a "Reference Case" analysis from the societal perspective, in addition to analyses done from more narrow perspectives.
The ethical foundations of CEA may not seem relevant to health care managers, but they speak to some of the most fundamental issues in health care delivery. Should persons of different ages be given equal priority? Should gains in life expectancy for persons in immediate need of rescue from imminent death be valued the same as gains due to prevention? Should a health benefit for 100 people be considered equal to a benefit 100 times as great for one individual? The calculus of QALYs leads to answers that are sometimes unsettling, but on further reflection may well reflect the views of the community.
The pharmaceutical industry uses CEA as a yardstick to evaluate its products. The health-care industry has been slow to incorporate CEA into their decision making, but there are signs that this is changing. Industry and academia need to work together so that CEA can fulfill its promise.

Milton C. Weinstein, Ph.D. Biosketch

Milton C. Weinstein is an internationally known authority on the methods of cost-effectiveness analysis in health care. He is the Henry J. Kaiser Professor of Health Policy and Management at the Harvard School of Public Health and Professor in the Department of Medicine at Harvard Medical School. He earned his A.B. (1970) and A.M. (1970) in Applied Mathematics, and his M.P.P. (1972) and Ph.D. (1973) in Public Policy at Harvard University. Professor Weinstein's research and teaching focuses on the cost-effectiveness of health practices and technologies, with particular application to heart disease, cancer, and AIDS. He co-chaired the Panel on Cost-Effectiveness in Health and Medicine for the U.S. Public Health Service. He is a member of the Institute of Medicine of the National Academy of Sciences, and a past president of the Society for Medical Decision Making. He is principal author of Hypertension: A Policy Perspective (with William B. Stason, Harvard University Press, 1976) and Clinical Decision Analysis (with Harvey V. Fineberg, W.B. Saunders, 1980). He consults with numerous pharmaceutical companies, managed care organizations, and government agencies.

Abstract #: SS2

Psychology of Judgment and Decision Making
Thomas R. Stewart, Ph.D.
After four decades of research on judgment and decision making, what do we know? We know that people can make remarkably bad judgments in onesituation and remarkably good ones in another. We know that behavior appears "rational" some of the time and wildly "irrational" at other times. We know that human ability to make complex inferences exceeds that of the most powerful computer in some situations, yet cannot match a simple hand calculator in others. In other words, human cognitive ability varies from person to person, from time to time, and from situation to situation.
Despite thousands of studies, research on judgment and decision making has had little impact on how judgments and decisions are actually made. This is not because little of value has been discovered, but rather because of a) the difficulty of making sense of a bewildering variety of seemingly contradictory results and b) the lack of a clear prescription for applying the research to real decisions.
Two developments, one new and one old but often ignored, offer promise both for resolving the apparent inconsistencies in the research and for understanding how to apply it to real decisions. The new development is the observation by Kenneth Hammond that there are two groups of judgment and decision researchers who implicitly use quite different standards for "good judgment." One group applies a coherence standard (Are judgments logically consistent?) while another applied a correspondence standard (Are judgments correct, according to some objective criterion?). Of course, this distinction is important in all of science as well as in everyday life. The older development, originating with Egon Brunswik and refined and extended by Kenneth Hammond, is that one cannot understand internal judgment processes without understanding the external environment that creates the need for judgment. With an understanding of the coherence/correspondence distinction and the importance of the environment for judgment and decision making, it is possible to derive some important lessons that can be applied to understanding and improving real decisions.

Thomas R. Stewart, Ph.D. Biosketch

Thomas R. Stewart, Ph.D. is Director for Research at the Center for Policy Research, University at Albany, State University of New York. He received his Ph.D. in quantitative psychology from the University of Illinois and was formerly with the Graduate School of Public Affairs and the Center for Research on Judgment and Policy at the University of Colorado. He specializes in theoretical, methodological and applied studies of judgment and decision making, and is particularly interested in expert judgment and medical decision making. His specific interests include methods of judgment analysis and the decomposition of judgmental skill; studies of judgment in medicine, weather forecasting, and global environmental change; and the role of scientific uncertainty and disagreement in policy formation.

Decision-making in Health Care: Bayesian Applications of Preventive Medicine

Abstract TS2
Analyzing Clinical and Pharmacoeconomic Data for Resource Allocation
Using Bayesian Analysis
Dennis G. Fryback, Ph.D

The decision to include a new agent or not in a formulary depend on many things and among these are the incremental magnitude of benefit to patients, and the likely incremental costs of securing this benefit, compared to currently employed treatments. The purpose of this presentation is to illustrate in a tutorial fashion elementary Bayesian analysis of data from clinical trials to compute probability distributions over quantities such as these that are critical for decision making.
Bayesian analyses require two ingredients, a prior distribution over the quantity of interest and observed data. These are combined by standard probability calculations into a posterior probability distribution over the quantity of interest. The posterior distribution quantifies the likelihood of all possible magnitudes of benefit (or cost) given both what was known before the trial and the observed data from the trial. An approximation for this calculation will be demonstrated using simple, spreadsheet based computations without specialized software, for the common case of trials with binomial outcome data bearing on a proportion of interest (e.g., a 1-year mortality probability), and for the case of continuous data relevant to inferences about a quantity with a Normal distribution.
The prior distribution reflects the beliefs of the decision maker before the experimental data are observed. Perhaps the most controversial point of Bayesian analysis is use of information other than the directly observed experimental data in statistical calculations. In this talk we will demonstrate how to systematically elicit a Beta prior distribution for a binomial proportion and the circumstances under which this prior distribution does or does not have negligible effect will be illustrated.
The simple Bayesian analyses illustrated here can serve in many circumstances, and the critical concepts of Bayesian analyses will be demonstrated by them. We will conclude with a brief survey of more advanced applications and frontiers Bayesian calculations for pharmacoeconomic data.

Dennis G. Fryback, Ph.D. Biosketch

Dennis Fryback received his BS degree from UCLA (Psychology-Mathematics;1969), and MA (Mathematics, 1973) and PhD (Psychology, 1974) from the University of Michigan. He joined the faculty of the University of Wisconsin-Madison in 1974 with appointments in the Department of Preventive Medicine and in the Department of Industrial Engineering. He has been Professor of Preventive Medicine and Industrial Engineering since 1984. Active as a teacher and researcher in medical decision making and technology assessment, he is a founding member, and fourth President, of the Society for Medical Decision Making. He was the second Editor-in-Chief of the society's journal, Medical Decision Making, and and in 1994 he was one of three people who received that society's EL Saenger Career Service Award in the first year the award was given. In July 1996 he was elected Fellow of the Association for Health Services Research. Dr. Fryback's research and consulting interests include medical technology assessment, health care cost-effectiveness analysis, measurement of health status and health-related quality of life assessment, and clinical decision analysis. He chaired the Board of Scientific Counselors of the National Library of Medicine from 1992-94, and the Health Care Technology Study Section for the Agency for Health Care Policy and Research from 1992-96. He was one of 10 members of the United States Preventive Services Task Force, appointed in 1990 by the Assistant Secretary for Health to prepare the second edition of the Guide to Clinical Preventive Services for publication in January 1996, and served from 1993-96 as a member of the Panel on Cost Effectiveness in Health and Medicine convened by the Office of Disease Prevention and Health Promotion of the US DHHS to prepare the definitive text on cost-effectiveness analysis methods for the Public Health Service (Cost-Effectiveness in Health and Medicine. New York: Oxford University Press, July 1996). Dr. Fryback regularly provides consulting services to private companies, and academic, state, and federal research programs to assist in designing and analyzing studies of medical decision making and medical technology assessment.

Abstract TS3
Estimating the Loss Function in a Bayesian Analysis
Josephine A. Mauskopf, Ph.D., M.H.A., M.A.

A hospital or managed care organization bases their decision as to whether or not to include a new drug on the formulary on estimates of the magnitude and uncertainty of the health and cost outcomes with the new drug compared to current treatments. The estimates of uncertainty may be expressed as probability distributions using a Bayesian analysis. Whether or not the decision maker includes the new drug on the formulary will also depend on their 'loss' function. This loss function is a representation of the decision maker's utility or value for each possible combination of health and cost outcomes based on their risk preferences. The decision maker will include the new drug on the formulary if there is a gain in expected utility or value from using the new drug rather than remaining with the current treatment. The parameters defining the 'loss' function will depend on factors such as the seriousness of the disease to be treated and the competitiveness of the market.
There are two standard economic approaches that have been used to estimate loss functions for consumers of durable goods such as automobiles and home heating systems. The first method is the standard gamble estimation approach which requires the decision maker to choose between a product with values of the outcomes that they can achieve with certainty or a second product where there is a chance of experiencing either the better or worse values of the outcomes. The second method is a random utility approach where the decision maker is asked to choose between products with different attributes including expected outcomes and their uncertainty. The data from either of these approaches can be used to estimate a 'loss' function.
Before these approaches can be applied to health care decision making, we need a better understanding of the role that different drug attributes play in hospital and managed care decision making. Such an understanding can be obtained from structured interviews with the decision makers.
Once loss functions have been estimated they can be combined with the Bayesian estimates of the probability distributions for the outcomes to determine the expected utility when adopting the new drug compared to staying with the current treatment. Estimates of the loss function can also be used to determine the probability of the decision maker choosing to adopt a new drug under different pricing scenarios and under different levels of uncertainty about the outcomes compared to current treatment.

Josephine A. Mauskopf, Ph.D., M.H.A., M.A. Biosketch

Dr. Mauskopf is Senior Director of the Pharmacoeconomics Program in Research Triangle Institute's Center for Economics Research. She has extensive experience both as a consultant and within the pharmaceutical industry designing and implementing pharmacoeconomics research strategies. She has estimated the cost-effectiveness of anti-retroviral drugs, as well as drugs for treating herpes zoster, epilepsy, neonatal respiratory distress syndrome, digoxin toxicity, and primary pulmonary hypertension. She has also managed projects to develop or culturally adapt psychometrically valid quality-of-life measures for migraine, genital herpes, and epilepsy. She has estimated the impact of an antidepressant on work and social disability. She has developed Markov models of disease progression for lung cancer and HIV infection and has developed a simulation model of time spent in the operating and recovery rooms. Dr. Mauskopf was previously Department Head of Economics Research at Burroughs Wellcome Co. and Director of Pharmacoeconomics Research for Anti-Virals and Anti-Infectives at Glaxo Wellcome Inc. She has published extensively in journals including The Journal of the American Medical Association, Journal of Pediatrics, Pharmacoeconomics, Medical Care, The American Journal of Public Health, Health Services Research, and American Journal of Cardiology.

Future Directions of Economic Evaluations in Healthcare

Abstract LS1
Making Economic Evaluations Respectable
Uwe E. Reinhardt, Ph.D.

Cost-effectiveness and cost-benefit analysis incorporate numerous assumptions and judgments by the analyst. These assumptions and judgment may reflect the analyst's own personal inclinations or perhaps that of the sponsor's. Whatever the drive, for these economic evaluations in health care to be taken seriously by the health care decision maker, the authors must be held accountable for the quality of their work.
A useful model in this respect is the accounting profession. The trust in and respect for the accountant's work rests on three major pillars.
1. STANDARDS: The accounting profession has developed a standard body of account techniques that are shared world-wide. Example: "the ledger"
2. CODE OF ETHICS: A widely shared set of principles guide the valuation of the asset and liabilities.
3. WILLINGNESS FOR AUDIT: The accounting profession is willing to subject their work to deeply probing outside audit at a moment's notice.
Road To "Respectability" For Pharmacoeconomics:
1. STANDARDIZATION: The first step in attaining "respectability" is to seek a much greater standardization of pharmacoeconomic evaluations. The product of this standardization must take on the force of binding rules.
2. CODE OF ETHICS: The second step is to develop a set of principles of conduct for analysts to follow. Empower professional health economic organizations to enforce the principles.
3. AUDITS: Establish an audit process.
In the next century, decisions in health care increasing will be driven by sophisticated information systems available to both sides of the market. Health care research, in general, and pharmacoeconomic analyses, may be able to play a significant role if the studies can overcome certain obstacles and reach a new level of :respectability.

Uwe E. Reinhardt, Ph.D. Biosketch

Uwe E. Reinhardt, a native of Germany, has taught at Princeton University since 1968, rising through the ranks from assistant professor of economics to his current position. He has taught courses in both micro- and macro-economic theory and policy, accounting for commercial, private non-profit and governmental enterprises, financial management for commercial and non-profit enterprises, and health economics and policy. Professor Reinhardt received the Bachelor of Commerce degree from the University of Saskatchewan, Canada in 1964, when he was also awarded the Governor General's Gold Medal as Most Distinguished Graduate of his graduating class. He received his Ph.D. in economics from Yale University in 1970. His doctoral dissertation was entitled Physician Productivity and the Demand for Health Manpower. He has received honorary degrees of Doctor of Science from the Medical College of Pennsylvania and from Mount Sinai School of medicine, City University of New York.
Although Professor Reinhardt's research interests since that time have centered mainly on health economics and policy, his work has also included topics in corporation finance, including benefit-cost analyses of the Lockheed L-1011 Tri Star and the Space Shuttle. In 1978, Professor Reinhardt was elected to the Institute of Medicine of the National Academy of Sciences, on whose Governing Council he served from 1979 to 1982. At the institute, he has served on a number of study panels, among them the Committee on the implications of for-profit medicine. He currently serves on the Institute's Committee on Technical Innovation in Medicine and on the Committee on the Implications of a Physician Surplus.
During 1987-90, Professor Reinhardt was a member of the National Leadership Commission on Health Care, a private sector initiative established to develop options for health-care reform and he continues to serve on that body's successor, the National Leadership Coalition of Health Care, co-chaired by former Presidents Carter and Ford. He is past president of the Association of Health Services Research on whose board he still serves. From 1978 to 1993, he has served on the Board of Trustees of the Teachers Insurance and Annuity Association where he has been a member of its Mortgage Finance Committee during the same period. Professor Reinhardt has served on a number of government committees and commissions, among them the National Council on Health Care Technology of the then U.S. Department of Health and Welfare (1979-82) and the Special Medical Advisory Group of the then Veterans Administration (1981-85). From 1986 to 1995 he served three consecutive three-year terms as a Commissioner on the Physician Payment Review Commission (PPRC), established in 1986 by the Congress to advise it on issues related to the payment of physicians. During 1996, he served as a member of the Committee on the U.S. Physician Supply of the Institute of Medicine (IDM) of the National Academy of Sciences,. He currently is a member of the Council on the Economic Impact of Health Reform, a privately funded group of health experts established to track the economic impact of the current revolution in health-care delivery and cost control. In 1996, he was appointed to the Board of Health Care Services of the Institute of Medicine, National Academy of Sciences. Professor Reinhardt was or is a member of numerous editorial boards, among them the Journal of Health Economics, the Milbank Memorial Bank Quarterly, Health Affairs, The New England Journal of Medicine and The Journal of the American Medical Association.

Abstract LS2
Harmonization of Pharmacoeconomic Guidelines in the EU
Wolfgang Greiner, Ph.D.

Until a few years ago, economic evaluations on medical services and pharmaceuticals were seldom conducted in Europe of Germany. Most of the publications on economic evaluation studies came from the US, followed by the United Kingdom, Canada and Scandinavia. Other EU countries were not particularly active in this field. However, this has changed. Although most of the studies are still done in the US, health economic studies are now also conducted in European countries for all major new pharmaceutical innovations. In some European health systems economic evaluations are even needed for price negotiations (e.g.. France and Sweden), or to negotiate the co-payment level (e.g. Italy, Netherlands of Switzerland), or to get drugs listed on a positive list of in treatment guidelines (e.g. Germany and France).
In Germany all the institutions in medical care are preparing for a time when the power of the sickness funds and supplier associations will decline and competition increase. Economic evaluations will then play a much bigger role, because economic considerations will have a larger impact on the decisions concerning resource allocation in medical care. In addition, health policies in the Eu countries are influencing each other. Therefore, drug companies employ an international price (Euro-Pricing) and research strategy.
Most of the EU countries are still "developing countries" with regard to economic evaluation. These studies are often criticized for being biased and, in many cases, not solidly conducted. The scientific quality of economic evaluations is often questioned, and in some cases the critics are right. Many commercial consultants do not meet minimum scientific standards. A precondition for economic evaluation studies to be accepted by decision makers in Europe in the future is for those studies to be of high quality.
The Hannover group published the Hannover Guidelines in January 1995, which have been translated into and published in English and Japanese. Recently, a consensus group published the German Recommendations of Economic Evaluation Studies, which are accepted by most relevant groups. However these recommendations are only a first step and need periodic revision to meet scientific developments and new standards in this field.
Guidelines have also been published in other EU countries (e.g.. Great Britain and Italy). A project financed by the European Commission was started some months ago to harmonize the recommendations as far as possible with respect to questions such as using quality of life measures, discounting rate or measuring indirect costs. Whether guidelines should guide those who commission economic evaluations, those who conduct these studies or those who use them in decision making is an open question. In any case they will guarantee minimum standards and will also make it easier to transfer study methods and study results to other countries, as the different European guidelines take into account the standards already developed overseas (e.g.. in Australia and the US).

Wolfgang Greiner, Ph.D. Biosketch

Wolfgang Greiner is currently senior research fellow at the Centre for Health Economics and Health System Research. He has published a number of articles in scholarly journals of health economics. He has worked in particular on health insurance issues and methodological questions concerning economic evaluations of health services and quality of life measurement. Wolfgang Greiner has been a member of the EuroQol group since 1995 and has led several major health economic projects (on asthma, cystic fibrosis, immunsuppression, liver and kidney transplantation). Before graduating in insurance economics he worked in the international department of a German bank.
The North German Center for Health Services Research (HSR), directed by Prof. Dr. J. -Matthias Graf von der Schulenburg, University of Hannover, is a university-based research institute specializing in economic evaluation of health services and the economics of health insurance. The centre currently has a staff of 24 researchers. It has worked on many projects in this field financed by the State of Lower Saxony, the Federal Research and the Federal Health Departments, health insurers, physicians' organizations and industry.

Abstract LS3
Developing A.P.O.R GAPP (Generally Agreed upon Pharmacoeconomic Principles)
Steven Finder, M.D., M.B.A., M.P.H.

Based upon the membership survey, APOR members believe that current pharmacoeconomic guidelines fail to meet their needs and that APOR should develop agreed upon criteria for the conduct and reporting of pharmacoeconomic analyses.
Pharmacoeconomics lies at a cross-road. Without a clear set of principles that define acceptable standards, pharmacoeconomic research loses validity among its users, yet as an emerging science it needs the freedom to be innovative and creative in developing new methods and approaches. To address this need, A.P.O.R is developing GAPP (Generally Accepted Pharmacoeconomic Principles). GAPP will be a constantly evolving set of principles that will define an acceptable level of standards for pharmacoeconomic research while providing a framework that fosters innovation.
This presentation will present the work of the Consensus Development Committee and A.P.OR.'s road map to GAPP. This presentation will address the following points.
1. Why the need for GAPP.
2. Are previous initiatives enough
3. Why one approach is not enough.
4. The perspectives and methodologies that GAPP must address.
5. The framework for GAPP. How APOR proposes developing GAPP.
6. How everyone can participate.
7. Where are we.
8. What is next.

Steven Finder, M.D., M.B.A., M.P.H. Biosketch

Steven Finder, MD, MBA, MPH, is President of MedAnalysis, a Pharmacoeconomic Research Organization bringing pharmacoeconomics and formulary management support to managed care organizations. Prior to MedAnalysis he was Deputy Director of the Department of Defense Pharmacoeconomic Center (PEC) at San Antonio, Texas, where he was primarily responsible for the development and oversight of cutting edge cost-effective evaluations of specific disease state pharmacotherapy. He developed the Functional Cost Based method for applied pharmacoeconomics being used by the PEC in its current disease state models and Best Value procurement process. While at the PEC he developed a number of disease states evaluations to include diabetes, BPH, osteoporosis, migraine prophylaxis, typhoid and hepatitis A vaccination, PCP infection in HIV, GERD, and candida vaginalis among others, and recently completed an analysis and model of major depression. Many of these evaluations have been presented at national conferences or are being published. In addition, Dr. Finder was elected to the Executive Board of the Association of Pharmacoeconomics and Outcomes Research and is the chairman of the Consensus Development Committee, which is developing standards and guidelines for the conduct and presentation of pharmacoeconomic evaluations. Dr. Finder received his BS in 1976 from the United States Military Academy at West Point. After a short stint as a military officer, he entered medical school at the University of Texas at Houston, where he graduated in 1982. He completed his internship at Walter Reed Army Medical Center, and was assigned duties as a Brigade Surgeon and Commander of a medical company (a mini-MASH), a rare honor in peacetime. Dr. Finder's experience during his two years of command led to his interest in management. He completed a full time MBA at the University of Texas in Austin Graduate School of Business in 1990, where he was recognized as a Sord Scholar. In 1991, he completed an MPH at the University of Texas School of Public Health in San Antonio and in 1992 a residency in Public Health and Preventive Medicine at Madigan Army Medical Center. He subsequently completed several economic evaluations of preventive medical services which led to his recruitment to the PEC in 1994, in an effort to bring management, business and epidemiological expertise to the new field of applied pharmacoeconomics.

Economic and Outcomes Issues in Mental Health

Abstract #: MH1


Russo, L, Curley, C, Yuan, Z, Singer, M
Case Western Reserve University, Cleveland, OH. US

BACKGROUND: Depression may be the most prevalent clinical problem presenting to primary care physicians (PCPs). Common treatment options are selective serotonin reuptake inhibitors (SSRI), tricyclic antidepressants (TCA), and psychotherapy. OBJECTIVE: Assess the cost-effectiveness of 4 treatment protocols for moderate depression; (1) TCA, followed by SSRI if unsuccessful (2) SSRI, followed by TCA (3) Psychotherapy, followed by SSRI (4) Psychotherapy, followed by TCA. METHODS: A decision analytic model, with Markov processes, was developed with 5 treatment arms representing the 4 treatment protocols and no treatment. The states for the Markov processes were: (1) death from suicide (2) death from other causes (3) better on treatment (4) depressed on treatment (5) better off treatment (6) depressed off treatment. The time horizon was 16 six-week cycles. Drug, consultation costs were obtained from hospital and pharmacy sources. Utilities of health states were obtained from a survey of PCPs. Other data were obtained from literature review.. The reference case was a 30 year old white female. A discount rate of 3% was used. Incremental cost-effectiveness (C/E) analysis was performed. The perspective is the health care system ( indirect costs were not available at this time). RESULTS: The incremental C/E ratios for TCA/SSRI and psychotherapy/TCA were $ 9,617.38/QALY and $ 56,637.40/QALY. TCA achieved virtually identical effectiveness as SSRI but at 20% less cost. Results were sensitive to the cost of psychotherapy and the utility of being better on psychotherapy. CONCLUSION: If an incremental C/E ratio of $ 56,637.40 is deemed cost-effective, then psychotherapy followed by TCA is the protocol of choice. Otherwise, TCA followed by SSRI is the best alternative.

Abstract #: MH2


Venturini F, Hay JW, University of Southern California, Los Angeles, CA, USA

Depression has been cited as the most common clinical problem that primary care physicians are called upon to diagnose and treat. Medication have been shown to be effective in all forms of major depressive disorder (MDD) OBJECTIVE: The purpose of this analysis was to evaluate the incremental cost-effectiveness of fluoxetine in comparison with amitriptyline as initial treatment of a MDD episode in primary care, from the perspective of a health maintenance organization (HMO). METHOD: A simulation model based on the theory of clinical decision analysis was constructed. The duration of the analysis was nine months, the average length of a MDD episode. Patients included were those without history of depression. The direct costs included in the model were: drug acquisition cost, outpatient medical care, laboratory tests and procedures, and hospital care. Costs were obtained from public sources and were expressed using 1996 dollars. Two different measures of effectiveness were utilized: years of life saved suicide related, and rate of treatment success. Treatment was considered successful if there was no need of withdrawal or switch to the alternative treatment, and no need to refer the patient to a psychiatrist. The efficacy and safety data, and the suicide figures were derived from published literature. Univariate sensitivity analyses were performed. RESULTS: The estimated incremental cost per year of life saved with fluoxetine treatment was $128. The estimated incremental cost per successful event was $1,446. The results were mainly stable when a sensitivity analysis was applied to the variables employed in the model. CONCLUSIONS: Despite its higher acquisition cost, fluoxetine was found to be a cost-effective treatment for MDD. These results are consistent with most of the available literature, and they should be taken into account by HMOs administrators in the formulary process and in the development of treatment guidelines.

Abstract #: MH3


Melfi, CA, Croghan, TW
Lilly Research Laboratories, and Indiana University School of Medicine, Indianapolis, IN.

Mentally ill Medicaid recipients represent a population which may be particularly vulnerable to limited access to adequate treatment their mental illness. OBJECTIVE: The purpose of this study is to compare depressed Medicaid recipients to those with private medical insurance, and to examine racial differences in the treatment of depression in the Medicaid recipients. METHODS: Data for the privately insured are derived from a database of claims for enrollees in a variety of employer-provided health plans. The Medicaid data are from a single state plan. We divide each dataset into people who have a claim indicating a depression diagnosis, but no prescription for an antidepressant; and people who have a diagnosis of depression and at the same time fill a prescription for an antidepressant. Comparisons are made using t- tests and chi-square tests of significance. We compare groups based on treatment received as well as patterns of use of antidepressants. RESULTS: The privately insured patients who are treated with antidepressants are much more likely to receive the newer selective serotonin reuptake inhibitors (SSRIs) rather than the older tricyclic antidepressants (TCAs) relative to Medicaid patients. Within the Medicaid group, African Americans are more likely to receive TCAs than are white Medicaid patients. Depressed privately insured patients are more likely to receive psychotherapy than are depressed Medicaid patients. There is also a much higher rate of continuous therapy on initial antidepressant in the privately insured group. CONCLUSIONS: Our results indicate that access to quality mental health care is significantly restricted among depressed Medicaid recipients. In addition, within depressed Medicaid patients, there are racial differences with regard to care and treatment access.

Abstract #: MH4


Wiener, MB, Kearney, K, Beck, A, Shedler, J, Kaiser-Permanente, Denver, CO, USA

The costs of psychiatric disorders are significant. It is estimated that depression costs Americans $44 billion in treatment costs, decreased productivity, lost income, and suicide. Depression and anxiety are among the most commonly seen disorders in primary care, but often go undiagnosed and untreated. It is estimated that 60 percent of patients with diagnosable psychiatric disorders are seen in primary care settings, termed the de facto mental health care system. There is a need for effective screening of depression and other psychiatric disorders in primary care settings. As of now, no psychiatric screening tool has gained real acceptance among primary care providers. METHOD: A protocol was developed which utilizes an automated, computerized screening procedure to diagnose depression and anxiety disorders in a large group HMO. The screening procedure made use of a computerized quality of life questionnaire (the SF-12), which included basic depression and anxiety screening items. When patients gave responses that indicated possible problems, the test branched automatically to the Shedler Quick PsychoDiagnostics Panel (QPD Panel), an automated test designed to diagnose a range of psychiatric disorders. The QPD panel provides physicians with a specific DSM-IV diagnosis, a numerical "severity score" indicating the patient's current level of distress, and a listing of the symptoms that gave rise to the diagnosis. Patients who met diagnostic criteria for a psychiatric disorder were treated according to the existing practice guidelines for the HMO. CONCLUSION: Of 200 patients screened, less than 1% indicated that the tool was inappropriate or declined to participate due to the confidential nature of the questions. Patient and physician satisfaction with the QPD Panel was near 100%.

Abstract #: MH5

Schizophrenic Patients' Views of Antipsychotic Medication

Conner TM, Rascati KL. The University of Texas at Austin, USA

Drug therapy for schizophrenia is costly and often produces bothersome side effects. However, noncompliance to therapy often results in rehospitalization, further increasing the cost to treat the illness. Several newer antipsychotic medications recently have become available which report lower incidence of side effects, and thus greater compliance. The need for outcomes research, including quality of life measurement and pharmacoeconomic analyses, to compare the drugs available is therefore evident but often challenging. OBJECTIVES & METHODS: As part of a larger study examining a quality of life questionnaire, we sought to assess treatment at an outpatient clinic from the patients' perspective. Forty- eight patients agreed to be interviewed or to self-complete a questionnaire. Of these patients, thirty-nine could provide coherent responses to questions. RESULTS: Patients reported they almost always took their antipsychotic medication as prescribed. Generally, patients reported their symptoms were somewhat to much better since starting medication. While on average, patients reported they felt happy, compared to how they felt before they began taking medication, 25% said they were unhappy. Surprisingly, over half the patients reported they experienced no side effects from medication. Thirty-five (almost 90%) patients correctly reported that they took medication and of these, most could name the drugs they were taking. However, six (15%) patients reported there were no benefits to taking their medicine. Several patients expressed a desire to stop taking or to decrease their medication, even if they knew of benefits for taking it. CONCLUSION: Most patients who were surveyed were knowledgeable about their medication and understood that it was important to continue antipsychotic treatment. The implications of these findings will be discussed.

Abstract #: MH6


Shani S., Haran B., Lev B.
Ministry of Health, Jerusalem, Israel.

The national Health Insurance Las was implemented in Israel in 1995. The law states that each citizen is entitled to receive health care treatments of established quality. Pharmaceuticals included in the list of health care treatments are reimbursed by the Sick Funds (HMOs). When the law came into effect, the list of reimbursed pharmaceuticals was defined as the 1994 formulary of Israel's largest Sick Fund. Since 1994, this list has not been updated although numerous new molecules have been registered. The Pharmaceutical Administration in the Ministry of Health set up a mechanism for updating the national list of reimbursed pharmaceuticals using pharmacoeconomic analysis as a tool for decision making. OBJECTIVE: Risperidone, a new antipsychotic drug indicated for schizophrenia was chosen as a test case for the proposed mechanism. METHODS: Data collected from a review of the literature and local data were used to perform a cost effectiveness analysis comparing risperidone with neuroleptics and clozapine. RESULTS: Both risperidone and clozapine are more effective than neuroleptics. The cost of treating a schizophrenic patient per year in Israel (in 1996 US dollars) is 200$ for neuroleptics, 1850$ for clozapine (including blood sampling) and 2550$ for risperidone. Risperidone is not more cost-effective than neuroleptics when used as a first line treatment for schizophrenia. Hence, risperidone should not be reimbursed for this use. Risperidone is cost-effective when used for treating schizophrenics non responsive to neuroleptics, though less than clozapine. Due to quality of life considerations, risperidone should be used before clozapine for treating these patients. CONCLUSIONS: The cost-effectiveness analysis let to a suggested clinical treatment guideline for reimbursing risperidone. This test case contributed to the establishment of a mechanism for reimbursing new pharmaceuticals.

Desease State Management: Episodes, Outcomes, and Cost of Care

Abstract #: CC1


Diamond LH, Gibson TB, Houchens RL, The MEDSTAT Group

OBJECTIVES: This presentation describes the development of a computerized inpatient/outpatient severity-adjusted episode of care grouper based upon the Disease Staging severity of illness system. This episode grouper extends the unit of analysis for cost and utilization reporting beyond the inpatient admission and outpatient claim level to link together all recorded services delivered for treatment of a specific disease in all settings (i.e., inpatient, outpatient) by all types of providers. METHODS: Episode creation requires classification of diseases into homogeneous units to identify claims that are clinically related to each other, and logic to link together related services. Disease Staging, a clinically-based disease classification system, serves as a framework by defining each condition and its severity levels. To link together related services, different approaches were used to construct episodes for acute and chronic conditions. While episodes for chronic conditions are well-defined, episodes for chronic conditions can be continuous or recurring, and the disease progression and underlying maintenance aspects of chronic conditions are reflected in the episode construct. In addition to clinical input, a national database of private pay medical claims was utilized to develop, test and verify episode logic. Episodes for each disease were developed separately, making certain to balance clinical rigor with feasibility of implementation. RESULTS: Using 1993 and 1994 data from the national database of medical claims described above, a comparison of treatment patterns and episode measures (cost, and use) for one chronic and one acute disease, diabetes and bacterial pneumonia, are presented. In addition, risk-adjusted, episode-based comparisons of populations and patients in different geographic regions are included. The application of these episodes for evaluating clinical performance measures will also be presented.

Abstract #: CC2


Fox, KM, Epstein, R, Suppapanya, N, Sutphin, M, Merck-Medco Managed Care, L.L.C., a subsidiary of Merck & Co., Inc. Montvale, NJ, US

Diabetes mellitus affects 16 million Americans and leads to long-term complications of retinopathy, neuropathy and nephropathy. Programs of patient education and self- management should achieve better glucose control thereby improving short and long-term outcomes. OBJECTIVE: The purpose of this study was to determine the economic outcomes of a patient-centered disease management program for diabetes. METHOD: A pre versus post comparison was conducted to examine the differences in health care utilization and associated costs between the year prior to the disease management program and the year during the program intervention. Subjects were insulin-taking diabetics who were identified through drug claims. 1,635 patients from one health care plan were continuously enrolled in the program from September 1994 through August 1995 and included in the analysis. Key components of the program included a personalized letter, quarterly newsletters, a risk assessment survey, and a toll-free support line. Medical claims data were used to track resource utilization and costs. RESULTS: All-cause hospitalizations were reduced by 14% (p<0.05) during program participation leading to a to $522,786 cost savings per 1,000 patients. All- cause emergency room visits increased by 9.5% while outpatient and physician office visits did not change significantly. A $457,396 per 1,000 patients cost reduction was found overall. For diabetes-specific resource utilization, a 10% decrease in hospitalization, 20% decrease in emergency room visits, 56% decrease in outpatient visits and 8% decrease in physician office visits was demonstrated. The diabetes- specific utilization reduction lead to an overall cost savings of $214,968 per 1,000 patients in diabetes costs. Drug therapy utilization did not change from pre to post program. CONCLUSION: These findings indicate that patient education targeted to improving self- management of diabetes lowers the cost of care by $458 per diabetic patient per year.

Abstract #: CC3


Zbrozek AS, Whalen E, Smith ME, Bayer Corporation, West Haven, CT.

Non-insulin-dependant diabetes mellitus (NIDDM) is known as a costly disease. The cost effect of initial drug therapy has not been previously reported.

OBJECTIVE: We wanted to understand the marginal effect of early diabetes treatment on direct medical costs. METHOD: An integrated pharmacy and medical claims data base was used to extract cohorts. Patients having a NIDDM ICD-9-CM code and an NDC code for any oral antidiabetic or insulin prescription were initially captured. Index year 1992 was used for all initial prescriptions and all patients were continuously enrolled in non-fee-for-service health plans. Patients had continuous eligibility for 24 months after the prescription index date. Patients had 6 months of continuous eligibility prior to index date without receipt of any antidiabetic drug during this period. Mean costs per month were tabulated during pre and post index periods in both cohorts. RESULTS: Patients aged 45-64 years (N=634) at end of post-index period averaged total costs (and diabetes-related costs) of $7692 ($2538)/year and $6360 ($1200)/year in the first and second post- index years, respectively. Patients 65 years and older (N=723) at pre-index averaged total (and diabetes-related) costs of $6408 ($984)/year and $6354 ($1074)/year during the first and second years since index, respectively. Mean annual total costs during the pre-index period were estimated at $3072/patient and $4248/patient in the respective cohorts. CONCLUSIONS: Patients with NIDDM incurred high mean direct medical costs prior to drug therapy onset. Mean direct medical costs after drug therapy onset were elevated in both cohorts. These elevations were explained in part by costs coded for NIDDM therapy or outcomes. During each post-index period, mean diabetes-attributed costs were proportionately lower than aggregate costs attributed to non-diabetes conditions.

Abstract #: CC4


Huse, DM, Richner, RE, Piercey, G, Hartz, SC, Medical Research International, Burlington, MA, USA, and SmithKline Beecham Pharmaceuticals, Philadelphia, PA, USA.

OBJECTIVE: To monitor asthma treatment, outcomes, and costs in managed care. METHOD: The Asthma Outcomes Registry is a longitudinal study of more than 2,500 patients aged 5-77 years who had received medical treatment for asthma within 18 months prior to identification at three U.S. managed-care organizations. Symptoms of asthma are assessed at baseline and then at six-month intervals, using mailed questionnaires. Questionnaires are completed by a parent or guardian for subjects under 12 years of age, and otherwise are self-administered. Prescribed therapy, test results, and the occurrence of acute exacerbations are documented through medical chart review. Data on health-care utilization, including prescriptions dispensed, are obtained from health insurance claims or similar administrative databases. To date, utilization and cost data have been collected and analyzed for 1,145 subjects at two sites, covering the 365-day period prior to each subject's date of enrollment. RESULTS: Mean annual costs of asthma treatment, including inpatient and outpatient medical care and prescription drugs were $623 per patient, including $68 for hospitalization, $228 for outpatient medical care, and $327 for prescription drugs. Total annual costs increased with age from $442 in asthmatics aged 5-15 years to $930 in those >50 years of age. Costs were also somewhat higher in females ($691) than males ($522). These results are higher than previously reported estimates of the annual cost of medical care for asthma. CONCLUSION: Experience to date shows the Registry to be a valuable tool for studying the both the epidemiology and economics of asthma in managed care.

Abstract #: CC5


Shih, Y-CT and Kauf, TL

UNC School of Pharmacy, Chapel Hill, NC, USA

OBJECTIVE: This study examines the impact of EPO coverage in terms of health care utilization and expenditures. Medicare expenditures for ESRD beneficiaries are decomposed into three parts. "EPO-direct" is the Medicare reimbursement for EPO administration. "EPO-indirect" is additional expenditure due to complications or savings from improved health conditions. "EPO-unrelated" is for covered care unrelated to EPO. If EPO coverage is cost-effective for Medicare, then "indirect" savings should outweight "direct" and "indirect" cost, and utilization due to anemia and related conditions should be reduced. METHODS: Several HCFA data sets providing claims information for Medicare beneficiaries are used. Claims made during the two year period before and after coverage of EPO were examined for dialysis-dependent ESRD beneficiaries. Using the pre-EPO beneficiaries as the control group, we use multivariate regression and share equations to study the variation in program expenditures as a result of EPO coverage. We also use logit to estimate changes in the probability of hospital admission and a count data model to investigate utilization variations. Powe et al (1993) showed an increase in per capita expenditures of ESRD patients over the 5 year post-EPO coverage period, though the magnitude of that increase was less than estimated previously. However, our estimation from expenditure decomposition shows that indirect savings outweigh the direct and indirect cost of EPO, and the expenditure increase observed above is mainly from "EPO-unrelated" expenditures. Moreover, medical complications as a result of anemia decrease after EPO coverage. CONCLUSION: This study provides claims-based evidence that EPO coverage has been cost-saving for Medicare. Our results also underscore the importance of distinguishing between variation caused by general trends vs. the policy under investigation. This is possible only with the use of claims-based data.

Abstract #: CC6


Dooley, JA, Dooley, GJ, Langley, PC, University of Arizona, Tucson, AZ, US

Cytomegalovirus (CMV) retinitis currently affects 30-40% of patients with end stage AIDS. Constantly evolving treatment approaches and limited long-term clinical data indicate a modeling approach to produce estimates of lifetime treatment patterns suitable for pharmacoeconomic evaluation. Markov modeling techniques are widely used to model disease prognosis and outcomes of therapy and clinical data is usually abstracted to a fixed probability for transitions between model states. OBJECTIVE: The purpose of this study was to develop disease models for the alternative therapeutic approaches for CMV retinitis and to evaluate the effect of using variable probability transitions based on all available clinical data rather than fixed probabilities. METHOD: A seven stage Markov model was developed representing all possible treatment states. Probabilities were assigned to the transitions between states using standard methods to convert clinical data to fixed probabilities. The model was evaluated using Monte Carlo techniques for the three primary therapies (IV Ganciclovir, IV Foscarnet and Oral Ganciclovir). Some transitions were recalculated as time-dependent probabilities to more accurately reflect the available clinical data and the models were reevaluated. RESULTS: The Markov models produced clinically plausible estimates of the average lifetime treatment state duration. Using time-dependent probabilities significantly changed the time spent in different treatment phases, the overall survival duration was reduced by an average of 41%. CONCLUSION: Markov models can be used as the basis for evaluation of CMV retinitis therapies and the approach is sufficiently flexible to accommodate new therapeutic developments. The significantly changed time in state estimates resulting from the use of time-varying probabilities indicates that caution should be exercised when clinical data is reduced to fixed probabilities to simplify the modeling process


Economic and Outcome: Issues in Pharmaceutical Care

Abstract #: PC1


Portner, TS, Gourley, DR, Gourley, GA, Solomon, DK, Bass, GE,
University of Tennessee College of Pharmacy, Memphis, TN, US

A multicenter outcomes study was conducted in 10 VA medical centers and 1 academic medical center across the United States. OBJECTIVE: The purpose of the study was to evaluate the impact of pharmaceutical care on select clinical, economic, and humanistic patient outcomes in COPD and HTN patients in ambulatory care environments. METHOD: Clinic patients treated for a diagnosis of HTN or COPD were randomly assigned to a treatment group which received pharmaceutical care or a control group which received standard care over a four month study period. Clinical pharmacists and pharmacy residents worked together to conduct the protocols, focusing on the management of COPD and HTN relative to physiologic measures, patient adherence to therapy, drug knowledge, utilization of resources, and functional status. RESULTS: For the HTN study arm, there were 153 evaluable patients (76, treatment and 77, control) and 115 evaluable patients in the COPD study arm (54, treatment and 61, control). For the hypertension study arm, treatment group patients made greater improvements over the 4 month study period than did control group patients on measures of systolic blood pressure, knowledge of disease and treatment, satisfaction with care, and self assessed adherence to medication regimens (p < 0.05). For the COPD study arm, treatment group patients made greater improvements over the 4 month study period than did control group patients on patient ratings of symptom interference with activities and dyspnea measures, ratings of bodily pain, satisfaction with care, and self assessed adherence to medication regimens (p < 0.05). CONCLUSION: Findings indicate considerable contribution of pharmacists to patients' improvement in hypertension and COPD disease states.

Abstract #: PC2


McMillan, KS, Gourley, DR, Portner, TS, Gourley, GK, Solomon, DK,
University of Tennessee College of Pharmacy, Memphis, TN. US

OBJECTIVE: The purpose of this pretest–posttest control group study was to evaluate clinical and humanistic outcomes in patients enrolled in a pharmacist–directed lipid management program in the community pharmacy setting. Outcome measures included total cholesterol, LDL–cholesterol, HDL–cholesterol, and triglyceride levels; patient satisfaction with pharmaceutical care; patient knowledge of hyperlipidemia; and patient willingness to pay for the service. METHODS: Patients at risk for coronary heart disease (CHD) or with known CHD identified from a search of the pharmacy prescription database and walk–in patients were invited to participate in a wellness and cholesterol screening. Patients found to have elevated LDL–cholesterol or triglyceride levels and at least one risk factor according to National Cholesterol Education Program (NCEP) guidelines were asked to participate in the study. Fifty–two subjects were randomized to a treatment (n=26) or control (n=26) group. The treatment group was followed in a 6 month pharmacist–directed lipid management program and the control group referred to their healthcare provider for care. The intervention was pharmacist provision of disease management and monitoring and education to patients with hyperlipidemia. Education involved instruction in lifestyle modification including diet and exercise therapy, the role of cholesterol in the development of CHD, and counseling on drug therapy. The pharmacist worked in collaboration with treatment patients' physicians to design an appropriate therapeutic regimen based on NCEP guidelines. RESULTS: The study will be completed by the end of February with final results available in April 1997. CONCLUSION: This project will serve as a model for community pharmacies for the establishment of lipid disease management programs.

Abstract #: PC3

An Application of the Time-Dependent Cox Regression Model in the KAISER/USC Pharmacists' Consultation Intervention Study

Yuan, Yong,1 Hay, Joel,1 Groshen, Susan2
1Department of Pharmaceutical Economics & Policy , and 2Preventive Medicine, University of Southern California

OBJECTIVE: The Kaiser Permanente/USC Patient Consultation (PC) study aims to analyze the impact of the three different models of pharmaceutical care on patient outcomes and quality of life. To evaluate whether pharmacist consultation has a favorable impact on drug therapy that improves patient survival, we consider a model for survival analysis that include multiple time-dependent indicator variables to predict the effect of pharmacy intervention on mortality rate. However, in the PC intervention sites, many patients switched within the model after initial randomization. We needed a strategy which takes account of intervention actually received. METHODS:The number of new prescriptions filled at PC intervention sites was introduced into Cox Models as intervention variables in order to measure the extent to which the patient was exposed to each model of pharmaceutical care. The baseline of comparison is the Control Group, which was not exposed to any formal PC intervention. Selection bias is then a potentially serious problem. The Adjusted Wu- test was used to test the model for endogeneity, if there is endogeneity and the Two-Stage method will be then performed. RESULTS: The "As treated " analysis didn't outperform the "Intent to treat" analysis (ITT). It is not surprising that "As treated " and ITT analysis defined totally different target population which showed the statistically significant intervention effects. However, the ITT analysis alone is not appropriate to answer the explanatory questions; it must be supplemented by analyses taking into account the treatments actually received. CONCLUSION: In such settings, one can use the Cox model for inference about the relative risk in such cases by using multiple time-dependent covariate structure to specify a continuous functional form for the relative hazard as a function of time since switch. We conclude that the time dependent model predicts better than our previous time-fixed model.

Abstract #: PC4


Munroe, WP, Kunz, K, Dalmady-Israel, C, Potter, L, Schonfeld, WH, MedOutcomes, Inc., Richmond, VA, US, and Technology Assessment Group, San Francisco, CA, US

To improve both health and economic outcomes for patients, an intervention program was started at three retail pharmacies. The interventions performed by specially trained pharmacists included targeted patient education, systematic patient monitoring and feedback, behavior modification, and regular communication with patients' physicians to allow early intervention for drug related problems. OBJECTIVE: The study was conducted to assess the economic impact of these patient-focused pharmacy interventions in the community retail setting among patients with hypertension, diabetes, asthma and/or hypercholesterolemia. METHOD: Prescription drug costs and total medical costs were analyzed by comparing claims data from 188 patients enrolled in the intervention program at the three participating pharmacies to data from 400 matched control patients at five non- participating pharmacies. All of the pharmacies were part of the same retail chain, and all patients used the same health care insurer. RESULTS: For all disease states, the average cost per prescription was significantly higher in the intervention group than in the control group (p<.001 for hypertension, asthma, and hypercholesterolemia; p=0.045 for asthma). However, differences in monthly prescription costs were only significant for patients with asthma, with monthly costs being higher in the intervention group. With regard to total monthly medical costs, substantial savings were demonstrated in the intervention group across all cost analyses. Savings ranged from conservative estimates of $144 per patient per month to $294 per patient per month when accounting for the possible influence of age, comorbid conditions and disease severity. CONCLUSION: Pharmacist interventions in this community pharmacy-based disease management program were associated with substantial reductions in monthly health care costs to patients with hypertension, hypercholesterolemia, diabetes and/or asthma.

Abstract #: PC5


Nol-Hiering L, Gause DO, Nathewitch A, Menz JM, Smith WP Novartis Pharmceuticals, Summit, NJ, Geisinger Health Plans, Danville, PA, USA

To contain healthcare expenditures, employer groups are placing presure on insurance carriers to minimize costs. Due to the misconception of the pharmacy benefit asa non-essential component of the healthcare delivery system, these programs are often targeted for cost containment. OBJECTIVE: The purpose of this study was to compare medical utilization between MCO members with or without pharmacy benefit. METHOD: A retropsective analysis of medical claims during the period July 1, 1995-June 30, 1996 was conducted utilizing data from Geisinger Health Plans (GHP), an MCO with approximately 180,000 members. Of these members, approximately 150,000 members have pharmacy benefits while approximately 30,000 do not. A random sample of patients from each group who had at least one medical claim during the 12 month period was obtained. The regression analysis was performed adjusting for age group, sex, and select illnesses as indicated by associated diagnostic codes (ICD-9-CM classification scheme). RESULTS: Patients without benfit expended annually an adjusted mean of $7034 while patients with benefit expended an adjested mean of $4778 on medical utilization. The number of medical claims also differed with an adjusted mean of 49.7 claims for paitents without benefit compared to an adjusted mean of 39.1 claims for patients with benefit. The potential annual savings for patients with pharmacy benefit unadusted by employer costs in sponsoring the program is $2256 per patient. CONCLUSION: Preliminary results demonstrate the value and potential savingss to employers providing pharmacy benefits. Further anlysis will adjust for income and incorporate employer sponsored program costs.

Abstract #: PC6


Rahman, A, Dighe, MS, Rappaport, HM. School of Pharmacy, Northeast Louisiana University, Monroe, LA.

Epilepsy is a disease state associated with severe disability, and limitations. The incidence of epilepsy in United States is less than 1%. However it represents a severe socioeconomic burden on the society. Drug therapy is the major method of controlling this condition, and is also the most costly category of service. In this era of cost containment and formulary restrictions, it is important to identify the factors associated with medication utilization for condition such as epilepsy. The data for the purpose of this study were obtained from the 1992 National Ambulatory Medical Care Survey (NAMCS). Data on eighty- two patients with a diagnosis of epilepsy (ICD-9- CM 345 to 345.91) were extracted using SETS version 1.22 and analyzed using Statistix version 4.1. Of these, the majority of patients received health coverage through either an HMO(17%), Medicaid(18%), Medicare(19%), or Commercial Insurance (36.6%). An ordinary least squares (OLS) regression model was developed to identify the role of demographic, comorbid, diagnostic, patient disposition and insurance coverage factors on medication utilization. The results of the statistical model suggest a significant positive slope in medication utilization. However it is further observed that there were a fewer number of medications prescribed to patients with Commercial Insurance (p=0.02). None of the other demographic, comorbid, diagnostic, and patient disposition factors were significant. The regression model explained 24.97% of the variance among the hypothesized relationships. The results of this study may hold important policy implications on the role of insurance status and drug prescribing. It suggests that Commercial Insurers may have better administrative control on drug prescribing. More studies of this type should be done in the future using large commercial health insurance databases.

Patient Preferences & Analytical Methods

Abstract #: AM1


Poon, AW, Gaylin, DS, Shapiro, JR, Mendelson, DN, Rubin, RJ, The Lewin Group, Fairfax, VA, US.

OBJECTIVE: We developed novel WTP methodology to assess the comparative costs and benefits of ostomy pouches; closed-end pouches have advantages in convenience, security, and comfort but can be more expensive than drainable pouches. METHOD: Through a survey of 90 colostomates and 100 Enterostomal Therapy Nurses, we analyzed the importance of 20 fundamental pouch attributes. The most important attributes were grouped into five categories based on physical or practical differences between closed-end and drainable pouches. 55 colostomates were surveyed about their willingness to pay for each of these benefits. The average WTP values for the five categories and retail pouch costs were inputs into a cost-benefit model, which calculates net benefit, accounting for individual differences in pouch use. RESULTS: The typical colostomate had a WTP of $34/month for the benefits that result from switching to closed-end pouches from drainable; the typical costs of switching were $24/month, yielding an average net benefit of $10/month. CONCLUSIONS: Relative to monthly pouch costs of approximately $60/month, a net benefit of $10/month is substantial. Closed-end pouches, therefore, should be the preferred product under most circumstances, even after considering differences in cost. WTP surveys are a theoretically preferred method of measuring the benefit of medical interventions, but have posed problems in application because of the difficulty many individuals have in valuing abstract health benefits, such as a risk reduction in mortality. This analysis demonstrates the applicability of WTP methods for evaluating benefits of a medical device that causes concrete, tangible changes in QoL.

Abstract #: AM2


Mucha, LM, Foreman, SE, Penn State University, Department of Health Policy and Administration, University Park, PA. US.

Estimating the demand for prescription drugs is a complex issue. While insurance coverage for prescribed medicines is important, it is not the only factor that impacts demand. OBJECTIVE: The purpose of this study was to build on prior drug demand studies by adding another piece of the puzzle: patient attitudes. We tested to see how patients' attitudes towards their physician and medicine impacts the decision to fill prescriptions. METHOD: Data from the 1987 National Medical Expenditure Survey were used to estimate a two part model of demand. The first part used logistic regression to estimate probability of use of drugs (N=25599). The second part used OLS regression on a semilog model to estimate level of drug use conditional on use (N=15185). The two attitudinal measures of interest were: (i) respondent agrees home remedies are better than prescribed medicines, and (ii) respondent agrees s/he can get well without the aid of the physician. The drug measure indicates how a patient perceives the effectiveness of drugs. The physician measure indicates the effect of the level of trust a patient has in the physician on the decision to fill a prescription. RESULTS: Results showed that, ceteris paribus, respondents who had favorable attitudes towards physician and drugs had a significantly higher probability and level of use of drugs than respondents who did not . Odds ratios from part one for both the drug (0.85) and physician measure (0.85) were significant (p.<.001). The same held true for the coefficients in part two (9% and 12% significantly [p.<.001] less drug use, respectively). CONCLUSION: Attitudes are as important predictors of drug demand as health status and insurance coverage are.

Abstract #: AM3

LINKING Patient's preferences and TREATMENT OUTCOMES in allergy:


Ricard, N. Health Economics Department Hoechst Marion Roussel Canada Research Inc, Laval, Quebec, Canada

Kind, P. Centre for Health Economics University of York, York, England.

Empirical evidence indicates that involving patients in decision making about their care improves their compliance with treatment, their perceptions of quality, and their outcomes. Such involvement can be achieved by explicitly monitoring patient's preferences particularly in regard to symptoms that impact on their quality of life. Methods of measuring patient-centered preferences do exist but no gold standard has been identified in this area. OBJECTIVE: To investigate the relationship between patient's preferences for relief of symptoms in allergy and efficacy/outcome following treatment. METHOD: In a clinical study of a new therapy, 400 patients rated their preferences for the relief of allergy symptoms prior to treatment. Patient recorded on a 10-point rating scale the level of desirability associated with the relief of sneezing, rhinorrhea, itchy nose, palate and/or throat, and of itchy, watery, red eyes. A preference scaling factor relating patient's impairment to each allergy symptom was recorded on a 10 centimeter feeling thermometer. Treatment outcome was determined by the change in allergy symptom score over a 7-day period. Treatment outcome was analyzed with respect to patient's preferences, and the functional equation linking these variables was used to determine an adjusted symptoms severity score. RESULTS: Patient's preferences will be presented and discussed with respect to the derived functional equation. CONCLUSION: Patient's preferences for symptomatic relief in allergy are important elements in determining outcomes to treatment, and can provide new information to assist decision making in selecting treatment of allergy.

Abstract #: AM4

Validation of the full and reduced batterY of scaleS for measuring the qol of patients ON ANTI-HYPERTENSIVE THERAPY

Nelsen, LM., Himmelberger, DU, Budd DW, Liss, CL. Markson, LE. Merck & Co., Inc., West Point, OA and Health Outcomes Group, Palo Alto, CA

Objectives: To examine the performance characteristics of a full and reduced battery of scales (BOS) designed to measure treatment- related QoL in hypertensive patients.

Methods: The BOS consists of 43 questions representing 6 domains (psychological general well-being, sleep disturbance, sexual function and cognitive function). Data included baseline and three one month timepoints following for 53 "stable" patients (blood pressures well- controlled with anti-hypertensive medication), 90 patients "changing" medications due to treatment-related side effects, and 84 "newly treated" patients. Performance characteristics examined included test/re-test reliability in "stable" patients, construct validity and responsiveness in "changing" and "newly treated" patients, and internal consistency in the "stable" and "changing" patients. Items to delete were identified using "Merit", developed to quantify the ability of a question to identify meaningful change in QoL over time.

Results: The performance characteristics of the full BOS indicated high levels of internal consistency (.74-.95); and acceptable reliability (ICC of .46 to .85). At 3 months compared to baseline, all domains demonstrated significant responsiveness. At 3 months, the change in all domains, except for sexual function, were significantly correlated with change in global QoL. Using the Merit Method, the full BOS was reduced from 43 to 21 questions. The performance characteristics of the reduced BOS did not differ significantly from the full BOS. External validation of the reduced questionnaire using data from two Phase III trials provided results similar to those from the full BOS.

Conclusions: The data from this study confirm that the performance characteristics of full and reduced batteries of scale are satisfactory for measuring QoL in hypertensive patients.

Abstract #: AM5


Berzon RA, Reilly M, Whalen E, Graham E, Smith ME. Bayer Pharmaceutical Division, West Haven, CT and Reilly Associates, Waccabuc, NY

Dementia is a chronic syndrome that is manifest by a progressive deterioration of cognition, function and behavior. Alzheimer's disease is the most common form of dementia in the United States. It and related dementias affect approximately four million U.S. residents.

It is widely recognized that Alzheimer's disease has a profound effect on families and close friends who devote time to caring for the patients. These dementia caregivers are burdened by emotional stress and anxiety that can be long-term and severe. We hypothesized that by intervening in the treatment of patients with probable Alzheimer's disease that these and other detriments to caregiver health-related quality of life (HRQoL) and time allocation could be moderated.

Objective: To investigate the effects of metrifonate, an acetylcholinesterase inhibitor, on caregiver HRQoL and time allocation.

Methods: This was a 26-week, randomized, double-blind, parallel group, placebo-controlled safety and efficacy study conducted at 25 sites by the Bayer Pharmaceutical Division. A total of 408 patients between 45 and 90 years of age with probable Alzheimer's disease of mild to moderate severity entered the study. A total of 357 caregivers completed HRQoL and time allocation surveys. Caregiver burden was assessed via the Screen for Caregiver Burden (SCB), the cognitive subscale of Poulshock and Deimling (1984), and the 17-item Mental Health Index (MHI-17) of the Medical Outcomes Study. Caregiver time use was measured with the Caregiver Activity Time Survey (CATS).

Discussion: The methodology by which HRQoL and time allocation data are collected and managed for caregivers for those persons with probable Alzheimer's Disease of mild to moderate severity is presented. Practical issues regarding questionnaire administration to subject surrogates are reviewed, as are modes of questionnaire delivery.

Abstract #: AM6


Vermeulen , LC, Windisch, PA, Vlasses, PH, Matuszewski, KA, University HealthSystem Consortium, Oak Brook, IL. US

OBJECTIVE: The study objectives were to assess the comparative effectiveness of granisetron, ondansetron, and other antiemetics by evaluating the functional health status; the degree of nausea and emesis rates; and use patterns in patients receiving cancer chemotherapy, in U.S. academic health centers. METHODS: A prospective, observational study was conducted in 14 hospitals. This study evaluated the use of antiemetics in hospitalized cancer chemotherapy patients. The primary clinical outcomes of interest measured were the impact of adverse effects to chemotherapy on the functional status of patients, using a validated instrument (MFLIE), and the occurence of post-treatment vomiting. RESULTS: Data was collected on 439 adult patients. The most common types of cancer chemotherapy regimens used consisted of cisplatin, paclitaxel, etoposide, and cyclophosphamide. Common antiemetics used for treatment of chemotherapy-induced emesis were ondansetron, dexamethasone, lorazepam, and granisetron. A total of 323 (74%) patients reported no episodes of emesis, and 35 (8%) patients reported one episode of emesis. Granisetron and ondansetron were associated with better overall outcomes (e.g., no vomiting, functional health status) for high and moderate emetogenic chemotherapy than standard antiemetics or no antiemetic therapy. In contrast, standard or no antiemetic therapy produced comparable outcomes to granisetron and ondansetron for low emetogenic chemotherapy. CONCLUSION: Both granisetron and ondansetron appear to be comparable in achieving positive antiemetic functional and clinical outcomes. The study data also suggests that 5-HT3 antagonist antiemetics be used in high emetogenic chemotherapy because of the benefit of positive outcomes. Non-5-HT3 antagonist antiemetic therapy or no antiemetic therapy achieved positive outcomes in patients receiving low emetogenic chemotherapy.

Comparative Outcomes and Cost-effectiveness Analysis

Abstract #: EA1


Mallick R, Pracon, Reston, VA. USA

OBJECTIVE: To examine the costs and benefits of universal calcium supplementation to prevent osteoporotic hip fractures. METHODS: The perspective is that of Medicare since 75% of hip fractures occur among aged enrollees. Data sources included the 1992 National Hospital Discharge Survey, and 1995 Medicare data on volume and payments for DRGs 210, 211, and 236. A MEDLINE search identified published studies of relative risk estimates of the association between calcium intake and hip fractures and surveys of daily calcium intake among aged US women. We calculated the percent of aged enrollees who have suboptimal calcium intake and projected the potentially preventable percent of hospitalizations and payments for treating hip fractures if the at risk population received optimal calcium supplementation for over one year. Potential dollar savings were discounted by 3%. Retail prices for calcium supplements in suburban Washington were used. Per capita net benefits of calcium supplementation were calculated by subtracting the cost of supplements and dividing by the size of the target population. RESULTS: In 1995, over 164,000 hip fractures occurred among aged Medicare enrollees. Medicare paid hospitals $1.2 billion for these fractures. Only 5% of aged American women met the US RDA of 1000 mg of calcium per day. If all aged women met this requirement, 38% to 41.7% of annual hip fracture hospitalizations could be prevented at an aggregate annual savings of $449-493 million. Net per capita savings from calcium supplementation for women over age 75 are between $15.50 and $19.40. CONCLUSIONS: Calcium supplementation for women over age 75 is cost effective if supplements cost Medicare less than $40 per beneficiary annually. These savings are conservative since related costs for surgeons, nursing homes and home care are excluded.

Abstract #: EA2


Orkin FK, Dartmouth-Hitchcock Medical Center, Lebanon, NH, USA

In tandem with concern for escalation of health care costs has been desire for greater value. Strategies in anesthesia care have focused on choosing anesthetic regimens (drugs, techniques) with lower pharmaceutical acquisition costs, ignoring the possibility that anesthetic management may influence resource use beyond the operating room (OR) and postanesthetic care unit (PACU). OBJECTIVE: This study explored the influence of anesthetic management on resource use in major joint replacement throughout the hospital stay. METHOD: Within an institution-wide, multidisciplinary quality improvement initiative, a data base was established, containing administrative, financial, patient, surgical, and anesthetic information, covering all patients having total hip or knee replacement in a university hospital during the period 1/94-7/96. Summary statistics were obtained for variables of interest, with P<0.05 considered statistically significant. RESULTS: The data base contains 832 patients: 437 received general anesthesia, the remainder received alternative techniques. Patients receiving general anesthesia were younger (mean 65.6 vs. 68.7 years, P<0.001). Use of general anesthesia vs. alternative had no influence on anesthesia, operating room, PACU, or intensive care unit (if any) time, number of blood transfusions, physical therapy hours, and charted adverse events. Alternative techniques were associated with shorter ward (6.2 vs. 6.9 days, P<0.003) and hospital (6.3 vs. 7.0, P<0.008) stay, with the greatest differences associated with adjuvants enhancing postoperative pain control. CONCLUSION: Efforts to control anesthesia expenditures should not be limited to acquisition costs of anesthetic drugs and supplies. Rather, a broader perspective must be taken to explore ways in which anesthetic management may influence 'downstream' resource use beyond the OR and PACU. That alternative anesthetic techniques were associated with shorter hospital stay prompts continuing analysis of the data set to identify pharmaceutical components of 'best practice'.

Abstract #: EA3


Kilburg, A, Bachinger, A, Rychlik, R

Institute of Empirical Health Economics, Altenberger-Dom-Strasse 16, D-51519 Odenthal, Germany

Ceftriaxone, Ciprofloxacin, Meropenem and Piperacillin-Tazobactam are some of the most preferred antibiotics to treat pneumonia. In a retrospective analysis the cost structure of above different antibiotic treatments was established. OBJECTIVE: This analysis centred around the question which of these alternative therapies incurs the best cost-effectiveness ratio. The comparison of costs and benefits was made on the basis of a cost-effectiveness analysis. METHOD: A decision tree with four endpoints was constructed: (1) successful therapy with adverse drug reaction (2) successful therapy without adverse drug reaction (3) failed therapy (4) death. Data regarding efficacy, duration of treatment and casualities underlying were taken from clinical studies. Health care costs included in the model were all medical costs (hospitalization, drug costs etc.) and indirect costs (absence from workplace, burial allowance). The benefit was defined as the cure rate of pneumonia. The following benefit rates were revealed in clinical studies: Ceftriaxone 0.6136, Ciprofloxacin 0.6939, Meropenem 0.8750, Piperacillin/Tazobactam 0.9549. For a better comprehension of the results a sensitivity analysis was conducted to determine the influence of the efficacy on the effectiveness adjusted costs and the effects of the frequency of adverse drug reactions using Piperacillin/Tazobactam. RESULTS: Therapy with Ceftriaxone resulted in total costs of 11,942.47 DM per successfully treated patient. Therapy with Ciprofloxacine incurred total costs of 13,352.41 DM, and therapy with Meropenem caused total costs of 7,720.39 DM. Concerning the therapy of pneumonia with Piperacillin/Tazobactam, total costs of 6,538.28 DM per successfully treated patient were calculated. CONCLUSION: This decision tree analysis reveals that therapy with Piperacillin/Tazobactam has the best cost- effectiveness ratio.

Abstract #: EA4


Dedhiya, SD, Kong, SX, Salmon JW, Crawford, SY. The University of Illinois at Chicago, Chicago, IL, US.

Multiple sclerosis (MS) is crippling neurological disorder that manifests itself in varied forms of disability and functional impairment, affecting young adults. Interferon- beta (IFNB) therapy was approved for treatment of relapsing remitting (RR) type MS. However, clinical trials using IFNB prior to its approval did not convincingly establish its effectiveness in reducing exacerbation rates. OBJECTIVE: To assess published clinical trial data and determine if IFNB therapy is effective in reducing exacerbation rates in RR-MS and to examine the strengths and limitations of carrying out a meta-analysis. METHOD: A meta-analysis of published clinical trials using IFNB therapy for MS was carried out. The results were summarized for-reduction in exacerbation rates, number of exacerbation-free subjects in treatment and control groups, and difference in disability status using the Expanded Disability Status Scale (EDSS) score. RESULTS: Six studies were identified based upon predecided inclusion-exclusion criteria. The summarized results revealed that there was a significant reduction in the exacerbation rate in the group treated by IFNB (p<0.001); no difference was observed in the number of exacerbation-free subjects between the treatment and control group. However, the change in EDSS score could not be estimated conclusively because of lack of information. CONCLUSIONS: IFNB therapy reduces exacerbation rates in MS, but issues such as identification of clinical end points for MS, reporting of results, etc., need to be addressed before conducting clinical trials for MS and for successful integration of information in future meta-analyses. As a statistically rigorous method, meta-analysis portends as a valuable tool to provide a comprehensive summary of efficacy and safety estimates--the critical elements, for cost-effectiveness analyses, necessary to establish the impact of drug therapy on patient outcomes.

Abstract #: EA5


Smith MD1, McGhan WF2, LaFrance ND3, Weaver M3, Harris GJ4, Renshaw PF5.

1Health Decision Strategies, 2Philadelphia College of Pharmacy and Science, 3Bracco Diagnostics, 4New England Medical Center, 5McLean Hospital

PURPOSE: The purpose of this study was to compare the cost-effectiveness of several functional neuroimaging modalities for the diagnosis of Alzheimer's disease (AD). METHODS: For the cost-effective analysis of DSC MRI and clinically-evaluated brain SPECT [cSPECT], 23 patients with probable AD and 18 age and gender matched elderly comparison subjects underwent a structural MRI scan and functional DSC MRI scan and cSPECT scan. For the cost-effective analysis of a quantitatively-evaluated SPECT [qSPECT], 20 patients with probable AD and 8 age and gender matched elderly comparison subjects underwent a structural MRI scan and a functional qSPECT scan. A Bayes' revision decision tree was constructed using the sensitivity, specificity, and disease prevalence from the clinical assessment studies. Out-patient direct medical, non- medical, and indirect costs were determined from survey, observational study and published literature. Cost-effectiveness ratios were evaluated from the perspective of a hospital and society. Sensitivity analysis was conducted to establish relationships of costs, disease prevalence, and neuroimaging modality sensitivity and specificity. RESULTS: Direct medical costs to the hospital for structural MRI+DSC MRI and structural MRI+SPECT were $586 and $865 and to society were $773 and $1159. Cost- effectiveness ratios were $650, $1040, and $967 for MRI+DSC, MRI+cSPECT, and MRI+qSPECT per accurate evaluation from a hospital perspective and $858, $1393, and $1296 from a societal perspective. CONCLUSIONS: Based on a Bayesian analytic model, MRI + DSC MRI was the more cost-effective imaging modality in evaluating patients for Alzheimer's disease both from the perspective of a hospital and from society.

Abstract #: EA6


Okano, GJ, Rascati, KL, Wilson, JP, Smeeding, JE, The University of Texas College of Pharmacy, Austin, TX. USA, Grabenstein, JD, Remund, D, The Pharmacoeconomic Center, Ft. Sam Houston, TX. USA

The Department of Defense has recently constructed a prescription database through the Uniformed Services Prescription Database Project. OBJECTIVES: For patients initiated on an ACE Inhibitor or Calcium Channel Blocker (CCB), the objectives of this study were to retrospectively: 1) determine those patients experiencing ONE change in therapy from initiation in therapy; and 2) estimate the marginal medication costs associated with the change. METHODS: Patient antihypertensive therapy was examined from February 1, 1994 through July 31, 1995. The unit of analysis was the patient. If patients were continuously enrolled for the 18 months of the study; had no antihypertensive medications from February 1, 1994 to July 31, 1994; but began therapy with an ACE Inhibitor or CCB in August 1994, their therapy was followed for one year. Marginal costs were estimated using Average Wholesale Price (AWP) and Average Manufacturers Price (AMP). RESULTS: Based on the criteria, 183 patients were identified as newly treated hypertensive patients. Of the 183 patients, 42 (23%) experienced one change in therapy. Specifically: 14 (8%) increased dose; 2 (1%) decreased dose; 17 (9%) changed medications; and 9 (5%) had another medication added. The estimated increase in marginal cost per patient above the cost of continuing on the same therapy for the year was calculated at $53.83 (AWP) and $38.79 (AMP). CONCLUSION: This results of this study provide insight into the utilization of antihypertensive therapy and the associated marginal costs as a result of the change. Analysis of data for a larger sample of patients is currently being conducted.

Hypertension: Identification of Patients and Costs

Abstract #: HY1

Utility and Validation of Administrative Data for Identifying a Hypertensive Cohort in a Managed Care Population.

James F. Murray, James B. Roehm, Charles A. Alexander, Nicholas A. Hanchak, Stephen J. Boccuzzi, U.S. Quality Algorithms, Inc.¨, Blue Bell, PA, and Merck & Co. Inc., West Point, PA.

Retrospective analysis of administrative data has proven to be an effective method for the characterization of provider practice patterns. A limitation in the use of this type of data is the ability to accurately identify the target population. To develop and validate a selection algorithm for the identification of hypertensive patients a retrospective study in a 2 million member independent practice HMO with an open formulary was conducted. Newly diagnosed members with hypertension were identified on the date of their first claim/encounter with hypertension-specific ICD-9 criteria, preceded by at least a 12 month interval without hypertension criteria. Additional claims or pharmacy records, if available, were used as corroborating evidence for hypertension (i.e., drug claim within +/- 6 months or a second ICD-9 encounter/claim within +12 months). Diagnostic validation of the selection algorithm was accomplished by auditing a 5% random sample of 174 patients through medical chart audit. The hypertension diagnosis was confirmed if it was recorded on the problem list or 2 recorded blood pressure readings exceeding 140 mm Hg systolic or 90 mm Hg diastolic were found. 157 of the 174 members were confirmed hypertensive for a positive predictive value (PPV) of 90.2%. The PPV was improved to 95.4% when corroborating evidence of the diagnosis was utilized. The algorithm was highly accurate for identifying a hypertensive cohort for describing the practice patterns in the pre and post JNC V time frames to include 1) the distribution of anti-hypertension agents in new starts, 2) discontinuation rate, 3) persistence and 4) compliance.

Abstract #: HY2


Hudson TJ, Gardner SF, Smith KF, Davis DA. University of Arkansas for Medical Sciences, Little Rock, Arkansas and Pharmacy Associates, Inc., Little Rock, Arkansas

Thiazide diuretics, although widely prescribed, are associated with a relatively flat dose response curve and a significant increase in metabolic disturbances at doses 25mg/day. OBJECTIVE: The purpose of this study was to determine the prevalence of prescriptions for hydrochlorothiazide (HCTZ) 50mg/day in a paid claims database. We also identified prescription profiles with concomitant medications to treat metabolic disturbances potentially worsened by thiazide diuretics. METHODS: We queried the database of a PBM using RxFocus¨ software (Argus Health Systems, Inc.). The database contained paid claims for prescriptions for 145,000 patients from Arkansas and Louisiana. Patient profiles with prescriptions for HCTZ 50mg/day from March 1, 1997 to May 31, 1997 were identified. Letters were mailed to prescribing physicians outlining the ceiling effect and metabolic disturbances associated with high doses of thiazide diuretics. Each profile was evaluated for the presence of medications used to treat metabolic disorders. RESULTS: Three hundred seventy-six patient profiles with paid prescription claims for HCTZ 50mg/day were identified. Thirty-six profiles were excluded secondary to incomplete data. The profiles (n=340) reflected patients aged 27 to 84 years (37% male). One hundred four (31%) profiles included at least one medication for treatment of a metabolic disorder. Medications included anti-diabetic agents (n=23), hypolipidemic medications (n=12), potassium supplements (n=46) and medications for gout (n=5). Eighteen patients were receiving medications from more than one category. CONCLUSIONS: Despite published data regarding the flat dose response curve of HCTZ and increased side-effects at higher doses, 340 patient profiles in this database contained paid claims for HCTZ 50mg/day. Nearly one-third (31%) of the profiles indicated the patient was being treated for at least one metabolic disorder potentially worsened by high-dose HCTZ. Although some individuals may respond favorably to HCTZ 50mg/day, patients should be closely monitored for the presence of new metabolic disturbances or worsening of known disorders.

Abstract #: HY3


Tasch, RF, Marentette, MA, West, R, Technology Assessment Group, Montreal, QC; Memorial University of Newfoundland, St. John's, NF, Canada

OBJECTIVE: To estimate the incremental cost of treating hypertensive patients in Canada. METHODS: Using the Saskatchewan Health databases, we measured the direct medical cost of treatment, in terms of physician visits (MD), prescription medications (Rx), and hospitalizations, of a case group and an age- and gender-matched control group. Rx were categorized as cardiovascular and non- cardiovascular drugs. We performed univariate and multivariate regression analyses using variables such as age, gender, and common comorbid conditions to predict the cost of MD, Rx, hospitalizations, and total treatment. Finally, we extrapolated the results from Saskatchewan to a national level, using Canadian population prevalence figures, in order to estimate the total annual cost of treating hypertension in Canada. RESULTS: For hypertensive patients, the annual cost of treatment per patient was $2,040, comprised of $500 for MD, $599 for Rx, and $940 for hospitalizations. For control group patients, the annual cost of treatment per patient was $1,360, comprised of $347 for MD, $256 for Rx, and $756 for hospitalizations. All these differences were highly statistically significant. In addition, we found that patients in the hypertension group had more comorbid conditions than control group patients. Controlling for the presence of comorbid conditions, as well as age and gender, reduces the difference per patient from $680 (unadjusted) to $603 (adjusted). CONCLUSIONS: Applying the adjusted cost difference to a national level yields an annual incremental cost of treating hypertension in Canada of $1.3 billion to $2.8 billion, depending on assumptions concerning prevalence. These figures may be conservative given that certain medical costs, such as laboratory tests and procedures, as well as indirect costs to society such as lost productivity, were not included.

Abstract #: HY4


Russell, MW, Huse, DM, Drowns S, Hartz, SC, Medical Research International, Burlington, MA. US

Huse, DM, Sytkowski, PA, D'Agostino, RB, Drowns, S, Hamel,EC, Hartz, SC, Medical Research International, Burlington, MA. US and Boston University, Boston, MA US

Many persons taking antihypertensive drugs continue to have elevated blood pressure relative to current guidelines. OBJECTIVE: To estimate the clinical and economic benefits of more rigorous adherence to current treatment guidelines. METHOD: We developed an economic model of cardiovascular disease (CVD, including coronary heart disease [CHD]. Stroke, and congestive heart failure [CHF]) and its relationship to blood pressure and antihypertensive treatment in the U.S. adult population (aged 40-79 years). This model forecasts CVD incidence and direct medical-care costs (discounted at 5% per annum) for demographic groups using national risk factor profiles. Principal data sources included the Framingham Heart Study, major national health surveys, and healthcare cost database. RESULTS: We estimate that 18% of US men and women aged 40-79 who ae currently free of CVD receive antihypertensive therapy and that 17% of treated persons continue to have diastolic bloood pressure of 90 mm Hg or higher. Among those receiving antihypertensive therapy, incidence of CHD, stroke, and CHF increase by 23%, 15%, and 65%, repsectively, with each 10-point rise in diastolic blood pressure. Expected per capita costs of CVD over five years among treated hypertensives range from $1,470 at DBP <80 mm Hg to $1,746 at DBP of 80-89 mm Hg to $2,270 at DBP> 90 mm Hg. CONCLUSION: More aggressive treatment of hypertension can be expected to yield substantial economic benefits in addition to reducing the risk of cardiovascular diasease and mortality.

Abstract #: HY5


Ricci JF, Oh S, Martin B., College of Pharmacy, Pharmacy Care Administration, University of Georgia, Athens, Ga. US

The purpose of this study was to determine the marginal cost-effectiveness of the pharmacological treatment of hypertension in the elderly. METHOD: A decision tree modeling the prognosis of hypertensive elderly patient between 60 and 75-years old was constructed. The model was run over 15 years in one-year increments. It included one decision node with two emanating branches (pharmacological treatment and no treatment of hypertension), each modeled with a Markov process. Transition probabilities were estimated using the Mantel-Haenzel pooled estimates for event rates for seven morbid and fatal clinical endpoints obtained from nine published randomized clinical trials on the treatment of hypertension for patients 60 years of age and older. Cardiovascular clinical endpoints were stratified in three categories - cerebrovascular, major coronary, and all other cardiovascular events. Cost data were obtained from a two-year retrospective study of the administrative claims of the Medicaid Program in Georgia. RESULTS: The average annual anti-hypertensive prescription expenditures was $428 per patient. Average annual hospital expenditures for patients without any prior diagnosis in one of the three cardiovascular categories was $2,150 vs. $4,030 for patients with at least one prior cardiovascular endpoint diagnosis. Over a 15-year time horizon, the non-discounted life expectancy in the treatment group was 11.79 years, a 3% discounted life-years gain of 0.37 compared to the no treatment group. The average marginal cost- effectiveness ratio indicated that pharmacological therapy increased life expectancy at a cost of $2,047 per life-year gained with a 95% confidence interval going from a cost-saving scenario to $7,197. CONCLUSION: Pharmacological treatment of hypertension in the elderly is cost- effective in comparison to many other medical interventions

Abstract #: HY6


Linde-Zwirble,W, Newbold,RC, Lidicker, J Health Process Management, Doylestown, PA. US

Hypertension affects almost 10% of the PA Medicaid population. Given the wide variety of hypertension agents available, prescribers must understand the efficacy, safety, and cost of each agent. OBJECTIVE: This study compared two once-daily, dihydropyridine Calcium Channel Blockers (CCBs) for number of healthcare encounters, use of healthcare resources, total and drug charges. METHOD: The 1994-96 comprehensive PA Medicaid claims database was reviewed for all adult patients ages 18-64 with a diagnosis of hypertension. All patients were retained for study with: at least one prescription for either Norvasc or Adalat CC, no other CCBs in the 8 months preceding the earliest prescription (REFERENCE) and no other CCBs in the 9 months following the earliest prescription (MONITORING), and continuous eligibility. RESULTS: The demographic characteristics and compliance of the two groups were similar. Adverse outcomes as well as resource use were largely similar. However, the Norvasc group had a higher rate of CV conditions (N=419, 14.6%) than Adalat CC (N=176, 7.4%) and a higher rate (23.4%) of hospitalizations than Adalat CC (15.9%). Interestingly, Adalat CC had a significantly lower average number of classes of concurrent antihypertensive agents (0.38 vs. 0.52); this correlated with lower cost of medications for treatment with Adalat CC. After controlling for pre-existing CV conditions, this effect remained. CONCLUSION: A direct cost benefit for Adalat CC was observed because of less use of concomitant drugs for hypertension. Differences in other kinds of resource use between the two groups were either rare (like hospitalizations), or unvarying (like physician visits). We believe that there may be an opportunity for cost savings with Adalat CC in managing hypertension because of lower acquisition cost and lower concomitant drug therapy.

Costs, Resources, and Utilities in Cardiovascular Disease

Abstract #: CD1


Russell, MW, Huse, DM, Drowns, S, Hamel, EC, Hartz, SC, Medical Research International, Burlington, MA. US

Despite significant reductions in CAD incidence and prevalence in the U.S. in recent years, the costs of treating CAD remain of great concern. New interventions that reduce CAD risk continue to be developed and may offset some of these costs. However, they compete with existing interventions for limited healthcare resources. These resource allocation decisions require quantification of the expected costs of treating CAD. OBJECTIVE: To develop current incidence- based estimates of the direct medical costs of CAD in the U.S. METHODS: A Markov model of the costs of CAD-related medical care among persons with and without prior history of CAD was developed. Risks of initial and subsequent CAD events were estimated using recently updated Framingham Heart Study coronary risk equations and national data on CAD risk factors. Costs reflected treatment of CAD events ("event-related") and surgical/nonsurgical follow-on care ("non event-related"), and were derived from national public-use databases and medical literature. RESULTS: First-year event- related costs are estimated to be $17,532 (fatal MI), $15,540 (nonfatal MI), $2,569 (angina), $12,058 (unstable angina), and $713 (sudden CAD death). Non event- related costs are estimated to be $1,051 annually. Five- and ten-year cumulative costs (1995 dollars) for persons initially free of CAD are estimated at $9.3 billion and $12.6 billion respectively; for all persons with CAD, these costs are estimated to be $71.5 billion and $126.6 billion respectively. CONCLUSION: The direct medical costs of CAD are substantial, and as new interventions that reduce the risk of CAD are introduced to the marketplace, healthcare providers and payers need to consider their net effect on the costs of managing this disease.

Abstract #: CD2


Zarkin, GA, Bala, MV, Calingaert, B, VanderLugt, JT, Research Triangle Institute, Research Triangle Park, NC, US

Pharmacological agents used in treating atrial fibrillation (AF) and atrial flutter (AFL) act much more slowly and are less effective than electrical cardioversion (EC). Recently, ibutilide fumerate, a new intravenous rapid- acting Class III antiarrhythmic agent, was approved by FDA for treating AF and AFL. OBJECTIVE: To compare the cost- effectiveness of a stepped treatment regimen of ibutilide first-line followed by EC second-line, to treatment with EC first-line, for AF and AFL. METHODS AND DATA: We developed a decision-tree model of the treatment of AF and AFL using outcome probabilities from a Phase III clinical trial. The resource use was based on the literature and physician clinical judgment. The resource costs were based on Medicare and hospital charge data adjusted by Medicare cost-to-charge ratios. RESULTS: For AF, a stepped treatment regimen of ibutilide followed by EC second-line is cheaper by $260 (95% CI: -$658, $27) and has a higher conversion rate (0.92 vs 0.78) than treatment with EC first-line. For AFL, the stepped treatment regimen results in a cost reduction of $395 (95% CI: -$685,-$206) and an increase in conversion rate (0.98 vs 0.86). Monte Carlo simulation showed that the stepped therapy was cheaper for 96 percent of AF patients and 100 percent of AFL patients. CONCLUSIONS: A stepped treatment regimen of ibutilide followed by EC second-line is both less expensive and more effective than treatment with EC first-line for AF and AFL. Our study illustrates cost- effectiveness analysis of a stepped therapy and provides guidelines for performing Monte Carlo simulations and presenting sensitivity results for such analyses.

Abstract #: CD3


Glick, H, Polsky, D., Willke, R, Schulman, K, University of Pennsylvania, Philadelphia, PA, Pharmacia and Upjohn Inc. Kalamazoo, MI 

Objective: Using data from a randomized trial of tirilazad mesylate we assess the differences between Canada and the U.S. in resource utilization and outcome in the treatment of aneurysmal subarachnoid hemorrhage during both initial hospitalization and post-discharge periods. Methods: Of the 877 patients randomly assigned to one of three treatment groups, 194 were enrolled in Canada and 683 were enrolled in the U.S. We analyze the difference between countries in patient characteristics, resource utilization, costs, and outcomes by comparing means and the 95% confidence interval around the difference in means. We also predicted utilization with multivariate regression analysis. Results: Average hospital stay was 4.7 days (95% CI, 1.1-8.2) longer in Canada. In general, however, hospital stays in Canada were substantially less intensive. Patients treated in Canada spent 4.1 days (95% CI, -6.5 to -1.7) fewer in nursing homes and rehabilitation centers than did those in the U.S in the 90 days post- randomization. The average cost of care in the U.S., using U.S. unit costs, was $11,234 more than in Canada (95% CI, $5,204 to $17,046). Using Canadian unit costs for Canada the cost difference increased to $15,328 (95% CI, $9,759 to $20,898). There were no statistically significant differences among the outcomes of Glasgow Outcome Score, death, and occurrence of vasospasm. When controlling for differences in the severity of illness at randomization, the results remained qualitatively the same. Conclusion: We found that the apparent difference in length of stay between the countries was due to a shift in the sites of formal care rather than in the number of days of this care. We found no significant difference in outcomes to justify the additional costs of care in the U.S. We also found that costs for the initial hospitalization in the U.S. were substantially higher than those in Canada despite the shorter lengths of stay.

Abstract #: CD4


Nguyen, M.H.1, Patterson, H.K. 1, Dlutowski, M.J. 1, Duong, P.T. 2, Johnson, N.E. 1, Stang, P.E. 2

1Office of Health Policy and Clinical Outcomes, Thomas Jefferson University, Philadelphia, PA, USA. Hospital, 2SmithKline Beecham Pharmaceuticals, Collegeville, PA, USA. Office of Health Policy and Clinical Outcomes

Despite its proven efficacy in congestive heart failure (CHF), there is some evidence which suggests an underutilization of ACE inhibitor therapy in patients with CHF. In June of 1994, AHCPR CHF Gguidelines for the management of patients with heart failure were published by the Agency for Health Care Policy and Research (AHCPR) in 1994 were established to to assist clinicians physicians in the management of patients with left left heart failure. and as a means of reducing unexplained variation in clinical practice. Guideline implementations can produce better patients outcomes. Despite the proven efficacy of ACE Inhibitor (ACEI) therapy in patients with left heart failure, there is evidence suggesting that ACEI therapy has been underutilized in these patients.

OBJECTIVE: The purpose of this study wasisto evaluatecompare the use trends of ACEIInhibitor usage for and treatment outcomes of patients hospitalized with exacerbations of for heart failure CHFafter over a one year period following the AHCPR publication of the AHCPR guidelines were disseminated. METHODS: Patients admitted to the hospital with congestive heart failure between 1994 and 1995 were identified from our clinical and financial accounting database using ICD-9-CM codes for CHF (428.0) and with DRG code 127 DRG codes for heart failure and shock (127)and ICD-9-CM code 428.0 for CHF were identified from our clinical and financial accounting database. We randomly selected and conducted a retrospective chart review of a sample of 50 patients' charts from each year to determine eligibility and use ACEIInhibitortherapy. eligibility and usage in 1995 and compared to the utilization reported in the previous year. Additional data were collected for all identified encountersextracted which including ed disease severity of illness, length of hospital stayLOS, resource utilization and re-admission rates. RESULTS:


(days)Re-admission RateDischarged Alive

1994 (n=208)6.713%96.2%

1995 (n=198)7.710.1%94.5%The use of ACEI for eligible patients increased in the year after the AHCPR guidelines were published. In 1995, 85% of eligible patients were discharged on ACEI Inhibitor therapy compared with onlyto 61% of patients in 1994. The LOS remained consistent for the two period. CONCLUSION: There appearsseems to be a trends towards increased utilization of ACEIInhibitorsamong eligible patients with no difference in LOS (excluding outliers) with decreased re-admission rates at our hospital., as recommended by the guidelines among eligible patients at our University Hospital

Abstract #: CD5


Bagchi, I, Taylor, TN, University of Iowa, Iowa City, IA, USA

OBJECTIVE: The objective of this study was to evaluate the use of the New York Heart Association (NYHA) functional classification for measuring utilities for health states associated with cardiovascular diseases. METHOD: The subjects for this study were 41 students from the College of Pharmacy who were presented with a simplified, four-state scenario for cardio-vascular diseases, based on the NYHA functional classification. Utility assessments were conducted by two interviewers using a computer program and a standardized protocol. Test-retest reliability was determined by correlation analysis. Regression analysis was used to identify factors explaining variation in utility scores. RESULTS: The utility scores were as follows:


LimitationSlight limitationModerate LimitationMarked Limitation

Visual Analog Scale88.6 (+7.5)65.9 (+10.0) 42.6 (+12.0)21.1 (+10.9)

Time Tradeoff89.5 (+12.3)78.7 (+17.0) 64.0 (+23.5)39.5 (+27.0)

Standard Gamble93.7 (+7.2)85.7 (+11.0) 72.2 (+16.4)47.3 (+26.0)For each health state, the mean utility values obtained by the three instruments were in the following order: SG>TTO>VAS, consistent with the literature. For each instrument, progressively better health states were associated with higher mean utility scores. The VAS exhibited higher test-retest reliability than TTO or SG though the number of subjects completing a second interview was small (n=12). Neither age nor gender were found to affect the scores. Interviewer effects were found to exist for TTO and SG, but not for the VAS scores. CONCLUSION: It is concluded that a simplified four-state scenario based on the NYHA classification, is useful for measuring utilities for health states associated with cardiovascular diseases.

Abstract #: CD6


Rappenhoner, B, Bachinger, A, Pfeil, T, Rychlik, R. Institute of Empirical Health Economics, Altenberger-Dom-Strasse 16, D-51519 Odenthal, Germany

OBJECTIVE: Both direct and indirect costs incurred by treatment of chronic venous insufficiency including qualitiy of life in Germany were illustrated on the basis of a cost utility analysis comparing compression therapy to an ascin containing drug.

By means of a decision tree analysis these two alternative clinical treatments were compared. METHODS: First a period of one year was considered, then the results were projected for a period of 10 years. The analysis is based on a simple blind, placebo-controlled, randomized clinical study. The benefit of these clinical treatments was defined as improvement of quality of life compared to placebo and measured in quality-adjusted life-years (QALY). RESULTS: For drug application both direct and indirect costs amount to DM 1,.370.45. The direct and indirect costs of compression therapy amount to DM 1,512.12. The enhanced quality of life compared with placebo for drug therapy revealed 0.1074 QALY and for compression therapy 0.0278 QALY. To enhance qualitiy of life by one QALY, the following costs are incurred: drug DM 14,079.32, compression therapy DM 433,126.62. Projected for 10 years, costs for drug application have been calculated at DM 123.702,51 for compression therapy at DM 433,126.62. A sensitivity analysis reveals that even if compression therapy is successful, costs of drug application per QALY will be lower than those of compression therapy at any presumption. CONCLUSION: Considering the implementation of quality of life (QALY) costs incurred by treatment of chronic venous insufficiency with drug compared to those for compression therapy will be lower, although the individual evaluation of direct and indirect costs revealed cost reduction more for compression therapy.

Economic and Outcomes Issues in Managed Care, Pharmacy Benefit Management, and Government

Abstract #: EO01


Khan, ZM, Rascati, KL, Smeeding, J, Koeller, J, The University of Texas, Austin, TX, US

The purpose of this study was to conduct a resource utilization and an economic analysis on the use of carboplatin versus cisplatin over multiple cycles in lung and ovarian cancer. OBJECTIVE: The objective was to measure and compare direct medical resource utilization and costs associated with carboplatin and cisplatin administration over three to six courses of treatment. METHOD: The perspective of this one-year prospective, multi-center, cost-minimization evaluation was that of the provider. A convenience sample of 16 sites representing a mix of cancer centers, outpatient clinics, medical centers, and managed care sites participated. Individuals involved with data collection at all sites were trained via on-site and/or teleconference training. Site visits were made to assure reliability of at least 0.80. Data were collected and compiled via a fax transmission process that scans directly through optical mark and character recognition into a computer data base. Outcome measures included costs of: medications, emergency room visits, physician/clinic visits, home health care visits, transfusions, special procedures, consultations, hospitalizations, and other/miscellaneous costs. RESULTS: Of 220 patients, 164 met the study criteria (response rate=75%) with 95 patients in the carboplatin group and 69 in the cisplatin group (average numbers of courses were 4.5 for both groups). The average number of visits per patient were as follows for carboplatin and cisplatin respectively: hospitalizations - 0.25 vs. 0.28; clinic visits - 0.95 vs. 1.67; ER visits - 0.11 vs. 0.09; and lab visits - 5.70 vs. 5.20. All outcome variables and their costs will be presented. CONCLUSION: When evaluating the costs of different drug therapies, it is important to evaluate other relevant costs associated with the treatment.

Abstract #: EO02

Costs and antihyperglycemic drug use over 8 years among HMO members newly diagnosed with type 2 diabetes (DM2) Brown JB1, Glauber HS1, Nichols GA1, Bakst AW2, 1Kaiser Permanente Center for Health Research (KPCHR) Portland, OR, US and 2SmithKline Beecham Pharmaceuticals, Collegeville, PA, US

Brown JB1, Glauber HS1, Nichols GA1, Bakst AW2, 1Kaiser Permanente Center for Health Research (KPCHR) Portland, OR, US and 2SmithKline Beecham Pharmaceuticals, Collegeville, PA, US

There is a lack of information about long-term patterns of antidiabetic drug use. Objective:

To describe patterns of antidiabetic drug use over an eight year period and evaluate costs of antidiabetic drug utilization in a large diabetic population. Method: In 1987, KPCHR began a registry of all members with diabetes, identified primarily from hospital discharge diagnoses and purchases of drugs and testing supplies. For the purposes of this study, decision rules were established to identify DM2 registry patients. Drug utilization data was obtained from records of drugs dispensed from Kaiser pharmacies during the first quarter (Q1) of each eight calendar years (1988-1995). Results: Of 1014 HMO members newly identified as DM2 in Q1 1988, 10% used no antidiabetic drug (NRX), 79% used sulfonylureas (SU) only, 9% used insulin (I) only, and 2% used both I and SU. Over the 8 year period the percentage of HMO members from the initial cohort transitioning to I therapy and SU + I had doubled. By 1995, 29% of the original cohort had died, 19% used NRX, 31% used SU only, 19% used I only, and 3% used SU + I. There was also a trend towards increasing SU dosage over time. In a separate analysis of all DM patients in 1995, the average annual antidiabetic drug/supply costs for patients receiving SU was $207 compared to $439 for I only patients and $511 for SU + I patients. Conclusions: Improved glycemic control of patients would reduce the rate of transition to insulin therapy in DM2 patients and may reduce diabetic drug and supply costs.

Abstract #: EO03


Brown, JB1, Glauber HS1, Nichols GA1, Bakst AW2, Kaiser Permanente Center for Health Research (KPCHR), Portland,OR, US and SmithKline Beecham Pharmaceuticals, Collegeville, PA, US

There is a lack of information about long-term patterns of antidiabetic drug use. OBJECTIVE: To describe patterns of antidiabetic drug use over an eight year period and evaluate costs of antidiabetic drug utilization in a large diabetic population. METHOD: In 1987, KPCHR began a registry of all members with diabetes, identified primarily from hospital discharge diagnoses and purchase of drugs and testing supplies. For the purposes of this study, decision rules were established to identify DM2 registry patients. Drug utilization data was obtained from records of drugs dispensed from Kaiser pharmacies during the first quarter (Q1) of each eight calendar years (1988-1995). Results: Of 1014 HMO members newly identified as DM2 in 1Q 1988, 10% used no antidiabetic drug (NRX), 79% used sulfonylureas (SU) only, 9% used insulin (I) only, and 2% used both I and SU. Over the 8 year period, the percentage of HMO members from the initial cohort transitioning to I therapy and SU+I had doubled. By 1995, 29% of the initial cohort had died, 19% used NRX, 31% used SU only, 19% used I only and 3% used SU+I. There was also a trend towards increasing SU dosage over time, In a separate analysis of all DM patients in 1995, the average annual antidiabetic drug/supply costs for patients receiving SU was $207 compared to $439 for I only patients and $511 for SU+I patients. Conclusions: Improved glycemic control of patients would reduce the rate of transition to insulin therapy in DM2 patients and may reduce diabetic drug and supply treatment costs.

Abstract #: EO04

Antidepressant Therapy in a Mental Health Managed Care Setting

Conner TM1, Trott J2, Rascati KL1 The University of Texas at Austin, USA1 The University of Texas at San Antonio, USA2

The newer antidepressant drugs available today are reported to be highly efficacious in reducing depressive symptoms and cause few side effects to the patient. However, few studies have been conducted on these newer drugs in "real-world" settings. OBJECTIVES & METHODS: In a retrospective review of medical records, we examined drug treatment for major depression at a mental healthcare organization. Specifically, we sought to describe 1) which antidepressant drugs were most frequently prescribed at initial outpatient visit; 2) how often dose (other than initial titration) or drug changes were made within six months of initial treatment; 3) how often these therapy changes were made because of side effects; and 4) how often adjunct therapy was prescribed for reported side effects from the antidepressant therapy. We identified 335 new patients with major depression who began treatment between September 1, 1995 and August 31, 1996. RESULTS: One hundred forty-four (43%) medical records have been reviewed. Of these, forty-seven met study criteria and have been analyzed. Preliminary results show that approximately 62% of patients were prescribed either fluoxetine (14 of 47) or sertraline (15 of 47) as first-line therapy; each patient averaged 1.4 (s.d. 1.10) dose or drug changes over a six- month period; approximately 40% of therapy changes were made because of reported side effects; and adjunct therapy was documented to have been prescribed for one patient specifically for a side effect from antidepressant therapy. CONCLUSION: Further analysis on a larger sample size is currently being conducted. Preliminary findings suggest that adjunct therapy for side effects of antidepressants is not common in this setting. Implications for future research, including cost analysis, will be discussed.

Abstract #: EO05


Kong, SX, Dedhiya, SD, G.D. Searle & Co. Skokie, IL, US.

Each year 9 to 12 million people suffer from asthma in the US. Asthma remains a chronic condition that has a significant effect on patients' personal and social life. OBJECTIVE: To evaluate the general and disease specific health-related quality of life (HQL) of asthmatic patients and to examine the major factors that may have a significant impact on the HQL. METHODS: Adult asthmatic patients (N=480; age Ň 17) living in a large metropolitan area were identified from the computerized claims database of a large HMO. A mail questionnaire survey which included the SF-36 Health Survey to measure generic HQL and the Asthma Quality of Life Questionnaire (AQLQ) for the asthma-specific HQL, was developed. Other variables included perceived asthma severity, impact on sleep, support from family members and friends, health care resource utilization (e.g., ER visits, hospitalization), work days lost and demographic variables. Regression analysis was used to identify the significant predictors of different domains of generic and asthma disease-specific HQL. RESULTS: A 42% response rate was achieved. The average age of the respondents was 39.4 (SD, 12.6) years. About 76% were female; 49% were black, and 33% were white; 27% were smokers, 27% were ex-smokers, and 46% had never smoked. Female patients reported lower scores on several domains of the SF-36 and AQLQ although there were no significant gender-related differences with respect to other variables measured. Patients with lower HQL scores reported significantly greater health care resource utilization and had incurred more loss in work days. CONCLUSIONS: Asthma can significantly reduce patient HQL measured by both generic and disease-specific instruments. There is a strong correlation between the HQL scores measured by the two instruments. Patients with lower HQL scores consumed more health care resource and had more work days lost.

Abstract #: EO06


Wisner, C, Abbott, T. Fischer, L, Berger, M,

HealthPartners, Minneapolis, MN, US, Merck & Co., Inc. West Point, PA, US

This study determined the incidence of fractures, pulmonary embolism and pneumonia in health plan enrollees over fifty years old. STUDY DESIGN: Enrollees were assessed for incidence of one of ten fracture types categorized into central (hip and pelvis) and appendage (femur, patella, ankle, tibia/fibula, proximal humerus, other humerus, distal forearm, and radius/ulna) fractures, pulmonary embolus, and pneumonia. Comparisons were made between men and women, fifty to sixty-four years and sixty-five plus years. RESULTS: In both age groups, women were twice as likely to experience fractures as men (.009 vs. .005 and .02 vs. .01, respectively). For both sexes, the incidence of fractures doubled in the older group. Complications of pulmonary embolus were experienced by six in the older female group (n=694) compared to one in the comparison group (drawn from a population with no fractures). Males had one incidence of pulmonary pneumonia compared to four in the comparison group. Males in the older group (n=217) had eleven cases of pneumonia after the fracture. CONCLUSION: Women are much more likely than men to experience fractures of all kinds in all age groups. Women also appear more likely to experience complications related to the fracture, particularly older women.

This study determined the incidence of fractures, pulmonary embolism and pneumonia in health plan enrollees over fifty years old. STUDY DESIGN: Enrollees were assessed for incidence of one of ten fracture types categorized into central (hip and pelvis) and appendage (femur, patella, ankle, tibia/fibula, proximal humerus, other humerus, distal forearm, and radius/ulna) fractures, pulmonary embolus, and pneumonia. Comparisons were made between men and women, fifty to sixty-four years and sixty-five plus years. RESULTS: In both age groups, women were twice as likely to experience fractures as men (.009 vs. .005 and .02 vs. .01, respectively). For both sexes, the incidence of fractures doubled in the older group. Complications of pulmonary embolus were experienced by six in the older female group (n=694) compared to one in the comparison group (drawn from a population with no fractures). Males had one incidence of pulmonary pneumonia compared to four in the comparison group. Males in the older group (n=217) had eleven cases of pneumonia after the fracture. CONCLUSION: Women are much more likely than men to experience fractures of all kinds in all age groups. Women also appear more likely to experience complications related to the fracture, particularly older women. This study determined the incidence of fractures, pulmonary embolism and pneumonia in health plan enrollees over fifty years old. STUDY DESIGN: Enrollees were assessed for incidence of one of ten fracture types categorized into central (hip and pelvis) and appendage (femur, patella, ankle, tibia/fibula, proximal humerus, other humerus, distal forearm, and radius/ulna) fractures, pulmonary embolus, and pneumonia. Comparisons were made between men and women, fifty to sixty-four years and sixty-five plus years. RESULTS: In both age groups, women were twice as likely to experience fractures as men (.009 vs. .005 and .02 vs. .01, respectively). For both sexes, the incidence of fractures doubled in the older group. Complications of pulmonary embolus were experienced by six in the older female group (n=694) compared to one in the comparison group (drawn from a population with no fractures). Males had one incidence of pulmonary pneumonia compared to four in the comparison group. Males in the older group (n=217) had eleven cases of pneumonia after the fracture. CONCLUSION: Women are much more likely than men to experience fractures of all kinds in all age groups. Women also appear more likely to experience complications related to the fracture, particularly older women.

Abstract #: EO07

THE COST OF OSTEOPOROSIS RELATED FRACTURES IN WOMEN AGED 50-64Abbott, T, Wisner, C, Rizzo, J, Fischer, LR,. Amundson, G, Hedblom B, and Berger, M, Merck and Company, Inc. West Point, PA, USA and Health Partners, Minneapolis, MN, USA

Abbott, T, Wisner, C, Rizzo, J, Fischer, LR,. Amundson, G, Hedblom B, and Berger, M, Merck and Company, Inc. West Point, PA, USA and Health Partners, Minneapolis, MN, USA

Osteoporosis is a chronic condition characterized by increased risk of fracture. Each year an estimated 1.3 million fractures can be associated with osteoporosis. Estimates of the direct medical cost of these fractures exceed $10 billion. OBJECTIVE: To assess the time pattern of health care expenditures associated with osteoporotic fracture and determine if there are important long term impacts. METHODS: Women between ages 50 and 64 continuously enrolled in a large Midwestern MCO from 1/93 to 12/95 were studied for osteoporosis related fractures. Cases were identified as incident fracture of the hip, pelvis, femur, patella, ankle, tibia, proximal humerus, other humerus, forearm, or ulna. Cases were matched like aged control. All charges for cases and controls were collected for a total of 18 months beginning 6 months prior to fracture. These charges were subsequently divided into three time periods: baseline (6 months prior), acute (first 3 months), and chronic (remaining 9 months). RESULTS: During the baseline period, those women who subsequently fractured had substantially higher health care utilization ($330 per month compared to $190). During the acute phase, costs increased dramatically for cases ($1719 per month). During the chronic phase, women who fractured continued to have increased health care utilization compared to baseline ($427 per month). CONCLUSIONS: First, women who fracture tend to have higher utilization of health care resources than women who did not fracture. Second, there is an increase in utilization even after accounting for acute care. Third, current estimates of the total burden of illness which focus on acute care may dramatically underestimate the true costs.

Abstract #: EO08


Fischer, L, Wisner, C., Abbott, T., Amundson, G., HealthPartners, Minneapolis, MN, US, Merck & Co., Inc. West Point, PA, US

OBJECTIVE: This paper addresses two questions: Are older women who experience fractures sicker and more impaired than an age-matched sample of women without fractures? What is the impact of fractures on health outcomes for older women? STUDY DESIGN: Samples were drawn of women age 65+ enrolled in a Social HMO which was part of a large HMO in the Twin Cities, Minnesota. Women who experienced fractures were compared to a sample of women without fractures, matched by age. Data were drawn from health status surveys before and after the date of fracture (fracture N = 42; matched N = 126). Four variables were examined: health status, ADLs, IADLs, and falls. The analyses compared age-matched samples and: (1) a general fracture sample, (2) a central fracture sample (hip and pelvis), and (3) an appendage fracture sample (non-hip femur, patella, ankle, tibia/fibula, proximal humerus, other humerus, distal forearm, radius/ulna). RESULTS: Preliminary analyses indicate older women with central fractures, compared to the matched sample without fractures, were significantly more likely to report health and functional problems both before and after the fracture. No differences in health or functional status were found in any of the comparisons of the appendage fracture group. Women with fractures were more likely to have had falls. Both the fracture and matched samples tended to have more functional deficits in the later time period. CONCLUSION: Older women and health and functional impairments are especially at risk for central fractures (hip and pelvis). There is also an aging effect. In this small sample, we were not able to detect an impact of fractures on health outcomes.

Abstract #: EO09

Academic detailing to affect market share of hmg-coa reductase inhibitors

Benson, SR, Torborg, SA, Boreen, DM, Magda, JJ, Burns, KD, Waltermire, RD, Pharmacy Gold INC., St. Paul, MN, US.

OBJECTIVES: To determine if academic detailing could influence physician prescribing patterns for HMG-CoA reductase inhibitors in the setting of a 1.5 million patient BlueCross BlueShield plan. Methods: This product strategy was the second phase of our Prime Therapeutics¨ program. We targeted top prescribers as well as top clinics throughout the state. The focus of the strategy was to shift patients from Mevacor and Zocor to Lescol, our preferred agent, or Pravachol as a cost effective alternative. The clinics were contacted by a clinical pharmacist and a presentation was given which detailed the benefits of the program and cost effectiveness of the preferred products. Quarterly reports of clinic and individual physician prescribing patterns were presented to the group. Follow-up material included patient lists of those taking the targeted agents. Results: We report on the results of 16 clinics which included 216 physicians who were given the presentation before the end of 3rd quarter 1996. The market share for Lescol changed from 16.9% to 20.1% (19% relative increase) from 2nd quarter to 3rd quarter. Pravachol market share rose by 0.6%, while Mevacor and Zocor had decreases in market share of 3.5% and 0.2% respectively. This market share shift corresponded to a projected annual savings of $70,000 for these 16 clinics. When extrapolating to the remainder of our targeted clinics, the total cost savings to the plan would be approximately $300,000 annually. Conclusion: Academic detailing is an effective technique to influence physician prescribing patterns and shift market share. This method has high physician acceptance and is cost effective to the PBM.

Abstract #: EO10


Torborg SA*, Benson SR , Juge D and Waltermire RD, Pharmacy Gold Inc., St. Paul , MN, U.S.A.

OBJECTIVE: To explore the utility of academic detailing for lowering drug expenditures in the angiotensin converting enzyme (ACE) inhibitors class and addressing under-utilization of this class for appropriate therapeutic indications. METHODS: ACE inhibitor prescribing represented the initial focus of a multi-step academic detailing program designed to influence physician prescribing within a 1.5 million member BlueCross BlueShield Plan. Physicians and Clinics in the top 20% for ACE inhibitor costs were targeted. The strategy was to shift prescribing from Vasotec and Capoten to a less costly alternative such as Zestril. Information about ACE inhibitor use for patients with congestive heart failure, post myocardial infarction and diabetic nephropathy was presented along with the cost information to stimulate appropriate utilization for these indications. Clinic and individual prescribing data were presented at all meetings. Each physician received an individual report and patient list following the meeting. Physicians that were not part of a targeted clinic were contacted directly. RESULTS: By the end of 3rd quarter 1996 we had presented to 70/379 (18.5%) of the clinics in the HMO network. Of the top 20% of targeted physicians, we had contacted approximately 80%. Overall market share for Zestril improved from 28.2% to 35.3% with a decrease for Vasotec of 36.6% to 30.5%. Capoten decreased from 11.1% to 1.8%. Overall ACE inhibitor utilization for the plan increased by 10% in this same period. CONCLUSION: Academic Detailing is an effective and well received method for affecting physician prescribing patterns in the ACE inhibitor class. Significant market share changes and cost savings were realized while utilization of the class increased through the first three quarters of 1996.

Abstract #: EO11


Benson, SR, Burns, KD, Waltermire, RD, Pharmacy Gold, INC., St. Paul, MN, US.

OBJECTIVES: To determine if operation of our retrospective drug utilization program called Gold Advantageń could result in direct cost savings for a 5,000 member employer group. METHODS: Pharmacy and medical claims data for all patients in the employer group were analyzed on a monthly basis. Five areas for intervention are examined: 1) drug-disease interactions, 2) drug-drug interactions, 3) over and under-utilization, 4) duplication and 5) duration. Each case is assigned a risk priority score (RPS) which estimates the likelihood of the patient being hospitalized due to the alert. The top 10% of cases are then reviewed by a clinical pharmacist. If the alert is deemed clinically significant, a letter is mailed to the prescribing physician and the pharmacist. This letter includes an explanation of the program, the patient's drug profile and an evaluation form to be returned to Pharmacy Gold, Inc. Using the evaluation forms, we can estimate the cost savings based on therapeutic intervention by the provider. RESULTS: As of 10/1/96, we had analyzed approximately 117,000 claims at a cost of $18,720.00 to the employer group. We mailed out 249 alert letters to providers with 164 (65%) returning the evaluation forms. Of these 164, 29.9% responded that they would change or stop the patient's therapy. The total potential cost savings for all 249 letters was $157,500.00. Therefore, the cost savings can be estimated as follows: $157,500 X 164/249 X 0.299 = $31,017.00. The direct benefit to the employer group is then $31,017.00 - $18,720.00 = $12,297.00. CONCLUSION: The use of Gold Advantageń as a retrospective drug utilization program can result in substantial direct cost savings for the client.

Abstract #: EO12


DuChane, J, Grebosky, B, Gunter, M, Brixner, D, Von Worley, A, Lovelace Clinic Foundation, Albuquerque, NM, Novartis Pharmaceutical Corporation, Summit, NJ, US

OBJECTIVE: The purpose of the study was to select and evaluate the reliability of an indicator or group of indicators that can be used to electronically identify managed care patients for inclusion in an epilepsy outcomes study. METHOD: Patients were identified electronically whose records over an 18-month period included: 1) use of an antiepileptic medication and/or, 2) a diagnosis pertaining to epilepsy or convulsions, and/or 3) a procedure relating to epilepsy. The selected patients were then grouped into 12 subcategories such as, those using one antiepileptic medication alone, those using one medication combined with the presence of one diagnosis code, those with the presence of one diagnosis code alone, etc. A predictive value was assigned to each of the 12 subcategories. Verification of a patient's seizure disorder was provided by correspondence with the patient's physician and through manual chart reviews. RESULTS: A total of 2,756 patients were identified electronically as potentially having seizure disorders. Preliminary results indicate that the most reliable indicator was the combination of all three electronic indicators. Diagnosis codes alone yield greater predictability than do medications alone; procedure codes alone are highly unpredictable. CONCLUSIONS: Patients with seizure disorders can be identified electronically by combining information relating to three primary indicators with physician verification and chart reviews. The use of medication alone was not found to be a reliable method for identifying persons with seizure disorders.

Abstract #: EO13



Bull, S, Shoheiber, O, Novartis Pharmaceuticals, Summit, New Jersey, US.

Therapeutic interchange can be associated with costly physician visits and patient dissatisfaction while titrating to an effective dose. To help minimize these cost, enormous amounts of money are often invested to determine therapeutically equivalent doses between target drugs. Although randomized, double-blind, controlled clinical trials are considered by many as the gold standard in establishing equivalent dosing efficacy between drugs, they are extremely time consuming and expensive to conduct. OBJECTIVE: The goal of this study is to test a retrospective methodology for determining effective dosing ratios between drugs. METHODS: The methodology uses retrospective longitudinal analysis of a large pharmacy claims database. The analysis involves the following steps: 1) Identify patients who switched between the two targeted medications; 2) Include patients with longitudinal data that indicate "pre" and "post" switch dose stabilization; 3) Calculate the preswitch-postswitch dose ratios. The following criteria can affect the interpretation of the switch ratio: 1) Long gaps between refills. 2) Unstable doses before and after the switch. 3) Titration of other medications during the switch. 4) Overall integrity of claims data. RESULTS: From a sample of 38 patients switched between Lotensin and Vasotec, approximately 20% of the data was of value in determining a dosing ratio. From this data, the milligram to milligram dose ratio between Lotensin and Enalapril was found to be 1.12:1. This finding trends well with conversion ratios that were found to be effective in other conversion protocols. DISCUSSION: This methodology can provide organizations with an easy systematic way to establish dose switching guidelines. Shortcomings associated with this approach are the lack of clinical efficacy indicators, and the need of a large database to identify a significant number of patients switched between the target medications.

Abstract #: EO14

Economic Outcomes Associated with Initial Treatment Choice in Depression: A retrospective Database Analysis

Edgell, ET, Hylan, TR, Eli Lilly and Company, Indianapolis, IN. US

Objective: The purpose of this study was to examine the economic outcomes associated with initial treatment choice faced by the physician and patient after a diagnosis of depression. Method: A database containing 1990-93 patient-level clinical and financial information was used to classify patients into one of four treatment cohorts: diagnosis-only, psychotherapy, drug therapy, and combination therapy. Sample selection bias was accounted for using a two-stage process where treatment choice was estimated using a multinomial logistic regression model in the first stage. From this model, a decreasing function of the probability of receiving treatment, the Inverse Mills Ratio (IMR), was calculated. The IMR was then included as a covariate in second stage regression equations estimating total costs. Log predicted costs from the second stage were compared across all observations and in observations whose initial provider was a non-mental health physician to determine the relative costs associated with each cohort. REsults: Significant differences (p < 0.008) in total costs were found between the combination therapy and psychotherapy cohorts in the analysis that included all observations (n = 9,110). In the analysis that included patients who initiated therapy with a non-mental health physician (n = 2,673) the drug therapy cohort was found to be significantly more costly as compared to the diagnosis-only cohort. Conclusions: There is not a significant difference in total health costs between depressed patients who initiate treatment with psychotherapy and depressed patients who initiate treatment with drug therapy.

Abstract #: EO15


Holzer S, Marder W, Ozminkowski RJ, Peters J. The MEDSTAT Group

OBJECTIVE: The purpose of this research was to analyze the prevalence, utilization, and expenditures for patients with 12 different diseases and conditions thought to be common and costly to private payers in the United States: asthma, coronary artery disease, cholecystitis/cholelithiasis, depression, diabetes, hypertension, low back pain, otitis media, pneumonia, sinusitis, upper respiratory infection, and urinary tract infection. We sought evidence that might identify opportunities for minimizing variation in healthcare costs. METHOD: We used the linked Medical-Rx MarketScan¨ database from 1992 through 1994. We identified study patients based on their having been diagnosed with each disease or condition under study, using a detailed clinical algorithm. Disease-specific and general medical and drug claims data were aggregated for each patient. We compared patients within each disease or condition group to other persons without the disease who submitted claims to MarketScan¨ clients. RESULTS: Prevalence of these conditions ranged from a low of 7 per 1,000 for cholecystitis and cholelithiasis, to a high of 195 per 1,000 for upper respiratory infection. Combined, these conditions represent more than 330,000 total cases. The disease-specific costs of these conditions totalled more than $164 million in 1994, or nearly 22 percent of the total healthcare expenditures for MarketScan enrollees. The total healthcare payments for the patients with each disease or condition were 1.5 -2 times higher than the payments for patients without each disease or condition. Variation in prevalence and expenditures over time and evaluation of the proportion of patients contributing to the proportion of healthcare dollars will be presented. CONCLUSIONS: The prevalence, utilization, and expenditure variation among patients with these disease leads to differential opportunities for disease management approaches.

Abstract #: EO16


Strauss, BM, Smith, ME, Schwartz, L, Linde-Zwirble,W, Newbold,RC, Lidicker, J

Bayer Corporation, West Haven, CT, Health Process Management, Doylestown, PA. US

Controlled trials have strict selection criteria to limit case-mix differences in order to evaluate specific interventions. However, this may result in study populations that are systematically different from 'real world' patients, treated in clinical practice. The utility of pharmacoeconomic information from such prospective studies is at best suspect. Retrospective studies using billing claims data sets can illuminate this situation. OBJECTIVE: To examine differences in demographic characteristics of hypertensive patients in PA Medicaid when applying a set of prospectively designed exclusion criteria in a retrospective cost study. METHOD: 15-months of 1994-96 PA Medicaid claims data were reviewed for hypertensive patients receiving at least one prescription for a Calcium Channel Blocker (CCB). A subset of patients were selected for inclusion in a cost study. Entry was restricted by age, eligibility, choice of CCB, a CCB-free period, and change in CCB. Patient demographics were examined for five subgroups.


Set Number % Female % Black Me an Ag e (ye ars )

1-Starting 44,997 70.9% 29. 9% 61. 6

2-All criteria but age 1,058 69.8% 22. 0% 60. 3

3-All criteria 595 60.7% 27. 2% 49. 8

4-Adalat CC 176 60.2% 33. 0% 48. 5

5-Norvasc 419 60.8% 24. 8% 50. 4

Selecting age results in a large decrease in percent Female. Selecting all criteria but age shows a significant decrease in percent Black. CONCLUSION: This study reveals how prospectively designed retrospective studies relying on claims data can be used to evaluate patient exclusion criteria and selection biases inherent to controlled clinical trials. Our selection process did produce a homogeneous sample representative of the overall population.

Abstract #: EO17


Einarson TR, Addis a, Iskedjian M, Shane L. Faculty of Pharmacy, University of Toronto

OBJECTIVES. Cost effectiveness analysis of drug groups in treating adults for acute major depressive disorder including SNRIs (venlafaxine), SSRIs (fluoxetine, fluvoxamine, sertraline, paroxetine), and TCAs (amitriptyline, imipramine, desipramine, nortriptyline). METHOD: Decision tree model over six months was constructed using an expert panel. The analytic viewpoint was for the Ontario Ministry of Health as payer for all direct costs, derived from standard lists, including cost of drug, medical care, laboratory costs, and treating adverse events (ADRs). Rates of success and dropouts were determined from a meta-analysis of published ramdomized controlled trials. Medine, Embase, and IPA were searched from 1984-1996, and references from retrieved articles and reviews. Outpatients and inpatients were analyzed separately. For SNRIs and TCAs, the backup drug was SSRIs; for SSRIs, it was SNRIs. Expected cost and incremental cost per success were calculated. Threshold analysis and Monte Carlo simulations tested sensitivity. RESULTS: Meta-analysis identified 56 treatment arms from 36 randomized controlled trials on 2953 patients (2380 outpatients and 573 inpatients). SNRIs had highest success rates and lowest dropout rates. For outpatients, expected costs were $4774, $5072, and $6181 for SNRis, SSRIs, and TCAs, respectively. Cost per success was $6044, $6633, and $9035, respectively. For inpatients, expected costs were $12,537, for SNRIs, $12,914 for SSRIs, and $13,400 for TCAs; cost per success was $16,702; $17,642; and $19,423, respectively. SNRIs were dominant for all incremental cost effectiveness analyses. Sensitivity analyses confirmed the robustness for outpatients, but sensitivity for outpatients on success rates. CONCLUSIONS: Venlafazine is a cost effective drug for the treatment of major depressive disorder in both outpatients and inpatients in Ontario.

Abstract #EO18


O'Brien JA, Shomphe LA, Caro JJ, Caro Research, Boston, MA. USA

Objectives. To examine the role behavioral disturbances (BD) play in transitioning a resident with dementia to a category of higher resource intensity. Methods. The Massachusetts (MA) Medicaid Nursing Home Case-Mix system determines ten per diem reimbursement rates based on a management minutes score that covers need for assistance with activities of daily living and allows special attention categories, such as BD. After assigning a score on admission, it is reassessed quarterly. Data from 1995 were examined to identify the prevalence of BD in each of the ten rate levels and those patients who transitioned to that level solely due to BD. Physician consultation data were also explored as the cost of visits is not included in the per diem rate. Results.. Of the 13,130 patients with dementia, 20% had documented BD. The prevalence of BD in this population ranged from 0.9% in the lowest rate level to 17.2% in the highest one. Of those with BD, 70% were in a higher care level solely due to BD. This represents an additional annual cost of 1995 US$667,243 to Medicaid. Patients with BD required 12.5% more medical consultations than those with dementia without BD. Conclusion. Although institutionalization itself is often considered to be the final major cost threshold for patients with dementia, it is clear that differential care costs can be quantified relating to specific disease elements, such as BD. BD is an identifiable cost factor in long term residential care. Controlling BD in a institutionalized dementia population could lower the cost of that care.

Abstract #: EO18


O'Brien JA, Shomphe LA, Caro JJ, Caro Research, Boston, MA. USA

Objectives. To examine the role behavioral disturbances (BD) play in transitioning a resident with dementia to a category of higher resource intensity. Methods. The Massachusetts (MA) Medicaid Nursing Home Case-Mix system determines ten per diem reimbursement rates based on a management minutes score that covers need for assistance with activities of daily living and allows special attention categories, such as BD. After assigning a score on admission, it is reassessed quarterly. Data from 1995 were examined to identify the prevalence of BD in each of the ten rate levels and those patients who transitioned to that level solely due to BD. Physician consultation data were also explored as the cost of visits is not included in the per diem rate. Results.. Of the 13,130 patients with dementia, 20% had documented BD. The prevalence of BD in this population ranged from 0.9% in the lowest rate level to 17.2% in the highest one. Of those with BD, 70% were in a higher care level solely due to BD. This represents an additional annual cost of 1995 US$667,243 to Medicaid. Patients with BD required 12.5% more medical consultations than those with dementia without BD. Conclusion. Although institutionalization itself is often considered to be the final major cost threshold for patients with dementia, it is clear that differential care costs can be quantified relating to specific disease elements, such as BD. BD is an identifiable cost factor in long term residential care. Controlling BD in a institutionalized dementia population could lower the cost of that care.

Abstract #: EO19


Ikeda, S, Inoue, S, Oliver, AJ, Ikegami, N, Department of Health Policy and Management and Department of Pharmacy, School of Medicine, Keio University, Tokyo, Japan

In Japan, as from August 1992, the submission of materials derived from pharmacoeconomic evaluation have been accepted at the time of new drug price negotiation. The pharmaceutical companies generally consider their submitted materials to be private information. Moreover, it is not clear how the government use these materials in their policy decisions. OBJECTIVE AND METHOD: In this paper, for the purpose of understanding the present condition of pharmacoeconomic evaluation in Japanese pharmaceutical companies, the results from an investigative interview of selected employees at twenty-two pharmaceutical firms are outlined. Also, an analysis of how the submitted economic data is reflected in drug prices is undertaken. RESULTS: Our analyses clarifies that in recent years an economic evaluation has been submitted for most new pharmaceutical products. However, a clear connection between the submitted economic materials and the issuing of a supplement to deserving new products was not found. For example, the prices of sixty-eight drugs with similar effectiveness were singled out by using a comparative method. Of these sixty-eight products, forty-eight (71%) had undergone a submitted economic evaluation. Five of the sixty-eight drugs were reported to offer advantages in terms of added innovation or added usefulness, but an economic evaluation had been submitted for only two of these five drugs. CONCLUSION: We conclude that there is an urgent need to construct a system in which the results of pharmacoeconomic evaluation can be appropriately used in government pricing decisions.

Abstract #: EO20

Cost-Effectiveness of Azithromycin for the treatment of Acute Otitis Media in a rural Population

Sansgiry Sujit S, Sansgiry Shubhada S, Cady Paul S, Weldezghi Bernardos K, College of Pharmacy, Idaho State University, Pocatello, ID. US

Acute Otitis Media (AOM) characterized by the presence of fluid in the middle ear affects almost 50% of children below the age of three. Amoxicillin, used predominately as the first line therapy for the treatment of AOM, could fail either due to bacterial resistance or poor patient compliance. Azithromycin is effective against beta-lactamase producing bacteria, has a short duration of treatment, and once a day dosing regimen. Objective: The purpose of this study was to compare the cost-effectiveness of azithromycin with amoxicillin, trimethoprim/sulfamethoxazole (tmp/smz), and erythromycin/sulfisoxazole (ery/sulfa) respectively, in the treatment of AOM. Method: A decision analysis approach was used to develop models using three drugs as initial therapies for the treatment of AOM. Idaho Medicaid Database (IMD), a retrospective antibiotic prescription data and information obtained from literature were evaluated. Efficacy rates were calculated by two methods using weighted means for cure rates reported in the literature and success rates calculated from the IMD. Drug cost was calculated by averaging retail prescription price obtained from four local pharmacies and average cost per prescription calculated from the IMD. Cumulative cost for each therapy was calculated by the 'folding back process' and a constant second line therapy using amoxicillin/clavulanate. Results: The results from the IMD and literature were consistent. Cost for the treatment of AOM using azithromycin was higher ($29.89 -- $34.03) compared to amoxicillin ($9.79 -- $20.00), tmp/smz ($10.73 -- $18.96), and ery/sulfa ($23.73 -- $26.20). Conclusion: The results indicated that as a first line therapy, azithromycin may not be cost-effective when compared with amoxicillin, tmp/smz, and ery/sulfa in the treatment of AOM.

Abstract #: EO21


Taylor, TN, Meyer J. College of Pharmacy, University of Iowa, Iowa City, Iowa.

OBJECTIVE: We sought to estimate costs of treating stroke over a two year period for persons with subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), and ischemic stroke (ISC). METHODS: Medicare claims data from a 5% sample of Medicare beneficiaries were used for this study. Persons aged 65 and older with a 1990 hospitalization with a principal diagnosis of SAH (ICD-9-CM code 430), ICH (codes 431-432), or ISC (codes 433- 434 or 436) were classified as stroke cases. We randomly selected two percent of Medicare beneficiaries in the 5 percent sample to serve as a control population. The first stroke hospitalization in 1990 served as the index event and cases were followed for seven quarters beyond the quarter of the initial hospitalization. Controls were also followed for two years. Cost of hospital inpatient, outpatient, Part B, Skilled Nursing Facility, and home health care services were estimated from the Medicare data. Medicare cost-to-charge ratios were used to estimate inpatient hospital costs. RESULTS: 18,768 cases met our criteria for stroke (305 SAH, 1,905 ICH, and 16,558 ISC). The mean first year total costs were $29,898 for SAH, $18,542 for ICH, $16,809 for ISC, and $3,134 for controls. The mean second year total costs were $2,962 for SAH, $3,710 for ICH, 5,796 for ISC, and 3,255 for controls. First year costs varied significantly by type of stroke and survival status. Inpatient hospital costs accounted for the majority of costs in each year for both cases and controls. CONCLUSION: Medical care costs incurred in the two years following a stroke are substantial and vary by type of type of stroke and survival.

Abstract #: EO22

Confidence in Cost of Illness Analysis: Focus on Asthma

Malone, DM, Smith, DH, Lawson, KA, University of Colorado Health Sciences Center, Denver, CO. USA, and University of Washington, Seattle, WA. USA.

Cost of illness (COI) analyses are often performed using multiple data sources, with numerous assumptions and typically provide only point estimates. Estimates are often taken at face value and may lead to improper public policy decisions. OBJECTIVE: The purpose of this study is to contrast U.S. asthma- specific COI estimates from a single data source, the 1987 National Medical Expenditure Survey (NMES), to those determined by Weiss et al. METHODS: The NMES data were used to identify persons with asthma using ICD-9 codes. Confidence intervals around point estimates were constructed by using the NMES weighting and stratification variables with SUDAAN¨ statistical software. Estimates from Weiss et al. and NMES were inflated to 1994 dollars. RESULTS: Using NMES data, we estimate the total cost of asthma to be $5.8 billion (95% confidence interval(CI): $3.6 to $8.1 billion). The Weiss et. al. estimate is $7.8 billion in 1994 dollars. While the two overall estimates are similar, there are differences in the proportion of costs accounted for by individual cost categories, particularly in estimates of indirect costs. The NMES total indirect cost estimate is $674 million (95% CI $270 million to $1.0 billion) versus over $2.9 billion in Weiss et al. Some of this difference is because Weiss et al. included a mortality cost and the NMES analysis did not. Even after accounting for the mortality cost difference, the Weiss et al. indirect costs are still greater by more than $1.3 billion. CONCLUSION: The results demonstrate that estimates of the cost of asthma in the U.S. can varying significantly depending upon the data source and assumptions.

Abstract #: EO23


Tretiak, R,* Grenier, J-F,† RiviŹre M.* *Quintiles Canada, Montreal, PQ, Canada; †Berlex Canada Inc., Lachine, PQ, Canada

OBJECTIVE: Multiple sclerosis (MS) is a common neurologic disease in young and middle-aged adults affecting approximately 35,000 Canadians. The objectives of this study were to estimate the cost of MS from the perspective of Canadian society, annually and over a patient's lifetime and to determine how the cost of the disease evolves with disease progression. METHODS: Patients were consecutively recruited by neurologists in 14 MS outpatient clinics across Canada. They were classified by the Expanded Disability Status Scale (EDSS) into three groups: mild (EDSS 2.5), moderate (EDSS=3.0-6.0) and severe (EDSS 6.5). Sociodemographic, clinical and resource utilization data were collected for the three months prior to patient inclusion. Costing of resources consumed was performed from Ministry of Health, private third party payers, patient and societal perspectives. Average Canadian costs ($CDN 1995) were valued from available provincial data. Using published Canadian epidemiological data on the evolution of the disease and the cost of MS by severity group, a lifetime cost model was developed.

RESULTS: A total of 198 patients were included in the analysis (mild: n=62, moderate: n=68 and severe: n=68). Costs increased over time with increasing EDSS scores from all perspectives. Results show that patients bear most of the economic burden of the disease. Indirect costs, including lost daily activity/leisure time and lost productivity, were the major societal cost drivers.

CONCLUSIONS: This burden of illness study provides a basis for cost-effectiveness analyses of new therapeutic interventions for MS. The importance of lost time as a cost driver suggests that MS patients' quality of life may be increasingly affected with disability progression.

Abstract #: EO24



Abernathy, GB, Kimberlin, K, Jaggers, R

Fairfax Hospital, Falls Church, VA. US

ISSUE: Several drugs coming onto the market have potential financial impact of over a million dollars/institution in 2-3 years time. Traditional Pharmacy & Therapeutics Committee review is no longer sufficient for these complex clinical/financial decisions. PROCESS: Reopro blocks the final common pathway of platelet aggregation and is used to prevent acute reclosure of coronary arteries in patients undergoing PTCA at a cost of $1350/patient. Our institution performs 1,300 PTCA procedures/year. An emerging drug therapies team (pharmacists, physicians, nurse specialists, financial analysts, administrators, cardiac cath technicians) was formed to decide whether to make Reopro available in the institution and for what indications. Financial impact was estimated based on physician assessment of usage, patient volume, and indications in clinical trials. Projected positive and negative outcomes were derived from clinical trial information. DECISION: After the initial analysis, the decision was made not to add Reopro to the formulary in 1995. The availability of additional clinical trial results with extension of positive outcomes and minimization of negative outcomes resulted in the decision to add Reopro in October 1996. In addition, hospital cardiology procedure-specific outcomes data became available in early 1996 improving our ability to contextualize clinical trial results. The emerging drug therapies team developed the structures needed to optimize outcomes (standard orders, physician and staff education, changes in clinical practices) and an outcomes monitoring plan to allow reassessment of cost-benefit every 3-6 months. The team will be used to analyze expected additional indications for Reopro. This model is being used for other emerging drugs with potentially large financial impact. The model could be adapted for use in a broad spectrum of healthcare provider or payer settings.

Abstract #: EO25


Kilburg, A, Bachinger, A, Rychlik, R

Institute of Empirical Health Economics, Altenberger-Dom-Strasse 16, D-51519 Odenthal, Germany

Ceftriaxone, Ciprofloxacin, Meropenem and Piperacillin-Tazobactam are some of the most preferred antibiotics to treat pneumonia. In a retrospective analysis the cost structure of above different antibiotic treatments was established. OBJECTIVE: This analysis centred around the question which of these alternative therapies incurs the best cost-effectiveness ratio. The comparison of costs and benefits was made on the basis of a cost-effectiveness analysis. METHOD: A decision tree with four endpoints was constructed: (1) successful therapy with adverse drug reaction (2) successful therapy without adverse drug reaction (3) failed therapy (4) death. Data regarding efficacy, duration of treatment and casualities underlying were taken from clinical studies. Health care costs included in the model were all medical costs (hospitalization, drug costs etc.) and indirect costs (absence from workplace, burial allowance). The benefit was defined as the cure rate of pneumonia. The following benefit rates were revealed in clinical studies: Ceftriaxone 0.6136, Ciprofloxacin 0.6939, Meropenem 0.8750, Piperacillin/Tazobactam 0.9549. For a better comprehension of the results a sensitivity analysis was conducted to determine the influence of the efficacy on the effectiveness adjusted costs and the effects of the frequency of adverse drug reactions using Piperacillin/Tazobactam. RESULTS: Therapy with Ceftriaxone resulted in total costs of 11,942.47 DM per successfully treated patient. Therapy with Ciprofloxacine incurred total costs of 13,352.41 DM, and therapy with Meropenem caused total costs of 7,720.39 DM. Concerning the therapy of pneumonia with Piperacillin/Tazobactam, total costs of 6,538.28 DM per successfully treated patient were calculated. CONCLUSION: This decision tree analysis reveals that therapy with Piperacillin/Tazobactam has the best cost- effectiveness ratio.

Economic and Outcomes Issues in Clinical Trials, Chart Reviews, Surveys, and Long-term Care

Abstract #: LC01


Alloul, K, Lafortune, L. Health Economics Department. Hoechst Marion Roussel Canada Research Inc, Laval, Que, Canada.

Suprefact Depot is a new 8 week LH-RH agonist formulation recently approved in Canada for advanced prostatic carcinoma. Like other LH-RH agonists, it has demonstrated equal efficacy and safety compared to orchidectomy. OBJECTIVE: The total direct medical cost of LH-RH agonist therapy was evaluated from the perspective of canadian provincial ministry of health and provincial drug formularies. METHOD: A cost-minimization compared Suprefact Depot to Lupron 30 day and Zoladex 12 week depot formulation. Drug acquisition cost, pharmacy mark-up on product, pharmacist dispensing fee, urologist visit fee and physician administration fee were entered into the cost equation. For each province, costs over a monthly, yearly and remaining life (2.5 years) time horizon were computed. Sensitivity analyses were performed to account for variations in practice patterns and in compliance to recommended dosing intervals. RESULTS: Based on national average costs, the use of Suprefact Depot compared to Zoladex and Lupron respectively results in monthly cost savings to the provincial formularies of $26.31 and $34.81 per patient; in 1996 yearly maximum aggregate cost savings to the drug plan of $77 186 and $104 028, based on current community pharmacy prescription numbers; and in annual cost-savings to the ministry of health of $279 and $525 respectively. Only in the unlikely situation of compliance to Suprefact 8 week dosing interval and non- compliance to Zoladex 12 week dosing interval for all patients receiving one of these two agents does Suprefact Depot become only slightly ($1.15 a month) more expensive than Zoladex. CONCLUSION: Amongst LH-RH agonists, Suprefact Depot remained a cost-saving treatment strategy for all provinces and for all scenarios envisioned.

Abstract #: LC02


Williamson, T.E.. Bence-Bruckler, I., & Kearns, B., Pharmacia & Upjohn, Toronto, Ontario, Canada

Hospitals face challenges in shifting treatments to outpatient settings to combat bed closures and reduced budgets. Outpatient treatment of selected AML with idarubicin oral may introduce institutional savings. OBJECTIVE: The study purpose was to examine costs of using oral versus intravenous idarubicin in treatment of acute myelogenous leukemia, assuming, for the purposes of this report, equal efficacy of oral and intravenous idarubicin. METHODS: A review was conducted at the Ottawa General Hospital to determine the number of cycles of AML chemotherapy administered over one year, and what proportion of these cycles contained idarubicin. Resource utilization and actual costs incurred by patients receiving idarubicin therapy were determined from records as captured by the institutional on- line case-costing system, and were utilized in the cost determinations. For the propose of this study, the costs retrieved included nursing and pharmacy costs, which had incorporated into them overhead costs for bed stay, building maintenance, etc. The impact of switching to oral idarubicin was modeled based on reductions in both nursing and pharmacy requirements. RESULTS: Intravenous idarubicin had a per patient cost of $4,295 (nursing 22.2%; pharmacy 77.8%) in an inpatient setting and $3,767 (nursing 10.0%; pharmacy 90.0%) in an outpatient day car unit. The corresponding costs for oral idarubicin were $4,195 (nursing 22.8%; pharmacy 77.2%) and $3,574 (nursing 10.6%; pharmacy 89.4%), respectively. CONCLUSIONS: This study demonstrates that there are potential savings in switching eligible patients from intravenous idarubicin inhospital to oral idarubicin in an outpatient day clinic of $721. Oral idarubicin may potentially save money in those patients eligible for outpatient day care treatment of AML (patients treated with curative intent would not qualify).

Abstract #: LC03

Cost-Effectiveness of Cyclophosphamide, Doxorubicin, and 5-Fluorouracil in Pre- menopausal Women With Lymph Node Positive, Early Stage Breast Cancer

SUng JCY,1 Hay J,1 Louie sG,1Popovian R,2 1Department of Pharmaceutical Economics and Policy, School of Pharmacy, University of Southern California, LOS ANgeles, CA. US.

2PFizer Inc., New York, NY. US.

Breast cancer is the most common female malignancy in the United States. Systemic adjuvant chemotherapy has been proven to prolong survival in breast cancer patients. One such regimen which combines cyclophosphamide, doxorubicin, and 5- fluorouracil (CAF) is considered a clinically effective adjuvant therapy. Objective: To evaluate cost-effectiveness (CE) of CAF therapy in a hypothetical cohort of pre-menopausal women with lymph-node positive, early-stage breast cancer. METHOD: A medical decision algorithm modeling was developed in accordance to possible clinical outcomes. Clinical probabilities used in this model were merged with medical cost data to evaluate the cost-effectiveness of CAF therapy. The clinical outcomes and health care utilization information were gathered using economic, epidemiological and clinical data from published studies. The average length of follow-up was five years. The primary measure of cost-effectiveness is incremental cost per life- year saved. The perspective of the study was societal. However, indirect costs were not included due to insufficient data. Medical costs were expressed in 1996 US dollars. A discount rate of 3% was incorporated to establish relative weights on deferred outcomes as opposed to immediate impacts. Univariate sensitivity analyses were performed on key parameters in the model. RESULTS: The average total health-care costs in the five-year period associated with CAF treatment of a breast cancer patient was $24,470. The baseline analysis found an average life-year gained of 4 months from CAF therapy at a cost of $48,292 per life-year saved. The results of the model were relatively insensitive to reasonable changes in parameter assumptions. However, the life-year gained and costs have non-linear impact upon the CE ratio. CONCLUSIONS: Although adjuvant CAF therapy is unlikely to produce net savings in medical resources, there is a survival benefit for adjuvant CAF chemotherapy. The cost-effectiveness ratio is high, but within the range of commonly reimbursed medical interventions.

Abstract #: LC04


Kilburg, A, Casser, U, Rychlik, R

Institute of Empirical Health Economics, Altenberger-Dom-Strasse 16, D-51519 Odenthal, Germany

In a clinical study Cametta et al. showed that the antibiotic Piperacillin/Tazobactam/Amikacin combination treatment of Febrile Neutropenic Cancer Patients was more efficient than a combination therapy with Ceftazidime/Amikacin. In a retrospective decision tree analysis the cost structure of these alternative treatments was established. OBJECTIVE: The purpose of this study was to analyse the medical and economic relevance of above alternative antibiotic treatments based on the multicentre clinical study. The comparison of costs and benefits was made on the basis of a cost-effectiveness analysis. METHOD: First a one-patient model was considered, subsequently the results were projected onto the total population of the Federal Republic of Germany. In the one-patient model (Markov model) three states were constructed: (1) successful therapy, (2) failed therapy, (3) death. Health care costs included in the model were all medical costs (hospitalization, drug costs etc.) and indirect costs (absence from workplace). The benefit was defined as the reduction of fever, or the elimination of the pathogenic microorganisms. The success rate of the therapy with Piperacillin/Tazobactam/Amikacin was 61.40 %, the success rate of the therapy with Ceftazidime/Amikacin was 53.85 %. RESULTS: Combination therapy with Piperacillin/Tazobactam/Amikacin resulted in total costs of 16,616.37 DM per successfully treated patient. Therapy with Ceftazidin/Amikacin caused 20,828.36 DM per successfully treated patient. CONCLUSION: This cost-effectiveness analysis reveals that a combination therapy with Piperacillin/Tazobactam incurs lower total costs per succsessfully treated patient. This cost differences is caused by a considerably difference in indirect costs which are reduced to the loss of human capital. The use of the human capital approach in case of the treatment of febrile neutropenia is discussed controversially.

Abstract #: LC05


Kind, P, Nabulsi, A, Maurath, C, Sarocco, P,

BACKGROUND: The measurement of health outcomes is critical to the clinical and economic evaluation of innovative therapies. It is increasingly recognized that such outcomes must take account of both quality and quantity of life. Measures of health-related quality of life that have the capability of capturing patient preferences for health states, enable us to capture outcome data which would not otherwise be included in the evaluation of new treatment. Generic measure of this type, such as the QWB, SIP and SF-36, have been developed for use in a wide range of settings. OBJECTIVE: To test the performance of EQ-5D as a measure of health status in the treatment of patients with HIV/AIDS. METHOD: The EQ-5D is a generic measure that can be used as a profile, and as an index. It defines health-related quality of life in terms of a profile of 5 dimensions – mobility, self care, usual activity, pain/discomfort and anxiety/depression. In two Phase III studies of Ritonavir, EQ-5D was used weekly initially the monthly to monitor patient reported health status. A total of 1434 patients took part in these studies. RESULTS: In a study of symptomatic AIDS patient (CD4 </= 100 cells uL) after 6 weeks of treatment, significant differences were found between experimental and control groups, with patients who received Ritonavir demonstrating drug and observations on EQ-5D. In a study of asymptotic, HIV-positive patients without AIDS-defining illness, there was not a correlation between observations on EQ-5D and clinical improvements in CD4 counts. CONCLUSIONS: Coefficients of reliability and sensitivity indicate that EQ-5D behaves exceptionally well in terms of conventional psychometric measures of performance.

Abstract #: LC06


Russell, MW, Huse, DM, Drowns S, Hartz, SC, Medical Research International, Burlington, MA. US

increasingly being incorporated into late- stage clinical trials. Nevertheless, commercial pressures to streamline drug development and speed up regulatory approval often hinder the implementation of prospective pharmacoeconomic studies. OBJECTIVE: To conduct retrospective economic evaluations of a new intravenous antibiotic based on three pivotal clinical trials. METHODS: Data for economic evaluations were collected for patients at U.S. study centers (hospitals). Resource use and associated costs were not collected during the trials, so it was necessary to collect the necessary data retrospectively through supplemental chart review and follow-up investigation. Since patient follow-up was incomplete for economic evaluation, data collected during the trials were supplemented by review of the medical records of study patients encompassing the entire hospital stay and any readmission within 30 days following discharge. Hospital costs were assigned to trial patients using "rapid estimation" methods, involving models that predict total hospital cost based on key variables, including length of stay, time spent in special care units, number of surgeries, and in-hospital mortality. These models were developed using regression methods and cost data from 66 U.S. hospitals, then validated using data from 33 different institutions. Cost estimates therefore reflected "usual" patterns of treatment in a large sample of hospitals, rather than protocol-induced resource use. RESULTS: Retrospective data collection was completed for 74% of the 711 patients at U.S. study centers. Rapid estimation models proved to be highly accurate in predicting the total costs of hospitalization. CONCLUSION: While theoretically less than ideal, retrospective pharmacoeconomic study designs continue to prove valuable in drug development.

Abstract #: LC07


Milne, Richard J1, Vander Hoorn, Stephen2; Adis International Ltd1; University of Auckland2, Auckland, New Zealand

OBJECTIVES: To compare the antihypertensive and lipid modifying effects and the lifetime effectiveness of 10 years' treatment of patients with mild to moderate hypertension with celiprolol versus atenolol and to estimate the incremental cost effectiveness.

METHODS: Antihypertensive and lipid effects were estimated from a pooled analysis of all published head-to-head comparisons of celiprolol versus atenolol. The baseline 10-year risk of coronary events, stroke events and cardiovascular death were computed from published Framingham Weibull model regression equations. The absolute and relative risk reduction for coronary and stroke events and cardiovascular death were estimated from the same equations as functions of antihypertensive and lipid effects, age, gender and smoking history. Direct medical costs included drug acquisition and treatment of coronary and stroke events.

RESULTS: Celiprolol (285mg/day) and atenolol (91mg/day) have equivalent effects on systolic and diastolic blood pressure, however celiprolol reduces the ratio of total serum cholesterol to HDL-cholesterol by 8.3% but atenolol increases this ratio by 10.3%. Celiprolol is predicted to be equally effective in reducing stroke risk but to be >2-fold more effective than atenolol against coronary risk and to confer a 1.2- to 3-fold greater survival advantage. Over 10 years, the cost of monotherapy with celiprolol versus atenolol is >$400 per patient lower (1996 NZ discounted dollars) despite a 19% higher tablet cost. These findings are sensitive to drug prices and lipid effects but robust to age, gender, the costs of treating coronary and stroke events and the lag time to treatment effect.

CONCLUSIONS: Compared with atenolol, monotherapy with celiprolol is predicted to reduce the risk of coronary but not stroke events, improve survival and cost substantially less over the 10-year treatment period.

Abstract #: LC08


StŚhl E, Liljas B, Pauwels R1

Astra Draco AB, Lund, Sweden

1Afd. Pneumologie, University Hospital, Ghent, Belgium

BACKGROUND: 1004 asthmatic patients (mean age 44 years and mean PEF 102% pred. normal value) in seven countries were randomised into an open, parallel-group study. For 52 weeks, they inhaled 2-agonists and/or corticosteroids either via Turbuhaler or a pMDI. All patients were considered adequately treated with inhaled corticosteroids and/or 2-agonists before inclusion into the study. The clinical conclusion was that budesonide and terbutaline in Turbuhaler offered a superior alternative to corticosteroids and bronchodilators in pMDIs in maintenance treatment of asthma.

COST-EFFECTIVENESS ANALYSIS: Canadian data from the study were used (440 patients being the largest subpopulation). Costs and effectiveness (number of exacerbations and number of days with exacerbation) were compared.

RESULTS: The total annual costs were, on average, CAD 457 less for a patient using Turbuhaler than for one using pMDI (p=0.057). The cost difference was mainly due to lower costs for hospitalization and loss of production. The cost differences for inhaled steroids and 2-agonists were both statistically significantly in favour of Turbuhaler (p<0.001). Both effectiveness variables show that Turbuhaler was better than using pMDI for the one year treatment.

CONCLUSION: Treatment using Turbuhaler is a cost-effective strategy in asthmatic patients using inhaled steroids and B2-agonists.

Abstract #: LC09

Comparison of octreotide and vasopressin in the management of acute bleeding varices: A cost-effectiveness study.

Hui RL, Louie GK, Wilson LS. Department of Pharmaceutical Economics and Policy, University of Southern California, Los Angeles, CA, US and Pharmaceutical Services, California Pacific Medical Center, San Francisco, CA, US.

Octreotide is a therapeutic alternative for the treatment of esophageal variceal bleeding and is reportedly less toxic than treatment with vasopressin. Octreotide is up to three times more expensive than vasopressin but has not been conclusively demonstrated to be clinically superior. OBJECTIVE: The purpose of this study is to determine which agent, octreotide or vasopressin, is more cost effective in the treatment of acute esophageal variceal bleeding with standard practice in a single institution. METHOD: Data were retrospectively collected from review of 21 patients who were admitted for acute esophageal variceal bleeding during 1990 to 1995. Cost analysis was done using length of stay in the intensive care unit, total drug costs, amount of blood transfusions and number of diagnostic tests. The analysis was done using the 1996 dollar figure. RESULT: There were no significant difference between patients' age, sex, cause of variceal bleeding and the overall Child's Grade among the two treatment groups. The vasopressin patient group accumulated a per patient charge of $21,830 with an overall efficacy rate of 60%. The octreotide patient group had a per patient charge of $13,128 with an efficacy rate of 45%. Efficacy was defined as a combination of variables. Vasopressin had an average cost- effectiveness ratio of $48,786 whereas the cost-effectiveness ratio was $18,635 for octreotide. The incremental cost- effectiveness ratio was $60,013, favoring octreotide. A small sample size and potential treatment bias due to the historical sample are weaknesses of this study. The study results were validated using simplified data comparisons of published studies and a marginal cost-effectiveness ratio of these data and showed similar results. CONCLUSION: The results of this study show that octreotide is more cost-effective than vasopressin.

Abstract #: LC10


Williamson, TE, Pharmacia & Upjohn, Toronto, Ontario, Canada.

Hospitals face challenges in shifting treatments to outpatient settings to combat bed closures and reduced budgets. Outpatient DVT treatment with the Low Molecular Weight Heparin, Fragmin¨ can introduce savings. OBJECTIVE: The study purpose was to examine resource utilization in the treatment of DVT in- hospital in a teaching and community setting, and develop a software model demonstrating the impact of switching to outpatient treatment with Fragmin¨. METHODS: Chart analyses were performed in two hospitals in Canada, one teaching and one community. Chart abstraction covered 6 months from January-June, 1996 and all patients treated for DVT in that time. 126 charts were abstracted to examine medical treatment, length of stay (LOS), and tests. A software model was developed where standard costs for each hospital could be applied to each component of resource utilization. RESULTS: 45% of patients in a community hospital were eligible for outpatient DVT treatment vs. only 39% in a teaching hospital. Mean LOS was 9 days in a community hospital vs. 5 in a teaching institute. Based on a cohort of 100 DVT patients, the annual institute savings in dollars and bed days would be $143,910 and 360 days in a community hospital vs. $63,960 and 156 days in a teaching hospital. With bed day cost reduced to zero, days saved remains constant in both types of institutions with total annual savings of $1,560 for a teaching hospital and a cost increase of only $90 for a community hospital. CONCLUSION: This study demonstrates that the higher cost of Fragmin¨ (vs. heparin) is more than offset by the institution savings in tests and bed days. Savings are higher in a community hospital, but could be implemented in many hospitals.

Abstract #: LC11


Chen, C, Danekas, L, Whiting, B, Ratko, T, Vlasses, P, Matusewski, K, University HealthSystem Consortium, Oak Brook, IL. US

OBJECTIVES: This multicenter drug surveillance study was conducted to characterize the current use of intravenous immunoglobulin (IVIG) at academic health centers. Evaluation of labeled and off-labeled indications was based on guidelines established through literature review and expert opinion distributed in May of 1995.

METHODS: Twelve US academic health centers prospectively enrolled up to 50 patients between April and November of 1996. Patient were included if they were receiving IVIG for the first time as an inpatient. Data collected included indications, outcomes, and adverse events associated with IVIG use.

RESULTS: A total of 254 patients were enrolled in the study (males 110, females 144). Non-labeled indications comprised 29% to 100% of IVIG usage at each institution (average 65%). Forty-two percent of patients received IVIG for labeled indications. Of those patients, one-fourth had at least one positive outcome associated with IVIG administration. Based on recommendations from UHC "model" guidelines, the remaining patients (58%) could be categorized into one of three groups: 1) patients who receive IVIG for unlabeled but acceptable indications (12%); 2) unlabeled but possibly indicated if alternatives fail or in specific situations (22%); 3) unlabeled and unacceptable indications (24%). Forty percent of patients receiving IVIG for labeled indications received prophylactic medication in order to prevent adverse events. Only 28% of patients receiving IVIG for unlabeled indications received prophylactic medication. Eight percent of patients in the labeled group reported adverse events, where as 14% of patients in the unlabeled group reported adverse events.

CONCLUSIONS: Results showed that 58% percent of patients were receiving IVIG for unlabeled indications. Forty-six percent of patients received IVIG for indications that were not specifically recommended by UHC "model" guidelines. Economic implications for institutions include closer evaluation of IVIG usage for unlabeled indications to identify areas of inappropriate prescribing of an expensive biological pharmaceutical.

Abstract #: LC12


Authored by: John Doyle and Neeta Sinha

Background: In recent years, chemotherapy is being administered increasingly in ambulatory settings. The shift away from the in-patient setting has been fueled by the cost-minimization goals, changing reimbursement patterns, and by rapid developments in the medicine and technology arena. Hence, acceptability of a drug in today's medical environment, depends on its ability to be administered effectively in an ambulatory setting. Ifosfamide (Ifex, Mead Johnson, Bristol- Myers Squibb Company), a synthetic analog of cyclophosphamide (Cytoxan, Mead Johnson, Bristol-Myers Squibb Company), is one of the most important anticancer drugs today. The adverse event profile of ifosfamide and the requirement of vigorous hydration during the therapy had erroneously led to the general perception that ifosfamide should always be administered in an in-patient facility. Since, it has already been shown that the effectiveness of Ifosfamide based therapy is independent of the setting it is administered in, we conducted an economic analysis to investigate the potential advantage of administering Ifosfamide in the out-patient setting. Methodology: Practice pattern analysis was conducted with clinical data collected from sites where VIP (Vepesid+Ifosfamide+Platinum) combination regimen was administered to Small Cell Lung Cancer patients, in the in- patient and out-patient settings. Information on the treatment pathways, resource utilization, and adverse events management was derived from the practice pattern analysis. Next, activity-based costing methodology was employed to assign appropriate financial values to the identified resources utilized. Cost-Minimization analysis was performed to compare treatment costs among comparators. Multi- Variate (Monte-Carlo simulation) sensitivity analysis was performed to test the robustness of the models under various treatment settings. Result: The treatment in the out-patient setting showed a cost advantage of approximately 18.40%. High in-patient chemotherapy administration cost was the major factor behind the cost difference. The reimbursement analysis showed that the difference between the two settings was insignificant.

Abstract #: LC13


Rough S, Carro G, The Evanston Hospital, Evanston, IL USA

OBJECTIVE: This study was performed to compare patient and cost outcomes of patients receiving oral and intravenous serotonin type three (5-HT3) receptor antagonists for prevention of acute chemotherapy induced nausea and vomiting (CINV). CINV, historically one of the most feared side effects of chemotherapy, can be controlled with 5-HT3 antiemetic therapy. As a result of product availability and clinical practice, 5-HT3s traditionally have been administered intravenously. However, oral 5-HT3 therapy is as safe and effective as intravenous formulation; and tablets offer advantages of convenience, and lower costs. This study describes the process of implementing oral 5-HT3 antiemetic guidelines for CINV prevention. METHODS: Patient and cost outcomes were evaluated in three phases: prior to implementing IV 5-HT3 use guidelines, following IV 5-HT3 use guidelines, and following oral 5-HT3 use guidelines. A total of 162 patients were evaluated. Clinical data collected included patient demographics, chemotherapeutic agent and emetogenic potential, cancer site, acute emesis incidence and nausea severity, 5- HT3 dose and route of administration. RESULTS: Following implementation of oral 5HT3 use guidelines, 64 of 81 patients (80.0%) received 100% oral 5-HT3 prophylaxis. Sixty-two, (97.0%) received granisetron and 2 (3.0%) received ondansetron. There was no statistical difference in the incidence of acute nausea and vomiting in patients receiving intravenous or oral 5-HT3 therapy, 14.1% and 7.8%, respectively. 5-HT3 doses per outpatient visit decreased 31.9% from 1.7 to 1.2 and drug acquisition cost per visit decreased 27.2% from $113.66 to $82.79. 5-HT3 doses per inpatient day decreased 68.0% from 0.16 to 0.05 and cost per day decreased 68.2% from $10.16 to $3.23. Annual drug cost savings to the institution totaled $115,182. CONCLUSION: When compared to intravenous 5HT-3 therapy in the prevention of CINV, oral 5-HT3 therapy is cost-effective in most patients.

Abstract #: LC14



The purpose of this study was to evaluate the economic benefits of routine monitoring of serum vancomycin concentrations and recommend monitoring guidelines. METHOD: An IRB approval was obtained, thereafter we conducted a 5-months retrospective chart review. Using a computer generated list of patients on vancomycin from August to December 1995, thirty-one patients charts were randomly retrieved from the medical record department and reviewed.Twenty four of these patients met the inclusion criteria. The data extracted from the charts were, serum vancomycin concentrations and frequency of determination, dose and duration of vancomycin treatment, and vancomycin untoward effects. Other data recorded were age, weight, height, sex, status of renal function, and other concurrent nephrotoxic or ototoxic drugs. Benefit-to-cost ratio and Return-on-investment for serum vancomycin concentration was calculated . RESULTS:The Benefit-to-cost ratio was 0.696, and the Return-on-investment was -30.4%. A mean of 2.83 + 3.1 serum vancomycin concentrations were obtained per patient during a mean therapy duration of 10.2 +6.8 days. The mean daily dose ( 27.8+ 7.0 mg/kg), peak (21.7 + 6.54mcg/ml ) and trough (7.4 + 5.98mcg/ml ) were within recommended standards. Also 2 (8.3%) patients experienced increased serum creatinine. CONCLUSION: Our results showed that the routine monitoring of serum vancomycin concentration is not cost beneficial. We recommend only one trough serum vancomycin concentration for patients with normal or stable renal functions for a full course of treatment. Patients with unstable renal function may require more than one trough concentration determination and should be evaluated on a case by case basis. Peak and trough concentrations should be reserved for patients with other risk factors for ototoxicrty.

Abstract #: LC15


McGhan WF,1 Corey R,2 Smith MD,3 Szeinbach S,4 and Oates M.1

1Philadelphia College of Pharmacy & Science, 2Novartus, Inc., 3Health Decision Strategies, and 4Univ. of Mississippi

OBJECTIVE: The purpose of this study was to develop and test an efficient instrument for gathering health state preference scores that could be used in quality-of-life adjustments for patients with breast cancer and multiple myeloma. METHOD: Alternative approaches were considered including time trade off, standard gamble and conjoint analysis. The assessment instrument chosen was a visual analog scale. Thirty-one health states were included in the instrument. Non-cancer scenarios were included to allow comparison with other published weights. A standard 100 point thermometer format was used with "0 = Dead (Least Desirable)" to "100 = Healthy (Most Desirable)". The visual analog scales were administered to two different groups of oncologists. Group 1 (n =10) included national experts in breast cancer and Group 2 (n =10) included national experts in multiple myeloma). RESULTS: Some of the general states and their corresponding mean score included: severe headaches (49.4 + 26.3), severe angina (31.3 + 14.0), general weakness (41.9 + 20.4), coma ( 4.7 + 3.3 ), chronic pain (38.3 + 15.8), hip fractures (30.8 + 21.7), and depression (41.4 + 16.5). Some of the cancer related health states included: chemotherapy (43.6 + 18.8), radiation for bone pain (49.6 +19.9), and cancer with metastases (23.5 +16.9). Between the two groups of oncologists slight differences in the perception of their patients' health state preferences occurred with spinal cord compression (p=.032), restricted from strenuous activity (p=.05), and ambulatory patient (p=.05). Further testing of preference assessment alternatives will be done with cancer patients. CONCLUSIONS: This study supports the use of visual analog scales as efficient ways to collect health state preferences for quality-of-life adjustments in breast cancer and multiple myeloma.

Abstract #: LC16


Dempsey, CL, Farley, PA , Shillington, AC, EPI-Q Inc, Oakbrook Terrace, IL

Cheff CM, The University of Illinois, Chicago, School of Pharmacy, Chicago, IL

OBJECTIVE: The objective of this study was to evaluate patient functional status with antiemetic prophylaxis for the prevention of chemotherapy induced nausea and vomiting (CINV). Appropriate documentation of functional status enables health care providers to evaluate a patient's perception of well-being when comparing anti-emetic drug therapies. METHODS: Functional status scores of patients who were administered a 5- hydroxytryptamine 3 receptor antagonist (5- HT3)--either granisetron or ondansetron--as prophylaxis for prevention of chemotherapy induced nausea and vomiting were evaluated and compared using the Functional Living Index-Emesis (FLIE) questionnaire. Pharmacists and nurses from six cancer centers collaborated to administer 230 FLIE questionnaires to 115 outpatients who received either highly, moderately high or moderately emetogenic chemotherapeutic regimens. Patients completed the questionnaire immediately prior to and 72 hours following chemotherapy. RESULTS: Results of the post- chemotherapy FLIE scores showed that patients experiencing emesis documented a much greater negative impact on functional status than those who had nausea alone. The post- chemotherapy scores of patients who did not experience either nausea or emesis were similar to the pre-chemotherapy FLIE scores, demonstrating no impact on daily functioning. Mean pre- and post-chemotherapy FLIE scores were 124.2 and 110.4 for granisetron, and 124.9 and 111.9 for ondansetron, respectively. CONCLUSION: These data suggest that there is no clinically significant difference in the impact on functional status whether patients received granisetron or ondansetron.

Abstract #: LC17


Welch GW1, Jacobson AM1, Bakst AW2, Revicki D3, Marquis P4, 1Joslin Diabetes Clinic, Boston, MA, 2SmithKline Beecham Pharmaceuticals, Collegeville, PA,3Medtap Int., Bethesda, MD, 4MAPI Institute, Lyon, France.

Previous clinical reports have described a complex process of emotional and psychological adjustment that confronts the patient with Non-insulin dependent diabetes mellitus (NIDDM) for whom diet and exercise prescriptions and oral hypoglycemic therapy are no longer sufficient to control elevated blood glucose levels. Although up to 2/3 of NIDDM patients transition to insulin therapy by the 10th year of treatment, little empirical data is currently available on the measurement of psychological aspects of this treatment transition. OBJECTIVES: Evaluate the psychological aspects of treatment transition. METHODS: We compared diabetes-specific psychological functioning of 132 NIDDM outpatients in an examination of the transition to insulin therapy. In this cross-sectional study, patients treated by oral agents only (O, n=40), oral agents plus insulin (O+I, n=16), and insulin only (I, n=76) attending the Joslin Diabetes Clinic took part. Duration of diabetes was 9.4±7.7, 11.5±10.1, and 15.5±9.1 years for the 3 groups respectively. Psychological adjustment was measured by the 20-item Problem Areas in Diabetes (PAID), a validated measure of diabetes-specific emotional functioning. RESULTS: Multiple regression analysis results showed the mean PAID score for the O+I group (38.6±22.9) was significantly worse (with a large effect size) compared to the I group (24.9±21.1), adjusting for age and duration of diabetes (T=2.69, p<.008). The O+I group scored non-significantly worse than the O group (29.1±21.4), with a small to moderate effect size. CONCLUSION: Psychological adjustment is poorer in NIDDM patients transitioning from oral agents to insulin monotherapy. Improved glycemic control in NIDDM patients may reduce the rate of transition to insulin therapy and improve patients' psychological aspects of diabetes management.

Abstract #: LC18


Abbott III, TA and Mucha, LM, Merck & Company, Inc., West Point, PA, USA

One challenge in the treatment of osteoporosis is to determine the optimal use of technology to diagnose disease. Currently the most preferred technique is hip DXA. However, cost and lack of availability may make this test prohibitive for some patients. SCORE (Simple Calculated Osteoporosis Risk Estimation) has been developed to triage candidates for further evaluation including densitometry. SCORE is a six item survey which identifies persons at high risk for osteoporosis. Although a screening tool can reduce the number of unnecessary tests, it may also screen out viable candidates for BMD testing. Thus, it is important to appropriately set the testing threshold. Prior studies have calculated this based on an arbitrary standards, such as 90% sensitivity. OBJECTIVE: This paper examines an alternative approach to determining the testing threshold. Specifically, we use a social welfare function to balance the costs of a false positive (FP), false negative (FN) and testing . METHODS: This objective function can be written as:

max B*TP(T) - C1*FN(T) - C2*FP(T) - C(T)(T)

where T is the testing threshold, B is the benefit of treatment for a patient with osteoporosis, C1 is the cost of a false negative, C2 is the cost of a false positive and C is the total costs of screening. RESULTS: This maximand was tested using data obtained from 845 employees of the Bank of America. SCORE values were calculated and BMD was performed for each study participant. We defined osteoporosis as being 2 or more standard deviations below peak adult mean. CONCLUSIONS: Preliminary analysis shows that use of SCORE as a screening tool would have lead to a reduction in the number of unnecessary BMDs by 30-40%.

Abstract #: LC19

Quality Assessment of Economic Evaluations Published in PharmacoEconomics, 1992-1995

Iskedjian M, Addis A, Bradley-Kennedy C, Ilersich AL, Kruk D, Lanct™t K, Trakas K, and Einarson TR, University of Toronto, Toronto, ON, Canada

OBJECTIVES: To assess the quality of reporting of original economic research articles in PharmacoEconomics from its 1992 inception to the end of 1995, to identify strengths and weaknesses, and analyze trends over time. METHODS: Each regular issue of PharmacoEconomicswas examined for original economic evaluations. Accepted articles were categorized by analytic type and publication year. We applied our previously developed 13-item quality scoring checklist that evaluated items on a scale of 0-4, 4 being correct, 3 acceptable, 2 doubtful, 1 not reported, 0 incorrect, and NA not applicable. Items were weighted equally and averaged for each article to produce an overall score (OS) having a maximum of 4.0. Quality scores were analyzed over time and by study type. RESULTS: We identified 54 articles for analysis. OSs ranged from 1.80 to 3.75, with a mean of 3.01 and SD = 0.47. The item "definition of study aim" scored highest (OS = 3.46, SD = 0.69); "ethical problems discussed and identified" scored lowest (OS = 1.44, SD = 0.92). Only four items scored lower than 3.0. No significant time trend was apparent for OS (R2 = 0.002). Cost-benefit (OS = 3.25, SD = 0.85, n = 5), cost-effectiveness (OS = 3.11, SD = 0.97, n = 27), and cost- utility (OS = 3.29, SD = 0.93, n = 6) analyses scored significantly higher than cost-analyses/cost-of-illness studies (OS = 2.51, SD = 1.14, n = 8). The mean OS for 8 cost-minimization studies was 2.74 (SD = 0.49). CONCLUSION: Despite some weaknesses in particular aspects of economic evaluations in PharmacoEconomics, it has published research papers with acceptable overall quality and adequate methodology.

Abstract #: LC20

Cognitive Function and the Costs of Alzheimer's Disease:

An Exploratory Pharmacoeconomic Study

Ernst, Richard L., Hay, Joel W., Fenn, Catharine, Tinklenberg, Jared,

Yesavage, Jerome A.

Objective: To estimate the dollar savings in illness costs attainable from drug treatments for Alzheimer's disease (AD) that stabilize or reverse patients' cognitive decline.

Methods: Medical and other disease-related utilization data were collected from the caregivers of patients diagnosed with probable AD. The quantities of utilization were priced at national levels to generate measures of illness costs. Costs per patient were then estimated as regression functions of scores on the Mini Mental State Examination (MMSE), which was employed as an index of patient cognitive function. Potential savings in illness costs were estimated by comparing predicted costs at various baseline and intervention-level values of the patient's MMSE score.

Results: The potential savings in illness costs attainable from treatment appear to be small for both mildly and very severely demented AD patients. However, for moderately to severely demented home-dwelling patients having (say) MMSE = 7 at baseline, prevention of a 2-point decline in the score would save approximately $3,700 annually, and a 2-point increase in MMSE score rather than a 2-point decline would save approximately $7,100.

Conclusions: Large savings in the costs of caring for moderately to severely demented home-dwelling AD patients may be achievable from disease interventions that have relatively minor impacts on patients' cognitive status. The results indicated that, even when the treatment effect is small, drugs that can impact the decline in cognitive function would be able to produce numerically significant economic saving in this large subgroup of AD patients.

Abstract #: LC21


Baran, RW, Fastenau, JM, Doyle, JJ, Erwin, WG

Institute for Pharmaceutical Economics, Philadelphia College of Pharmacy and Science, Philadelphia, PA. USA

Osteoarthritis is one of the most chronic conditions in persons 65 years of age and older. NSAIDs are the predominate treatment for osteoarthritis. Although effective, NSAIDS are associated with serious GI complications. The recently completed MUCOSA trial identified 1) age greater than 75, 2) history of previous peptic ulcer, 3) history of GI bleeding, and 4) history of cardiovascular disease as risks for NSAID-induced GI complications. This risk profile is consistent with many LTC residents in whom NSAID use has been identified to be greater than 13%. OBJECTIVE: To determine a correlation between risk factors reported by the MUCOSA trial and the following outcomes in LTC: 1) resource utilization, 2) clinical/biological markers, 3) traditional clinical endpoints, 4) HRQol and 5) resident satisfaction. METHODS: Outcomes associated with NSAID-induced GI complications were identified. A retrospective chart review of 340 residents prescribed NSAIDs was conducted with 800 to be completed. Outcomes according to the five categories above were extracted from resident charts. Outcomes were correlated to MUCOSA risk factors using SAS 6.11. RESULTS: Results show that at least one risk factor was present in all residents, 74% had more than one risk factor. Osteoarthritis comprised 56.8% of all NSAID indications. Perforated duodenal ulcer developed in 21.2% of residents. Pathology/lab usage increased in 8.8% of residents and level of nursing care increased in 4.7% of residents. These preliminary results will be further investigated in the larger population. CONCLUSIONS: The risk factors for NSAID-induced GI complications identified by the MUCOSA trial are predictive of negative outcomes in LTC. In addition, multiple MUCOSA risk factors are common in this population.

Abstract #: LC22


Lemke, P, Joseph, D, LaPalio, L, Stahr, P, Abbott T, Vitalink Pharmacy Services, Allentown, PA, USA and Merck & Company Inc., West Point, PA, USA.

Osteoporosis can be defined as "a disease characterized by low bone mass and microarchitechtural deterioration of bone tissue, leading to enhanced bone fragility and a consequent increase in fracture risk." Epidemiological studies suggest that a significant percentage of women living in nursing homes have osteoporosis. Unfortunately, in many cases these women are not diagnosed or treated. OBJECTIVE: This study evaluated the extent residents could be diagnosed using information already listed on their chart. One can operationally define a chart based diagnosis of osteoporosis as either: an existing diagnosis, comments about osteoporosis on radiological reports, or evidence of fracture. METHODS: A review of 110 female resident's charts at a single home was conducted. The data collected included whether the resident had a current diagnosis of osteoporosis, whether they had comments concerning osteoporosis on radiological reports, and whether they have existing fractures. RESULTS: Of the 110 patients examined, 59 (53%) could have been diagnosed with osteoporosis based on the chart, of those only 16 (27%) had a diagnosis listed, and only 5 were treated appropriately (Vitamin D, Calcium, and antiresorptive agent). CONCLUSIONS: These findings suggest substantial under diagnosis and treatment of osteoporosis in the nursing home and that a simple case finding algorithm could increase the number of residents diagnosed at very little cost.

Abstract #: LC23


Garfield FB, Caro JJ, Caro Research, Boston, MA. USA

Stafford JL, Bristol-Myers Squibb, Princeton, NJ. USA

The outcome of treatment of hypertension, an asymptomatic condition is heavily influenced by the patient's compliance behavior. The health belief model attributes noncompliance to static factors, such as the patients' beliefs about the disease or their inability to pay for treatment. Another model views compliance to treatment as a process in which individuals go through 5 recognized stages of behavior change. When patients interact with their physicians, the patients may be in any one of the 5 stages or even have relapsed and interrupted treatment. Objectives. To determine 1) whether viewing compliance as a process of behavior change would conflict with the more static health belief model 2) whether the behavior change approach would provide measures that were applicable to compliance 3) whether using these measures, researchers and clinicians could develop appropriate interventions to increase compliance. Methods. Literature Review. Results. The review found that the models describe different aspects of compliance with antihypertensive therapy. Outcomes of therapy may be influenced by behavior change as much as by beliefs or the barriers that prevent patients from taking their medication as directed. Conclusion. In order to accurately measure compliance in hypertensive patients, assessment instruments must incorporate elements from the process as well as the static models of compliance behavior. More inclusive measures will differentiate patients, which in turn will allow clinicians to more accurately direct their interventions to each of their patients. Such focused interventions have a greater probability of increasing compliance and improving the outcomes of antihypertensive therapy.

Abstract #: LC24


Tretiak, R,* Breton, M-C,* Grenier, J-F,†RiviŹre M.* *Quintiles Canada, Montreal, PQ, Canada; †Berlex Canada Inc., Lachine, PQ, Canada

OBJECTIVE: To measure quality of life (QoL) in multiple sclerosis (MS) patients in Canada and to assess the QoL impact of disease progression. METHODS: MS patients (198), were recruited on a consecutive basis from 14 MS clinics in Canada and separated into three levels of disease severity, based on their Expanded Disability Status Scale (EDSS) score (mild, EDSS < 2.5; moderate, EDSS 3.0-6.0, and severe, EDSS > 6.5). QoL was assessed cross-sectionally using the 36 Item Short-Form Health Survey (SF-36), a validated generic QoL instrument. Statistical tests were used to compare QoL scores between severity groups and to identify possible relationships between QoL and patient parameters. RESULTS: QoL of MS patients collapses early in the disease when physical disability is not yet a major symptom (EDSS < 2.5). With disease progression, only physical functioning scales show further decreases in QoL.

DISCUSSION: The results of the study suggest that, either the SF-36 is insensitive to the evolution of patient QoL as the disease progresses, that their is an adaptation of the patient to the disease, or, that the immediate shock of diagnosis is so severe that patients do not recover. A disease-specific instrument may provide additional information on QoL. Cross- sectional observation was shown to be an interesting and effective approach to estimating QoL in a chronic disease.

Abstract #: LC25


Strauss, BM, Glick, H*, Smith, ME, Kinosian, B*, Bayer Corporation, West Haven, CT., *University of Pennsylvania, Philadelphia, PA. US

Economic modeling of health care intervention is a framework in which the effect of the intervention on the target disease, in terms of health outcomes is made explicit. This approach provides a tool to answer research questions regarding the effectiveness of health care strategies and also determine efficient allocation of health care resources. Decision analytic models have been used to estimate the costs and effects of therapy for many years and their use is increasing. OBJECTIVE: This paper describes and illustrates the use of decision analytic modeling to evaluate the efficacy and effectiveness of a cholesterol- lowering agent for the primary prevention of coronary heart disease (CHD). The model construct is based on intervention study data from the Lipid Research Clinic Coronary Primary Prevention Trial (LRC-CPPT). Previous CHD models have been based on observations from epidemiological studies, such as the Framingham Heart Study, plus intervention assumptions regarding efficacy and effectiveness. This approach was adopted because information regarding the intervention, patient demographics, and the relationship between changes in cholesterol and patient risk for CHD, which are critical to the model, were not always available. METHOD: The model is based on CHD risk equations, which are directly estimated from the experiences of the LRC-CPPT participants. RESULTS: Concepts such as time delay, maximum benefit, and the cost- effectiveness of the test therapy are estimated. CONCLUSION: Previous work which based the probabilities and risk assessments on assumptions questions the reliability and credibility of the results and therefore confidence in the model. Using intervention data produces a model which is robust and thus allows health care decision makers to base theirdecisions on actual experience.

Applications Workshops

A.P.O.R. Consensus Development Committee’s GAPP (Generally accepted Pharmacoeconomic Principles): Practical issues

Abstract #: AP1


A.P.O.R. Consensus Development Committee

A.P.O.R, through the Consensus Development Committee, is developing methods and guidelines that will become GAAP (Generally Accepted Pharmacoeconomic Principles), a constantly evolving set of standards and principles. The purpose of GAPP is to define a common set of principles, methods, and standards under which all parties can conduct and interpret any type of pharmacoeconomic research. The intent of GAPP is to not prescribe how pharmacoeconomic research should be conducted but to provide a framework of principles under which any rational approach can be used. Though previous initiatives have been conducted, each focused on only a particular perspective or method, and often failed to provide guidance to the broader continuum of user needs. APOR proposes to incorporate previous guidelines into GAPP while expanding its principles to include research focused on varying perspectives and methodologies. The A.P.O.R. Consensus Development Committee has identified two major perspectives and three broad categories of methodology which underpin the current practice of health economics. In this session, health outcomes and economic issues relating to "broad focused" (societal) perspective and how it relates to the three methodolocigal categories: clinical trials, modeling, and database health economics analysis will be discussed. The "broad focused" perspective is concerned with allocation of scarce resources and social good across society and is best identified with the societal perspective. Its primary question is: should society pay to treat this condition?

Abstract #: AP2


A.P.O.R. Consensus Development Committee

A.P.O.R, through the Consensus Development Committee, is developing methods and guidelines that will become GAAP (Generally Accepted Pharmacoeconomic Principles), a constantly evolving set of standards and principles. The purpose of GAPP is to define a common set of principles, methods, and standards under which all parties can conduct and interpret any type of pharmacoeconomic research. The intent of GAPP is to not prescribe how pharmacoeconomic research should be conducted but to provide a framework of principles under which any rational approach can be used. Though previous initiatives have been conducted, each focused on only a particular perspective or method, and often failed to provide guidance to the broader continuum of user needs. APOR proposes to incorporate previous guidelines into GAPP while expanding its principles to include research focused on varying perspectives and methodologies. The A.P.O.R. Consensus Development Committee has identified two major perspectives and three broad categories of methodology which underpin the current practice of health economics. In this session, the "narrow focused" (provider/payer) perspective and how it relates to the three methodological categories: clinical trials, modeling, and database health economics analysis will be discussed. The "narrow focused" perspective is concerned with maximizing the efficiency of a particular set of treatments from a specific point of view and is best identified with the payer/provider perspective. Its primary question is: given that one chooses to treat a condition, which treatment provides the best value?

Pharmacoeconomic Terminology: Generating a Common Language

Abstract #: CL1


Pashos, CL1, Klein EG2, Wanke L3, 1Abt Associates Inc. and Harvard Medical School, 2Eli Lilly and Co, 3Immunex Corp.

APOR has initiated a project to develop the most comprehensive glossary yet assembled for its field: The APOR Pharmacoeconomics and Outcomes Research Lexicon. APOR's goal in this effort is to develop a resource that will be a useful tool both for its members, the practitioners of pharmacoeconomics and outcomes research, and for those who are users or consumers of this research or whose own work may be affected by it. The Lexicon will be a comprehensive listing of the terms used in pharmacoeconomics and outcomes research. The Lexicon will include examples of terms and bibliographic references to key articles in the published literature, where appropriate. The initial version of The APOR Pharmacoeconomics and Outcomes Research Lexicon is organized into twelve headings as follows: The Field, Outcomes / Endpoints (major categories), Data and Data collection, Analysis Attributes, Strengths and Limitations, Modeling Frameworks, Clinical Outcomes Analysis, Economic Outcomes Analysis, Quality of Life Outcomes Analysis, Satisfaction Outcomes Analysis, Applications of Outcomes in Practice, Healthcare Environment, Pharmaceuticals. Copies of the draft Lexicon can be requested from the APOR office. This interactive workshop will provide an overview of The APOR Pharmacoeconomics and Outcomes Research Lexicon. The organization of the Lexicon and the definition of key terms as well as examples and references to include in the Lexicon will be discussed at this workshop. Additional terms and their definitions, improved definitions of currently included terms, as well as examples of terms or bibliographic citations of landmark publications related to terms are welcomed. Feedback on the Lexicon can be made in person at the APOR Second Annual Meeting or by e-mail to: chris_pashos@abtassoc.com.

Abstract #: CL2


Brier, K.

Light, P.

Moore, M.

Micromedex, Inc.

Englewood, Colorado, USA

An on-going challenge exists in healthcare to integrate disparate work systems and to provide the seamless transfer of clinical information as patient records and databases merge when health systems consolidate. Additionally, these disparate work systems complicate determining appropriate level of care, best practice, physician benchmarking, and health plan performance. A logically modeled lexicon provides the linkage of the medical record with the pharmaceutical intervention, and facilitates the ability to associate encounters with outcomes. Furthermore, a lexicon would facilitate the exchange of clinical data regarding patient symptoms, diagnosis, procedures, demographics, patient identifiers and other specific factors. This allows the data to be passed between clinical applications for use by several systems without the need for reentry of data. This workshop will compare and contrast existing coding systems (e.g., ICD- 9-CM, CPT-4, UMLS, SNOMED, etc.) and discuss the attributes of a logically modeled lexicon that will preserve the original intent of the clinician as data is captured and retrieved from data repositories. For example, key attributes include accommodating synonymous concepts (Prinzmetal and vasospastic angina), overlapping categories (duodenal ulcer can be both gastrointestinal and infectious diseases), and medical intent based on position of words (left to right shunt is not the same as right to left shunt). A data model and initial standardized set of medical concepts have been completed. The data model serves as a long term enterprise solution, accommodating both the continued advances in externally coded input from a variety of sources, the natural evolutionary process of medical language, and applications of natural language processing and their uses in healthcare decision support.

Use of Pharmacoeconomics in Decision-making: Practical Issues

Abstract #: PI1


Leader, S, Mallick, R, Pracon, Reston, Virginia, USA

The advantages and disadvantages of using Medicaid claims for studying economic and clinical outcomes of community based pharmacotherapy will be reviewed as well as techniques for data verification and adjustment for missing clinical data. In the workshop, potential pitfalls of confounding by indication, misclassification of drug regimen, and mismeasurement of drug exposure will be discussed and study design techniques for reducing these sources of bias will be presented. In addition, validation techniques that enhance the reliability of coded diagnoses and outcomes such as data linkage to death records and to prescription claims will be offered. The lessons learned from the presenters' own novel use of Medicaid data to reassess the risks of calcium channel blocker therapy for hypertension will be described. The utility of this approach for examining other clinical controversies will be explored. The workshop is aimed at hands on researchers and sponsors interested in learning more about potential applications of insurance claims data for outcomes research.

Abstract #: PI2



Fairman, K.
Express Scripts, Inc., Maryland Heights, Missouri

Prescription claims data are often used to compare important outcomes across different drug classes and products. These outcomes include compliance (number of days between fill and refill dates), dosing adequacy (average daily dose versus practice guidelines), and treatments' administrative intensity (events requiring physician intervention, e.g., dose titration and product switches). This workshop will demonstrate that drug outcomes analyses can produce misleading results if "real life" factors affecting prescription filling behaviors are not considered. Using antidepressant drug data from a large claims database, strategies for detection and resolution of problems will be presented. The presentation will include evidence that: 1. Gaps between antidepressant prescription depletion and refill are not always due to non-compliance. Late refills are more likely than on-time refills to be followed by dose decreases, suggesting that some gaps may be attributable to verbal orders to reduce dosage by bisecting pills or taking fewer pills. Thus, comparing compliance across products with different dose titration or "tapering" patterns may be problematic. 2. Calculations of dose per calendar day from fill to refill can produce anomalous results because refill timing is related to arbitrary factors such as the weekday on which the prescription depletes. Using the entire therapeutic course to measure dose adequacy, or calculating dose per day of dispensed supply for each prescription, may produce more interpretable results. 3. Dose titration analyses produce different results, depending on the calculation method used. A comparison of titration patterns for three antidepressant products, using five different methods, will be presented. The workshop will conclude with a discussion of selecting appropriate methods, depending on study purpose, policy decisions to be made, and the target population

Abstract #: PI3


Mitchell, JB, McCall NT, Health Economics Research, Inc., Waltham, MA. US

Much of the methodological work surrounding cost-effectiveness analyses (CEAs) has focused on developing appropriate measures of effectiveness. Effectiveness represents only one-half of the cost-effectiveness ratio, however, and costs have received comparatively little attention. One of the challenges of obtaining medical cost data for CEAs is to find data that are not only reliably measured but that also are a valid representation of the costs actually incurred among all patients with a given condition. The costs associated with treating a selected number of patients enrolled in a RCT will be very different from those incurred by patients generally. Medical claims data, have the advantage of generating cost estimates for large numbers of patients in a wide range of treatment settings nationwide. Compared to primary data collection, they also have the advantage of being relatively inexpensive. This workshop will discuss three types of claims data: (1) Medicare; (2) Medicaid; and (3) private health insurers. Together, these three data sources can be used to construct a composite measure of US health care costs for a given illness or treatment. We will describe how these different types of claims can be used to derive not only point estimates of costs for individual services, but also the costs of entire episodes of care. This enables the researcher not only to construct the costs of alternative treatments, but also the costs of alternative ways of sequencing treatments. For example, is bypass surgery performed following an unsuccessful angioplasty procedure more expensive than when performed as the initial revascularization procedure? The workshop also will discuss the strengths and weaknesses of administrative data for different types of CEAs, e.g., drug therapy vs. surgery, preventive services, etc.

Abstract #: PI4


Basskin, L, Butler University, Indianapolis, IN, USA

While hospitals want to offer services which maximize benefits within a given cost (labor) constraint, selection of clinical services is often based on anecdotal reports of patients, physicians, or pharmacists at other institutions. Existing pharmacoeconomic models are of limited value in this process because of the difficulty in both valuing benefits in dollars and measuring effectiveness in a single unit. In this workshop, participants will learn to select clinical services which maximize benefits from a given perspective at the lowest cost using a new pharmacoeconomic model developed by the author. The model allows one to quantify, in a logical, systematic manner, the benefits of a variety of clinical services. Participants will learn how to: (1) identify potential services, (2) identify benefits (defined as objectives necessary to meet the pharmacy department's goals), (3) quantify the extent to which each service satisfies each objective, (4) determine the time/cost required to service each patient, service the entire hospital, and compare it to the total time/cost available, (5) calculate total benefits to the hospital by "weighting" each objective (based on the relative importance of the objective to management, and the probability of that objective being satisfied), (6) determine the cost/benefit ratio for each service, and (7) select the services to be offered based on each service's "cost per weighted objective met". Total benefits, costs to the pharmacy and number of patients who can be serviced are determined. The information obtained from using this model can also be used for pharmacist scheduling and patient care assignments. This workshop will benefit people in universities, disease state management programs or healthcare institutions who need to evaluate, justify or recommend clinical services on the basis of cost-effectiveness.

Abstract #: PI5


Heithoff, K.; Crown B.; Cuffel, B.; Holzer, S.; Frank, L.; Mohr, P.
The MEDSTAT Group, Washington, DC, USA. Lehman, A.; University of Maryland, Baltimore, Maryland, USA

The workshop describes the development of methods and measures suitable for conducting multi-site research on the costs and outcomes of schizophrenia treatment. This workshop has four key objectives: 1) to provide an overview of cost-effectiveness research findings regarding the treatment of schizophrenia, 2) to present issues involved in selection and measurement of outcomes for the severely mentally ill. Methods for measuring outcomes, treatment types, treatment costs, organizational characteristics of care settings, and technology diffusion will be presented 3) to present advantages and disadvantages of randomized clinical trial designs relative to observational designs in cost-effectiveness research for this population, and 4) to present a brief overview of the potential applications of the methods presented for disease management, bench-marking patient and provider outcomes, and for providing essential information on cost- effectiveness for pharmaceutical manufacturers, health care payers, care providers, and people with schizophrenia. Participants will obtain specific information about:

the current status of the cost-effectiveness literature regarding the treatment of schizophrenia;
the state of the art of outcomes measurement for severe mental illness;
practical and statistical considerations in research design selection;
the value of comprehensive cost- effectiveness research for understanding and improving mental health care.

Pharmacoeconomic Methods: Practical Issues

Abstract #: PM01


Ward, RE, Petitta, A, Beis, SJ, Henry Ford Health System, Detroit, MI. US

Dedicated decision analysis software programs are commonly used to create health economic models. However the constraints of these pre-programmed packages can limit the flexibility of the modeling process and sensitivity analysis. The use of general software programs, such as spreadsheets, can overcome these limitations. This workshop will demonstrate the creation of two cost- effectiveness models with a commonly used electronic spreadsheet (Excel 5.0, Microsoft Inc.) This session is intended for individuals that already have a basic understanding of the basic concepts and terminology used in cost-effectiveness analysis who wish to learn more about an alternative modeling technique. A standard decision tree type model and a markov type model will be created. The subject areas that will be covered include model design, modular spreadsheet set-up, defining assumptions, discounting, and presentation of outcomes in a balance sheet type format. Special emphasis will be placed on model design techniques that enhance the ability to discount and conduct sensitivity analysis.

Abstract #: PM02


Caro, JJ, Klittich JS, Caro Research, Boston, MA, USA

To objectively and explicitly compare pharmaceutical interventions, a single index of value is useful. While this is straightforward for costs, the same can not be said for the non-monetary outcomes, Much effort has been put towards the pursuit of a simple outcome measure that can be used across all interventions. As the length of life figures so prominently in out culture, the choice of scale has seemed obvious: life years gained (LYG). The use of this measure, however ignores most of the adversity clinical practice seeks to prevent or alleviate. Presumable, we try to ignore most of the adversity clinical practice seeks to prevent or alleviate. Presumable, we try to treat all illnesses, even ones that do not prove fatal. Credible comparison across interventions requires a measure that values all benefits of treatment, not only the ones that impact on duration of life. Although the intent of quality adjusted life years (QALY) is to move towards this goal, they too have a major defect. Both extremes used to anchor this scale, 1 (perfect quality of life) and 0 (death), are unmeasurable. The definition of perfect quality of life, for an individual or for society, has defied even the most adept philosophers. Second, measuring quality of life in reference to death is nonsensical, as death does not speak to quality, or the lack there of, but rather directly to the presence or absence of life. Although, cost per LYG may be an attractive measure for some interventions, it is far from the all encompassing outcome measure that it is currently being sought. Researchers need to continue in the pursuit of a unifying outcome measure or rethink the purpose of establishing such. While the quest for comparability is laudable, achieving a general measure is not so easy.


Abstract #: PM03


Josephine Mauskopf, Research Triangle Institute, RTP, NC USA,
Kit Simpson, University of North Carolina at Chapel Hill, NC, USA.

Prescription claims data are often used to compare important outcomes across different drug classes and products. These outcomes include compliance (number of days between fill and refill dates), dosing adequacy (average daily dose versus practice guidelines), and treatments' administrative intensity (events requiring physician intervention, e.g., dose titration and product switches). This workshop will demonstrate that drug outcomes analyses can produce misleading results if "real life" factors affecting prescription filling behaviors are not considered. Using antidepressant drug data from a large claims database, strategies for detection and resolution of problems will be presented. The presentation will include evidence that: 1. Gaps between antidepressant prescription depletion and refill are not always due to non-compliance. Late refills are more likely than on-time refills to be followed by dose decreases, suggesting that some gaps may be attributable to verbal orders to reduce dosage by bisecting pills or taking fewer pills. Thus, comparing compliance across products with different dose titration or "tapering" patterns may be problematic. 2. Calculations of dose per calendar day from fill to refill can produce anomalous results because refill timing is related to arbitrary factors such as the weekday on which the prescription depletes. Using the entire therapeutic course to measure dose adequacy, or calculating dose per day of dispensed supply for each prescription, may produce more interpretable results. 3. Dose titration analyses produce different results, depending on the calculation method used. A comparison of titration patterns for three antidepressant products, using five different methods, will be presented. The workshop will conclude with a discussion of selecting appropriate methods, depending on study purpose, policy decisions to be made, and the target population


Abstract #: PM04


Harji I. Patel, Berlex Laboratories, Inc., Wayne, NJ. US

In this workshop we review existing procedures and propose a new procedure for comparing medical costs of interventions in a longitudinal trial in the presence of censoring. Here censoring accommodates both deaths and premature withdrawals. We briefly review design and analysis of a longitudinal study for a conventional endpoint and types of missing data along with consequences of ignoring them from the analysis. In pharmacoeconomic studies, the total cost during a trial is of interest and is different from a conventional endpoint in that it can be measured cumulatively over time. So in the presence of censoring the analysis of total cost needs some modifications. Suppose the entire follow-up period is partitioned into a number of small intervals for which the cost data are available. Lin et al. (1997) recently proposed a mean based procedure to analyze such data after making adjustment for the intervention specific survival distributions. Their method relies on the central limit theorem which may be unsatisfactory when the populations sampled are highly skewed and the sample sizes are small. Furthermore, the sample mean is sensitive to outliers which often exist in medical cost data. We therefore replace the intervention mean difference for each interval by the Hodges-Lehmann estimator after making adjustment for the survival distributions. Also, prior to computing this estimator we make some refinement in the cost for an interval in which censoring occurred. We then estimate the total cost difference between the interventions by adding the Hodges-Lehmann interval estimators. The properties of the proposed estimator are examined through a Monte Carlo study. Since the distribution of this estimator is not known, a bootstrap confidence interval for the difference between the intervention costs is computed. A bootstrap confidence interval method is reviewed.


Abstract #: PM05


Rajagopalan, R. Glaxo Wellcome Inc.,
Research Triangle Park, NC, US

To evaluate the structure and functional capabilities of various pharmacoeconomic models that help analyze the issues hand build an analysis model to realize a meaningful solution.

Pharmacoeconomic issues are examined from different perspectives: 1) patient, 2) provider, 3) managed care organization, and 4) society. Mostly, policymakers consider the societal perspective. Then, one of the following pharmacoeconomic models is chosen to assess the economic value of the health care product at hand: 1) cost-benefit analysis, 2) cost-effectiveness analysis, and iii) cost-utility analysis. The advantages and disadvantages of each of the model will be discussed so that an appropriate model will be chosen for the evaluation at hand.

Further, for any one of the pharmacoeconomic model, Markov chain model, econometric (multiple regression) model, extrapolation, data transformation, stratification, scoring and grouping model etc. Again, pros and cons of each analytical model will be discussed. Audience participation will be encouraged. Analysis models will help analyze the issues allowing for limited assumptions to make up for absence of necessary data. Thus, the models help the analyst complete the analysis in spite of missing data so that available information and time are made use of in an optimal way.

An example of logistic regression model will be discussed.

Abstract #: PM06


Petitta, A, Ward RE, Beis SJ,
Henry Ford Health System, Detroit, MI. US

This session is a continuation of the Developing Cost-Effectiveness Models Using a Computer Spreadsheet Workshop. In this session sensitivity analysis will be used to assess the uncertainty in the previously developed models. Three types of sensitivity analysis will be conducted: one-way sensitivity analysis, break-even analysis, and multi-way sensitivity analysis. Subject areas to be presented include spreadsheet set-up of the sensitivity analysis, interpretation of the results, and presentation of sensitivity analysis results. Emphasis will be given to the multi-way simulation since this method better estimates overall model uncertainty than traditional one to three variable sensitivity analysis. The multi-way sensitivity analysis will be performed by Monte-Carlo simulation using spreadsheet add-in forecasting software (Crystal Ball, Decisioneering Inc.) In the simulation, every variable in the model will simultaneously be allowed to vary randomly and independently over a pre- defined pattern for 10,000 iterations. The simulation will result in a range of 10,000 forecasts for each of the model outcomes. The results will then be used to attach 90% certainty ranges to each of the health and economic outcomes of the model.

Abstract #: PM07


Wood, LL, Bala, MV, Mauskopf, JA,
Research Triangle Institute, Research Triangle Park, NC, US

Methods have been developed in environmental economics to estimate willingness to pay (WTP) for goods not traded in private markets. The development of such methods has led to a growing interest in using WTP to value health benefits. However only a small number of studies have been done in the health area and the design of such studies for valuing health benefits and the interpretation of results are not well understood by health care researchers. In this workshop, we will describe the use of WTP as a method for valuing health benefits. We will present the theoretical foundations of WTP, different ways in which WTP questions can be framed and when to use each approach, and interpretations and use of WTP results. WTP can be used to measure health benefits for a specific patient group or health benefits at the societal level. We will discuss the use of WTP as a measure of benefits, present the format of questions used to estimate WTP, and define how different approaches may be appropriate depending on the benefit being measured. Finally, we will compare cost-benefit analysis with cost-effectiveness analysis for conducting economic evaluations of health interventions and show how cost-benefit analysis (for which WTP provides the benefits measure) offers several advantages over the more commonly used cost-effectiveness analysis for measuring the desirability of new treatments or drugs. This workshop will be of value to industry pharmacoeconomists. It will show them how to use WTP as a method of valuing benefits, how to frame the appropriate questions, and how to interpret results.

Abstract #: PM08


Polsky, D, Glick, H., University of Pennsylvania, Philadelphia, PA

Recently researchers evaluating cost effectiveness ratios in clinical trials have been reporting
confidence intervals along with their estimates of cost effectiveness. While the use of confidence intervals has been expanding, standard methods for their calculation have not been well articulated in the literature. In this workshop we will provide detailed instruction as to how to use bootstrap methods to compute confidence intervals. Particular attention will be paid to cases where the costs and effects do not result in an interpretable ratio and to cases where the resulting confidence intervals may be difficult to interpret. Topics will include: (1) A review of the procedure for computing confidence intervals for cost effectiveness ratios using the nonparametric bootstrap; and (2) a review of how these intervals should be reported. The workshop will provide a forum to discuss two special cases regarding reporting of confidence intervals. The first is when the study does not produce a cost effectiveness ratio, per se, but produces evidence that the treatment dominates the standard or the standard dominates the treatment. In this case we propose using the bootstrap method to compute the confidence of being dominated or dominant and then reporting this confidence level. In addition, when the limits of the confidence interval are not an interpretable ratio, we propose using the bootstrap to compute the probability of being below a predetermined threshold value. The goal of this workshop is to introduce methods of computing and reporting confidence intervals to improve the quality and reliability of measures of statistical uncertainty surrounding cost effectiveness ratios. Both analysts who calculate cost effectiveness ratios in randomized trials and decision makers who use cost effectiveness ratios will benefit from this workshop. 

Abstract #: PM09


Strauss, BM, Smith, ME, Schwartz, L, Linde-Zwirble, W, Newbold, RC, Lidicker, J
Bayer Corporation, West Haven, CT, Health Process Management, Doylestown, PA US

Prospective data (PD) studies often have limitations due to the selection/exclusion criteria. Sensitivity analysis and modeling techniques are often educated guesses and potentially important factors are not identified prior to study initiation. OBJECTIVE: The objective of the study was to overcome limitations of PD studies by using population- level, retrospective data (RD) sources, and appropriate research methods. METHOD: The key to productive RD research implementation is to identify the treatment groups within the database only after all other exploration has been accomplished. Sensitivity analysis on the study exclusion criteria can be performed directly because the entire data set is already gathered. Furthermore, retrospective databases contain the entire set of factors and procedures actually recorded or performed. RESULTS: Prospectively designed retrospective data studies are key to the appropriate use of a given data resource, enumerating the primary and secondary results. The most important limitation of these studies, is that subjects are not randomized to treatment groups. However, real-world use of therapeutic interventions never takes place in a randomized population. Results from RD studies may therefore be more representative of potential real world results. CONCLUSION: Together, RD and PD studies have a natural synergy. The RD study can illuminate unforseen factors, selection bias, data and factor distributions and modeling techniques in a shorter time and at lower cost. The results of RD studies can then be used to design better PD studies: much of the guess work currently undertaken to estimate the size of effects, sample sizes and analysis methods could be avoided. PD studies could then be performed with a higher likelihood of producing more timely, cost-effective, and useful information.

Abstract #: PM10

Effectively Reporting Pharmacoeconomic Data: clarity and content

Rindress D, Goetghebeur M, Welner S
BioMedCom Consultants inc, Montreal, Quebec, Canada

The content, clarity and readability of a document has a critical impact on the delivery of its message. In pharmacoeconomic research, as in any developing field, there have emerged over the last few years many opinions about standards necessary to assure quality and credibility. A plethora of guidelines, regulations and position papers have been issued by government agencies, academic committees and biomedical journals concerning the realization and reporting of pharmacoeconomic studies. A synthesis of those relating to the reporting of data will be presented along with practical tools for communication of pharmacoeconomic data to the biomedical and health care community. This interactive workshop will provide an overview of the accumulated body of knowledge of reporting guidelines and their application to two modes of reporting, i.e., publication in biomedical and health care journals and pharmacoeconomic study reports. Included in the materials will be copies of relevant overheads, practical templates for the production of manuscripts and reports, and useful checklists for the review of documents for content, clarity and consistency. Anyone who needs to effectively communicate pharmacoeconomic information may benefit from the material compiled for this workshop; a rudimentary knowledge of pharmacoeconomic research methodology is recommended. Participants should walk away with a basic organizational plan for the preparation of pharmacoeconomic research reports and manuscripts.

Past Meetings Index