In 2008, the Medicare program provided health insurance coverage for over 45 million people—37.8 million ages 65 and older and 7.4 million disabled people, with total Medicare expenditures of $468 billion (Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds, 2009).1 The 2009 Medicare Trustees report projected that, under its intermediate assumptions, expenditures will increase from 3.2 percent of gross domestic product (GDP) in 2008 to 11.4 percent by 2083, raising serious solvency issues for the program by 2017. Policy makers face significant challenges managing the program, given not only increases in the size of the beneficiary pool with the increasing size of the aging population, but also increases in average rates of spending per beneficiary resulting from factors specific to health status and health care, such as increasing rates of obesity and the development of new drugs, medical care technology, and medical treatments.
Likewise, policy analysts confront a difficult task in developing credible short-term as well as long-term projections of Medicare costs in the face of uncertainty with regard to the many factors that are likely to affect future costs. There is uncertainty not only in the underlying economic and demographic assumptions used in projection models, but also in what a policy
The Medicare program has four major components or parts: Part A, which helps cover hospital costs; Part B, which helps cover costs of physicians and outpatient services; Part C, which covers enrollment in Medicare Advantage Plans (health maintenance organization type private plans); and Part D, which helps cover prescription drug costs (see http://www.medicare.gov/navigation/medicare-basics/medicare-benefits/medicare-benefits-overview.aspx).
modeler assumes about future changes in the health status of the population, the extent and pace of scientific and technological breakthroughs in medical care, the preferences of the population for particular kinds of care, the likelihood that policy makers will alter current law and regulations, and how each of these factors relates to health care costs. There is need for better understanding of the factors contributing to the growth of health care spending and how these factors might be moderated in the future. There is also need to consider for policy models the trade-offs between simplicity and comprehensiveness and what is required for short-term and longer term projections.
Given the substantial growth in the Medicare population, fueled by the aging of the baby boom generation and rising life expectancy at age 65, and the continued increases in Medicare, Medicaid, and private health insurance spending, the availability of well-specified models and analyses that can provide useful information on the likely cost implications of health care policy alternatives is critical for public- and private-sector policy planning. Current models for health care cost projections range from a simple projected rate of GDP increase plus a specified percentage, to dynamic microsimulation models that “age” population cohorts over time, to computable general equilibrium models of the health care sector of the economy and long-term health care spending (see examples in Box 1-1).
It is therefore timely to review the capabilities and limitations of extant health care cost models and to identify areas for research that offer the most promise to improve modeling, not only of current U.S. health care programs, but also of policy alternatives that may be considered in the coming years. Understanding the factors that affect health care costs for the elderly and how to develop improved projection models for policy is an important area for the behavioral and social research program of the National Institute on Aging (NIA), given its concern with the health and socioeconomic well-being of the elderly, which could be significantly affected by changes in Medicare and other health care programs. NIA consequently asked the National Research Council’s Committee on National Statistics (CNSTAT) to conduct a public workshop on needed research to improve health care cost projections for the Medicare population and on the strengths and weaknesses of competing frameworks for projecting health care expenditures for the elderly.
The workshop was to consider major classes of projection and simulation models that are currently in use and the underlying data sources and research inputs for these models. It was also to consider areas in which
Three Models for Projecting Medicare Costs
There are currently several models and data sources for projecting future health care expenditures that vary in their capabilities, complexity, and limitations, three of which are briefly summarized below.
additional research and data are needed to inform model development and health care policy analysis more broadly, such as:
The relative merits of various cost projection approaches with regard to short-term versus long-term projections, the ability to model what-if scenarios, and other features for the major modeling categories.
Trends in socioeconomic status and in mortality and morbidity and how they affect health care cost projections.
Medical technology as a driver of costs and the policy responses to this trend.
Factors affecting health status, such as obesity, disability, and chronic diseases, that may affect costs over the longer term.
Addressing uncertainty and bias in model projections.
In response to NIA’s request, CNSTAT appointed a steering committee to plan a public workshop to identify research that can improve models for projecting health care costs for the population 65 years and older and, more broadly, address factors that drive health care spending. The workshop was structured to combine invited presentations and discussions among the participants to consider the uses and limitations of alternative modeling approaches, as well as factors that affect health care spending and suggest priorities for research that could support improved projection models, including a long-term research agenda in this area for NIA and others in the field. To set the context and provide background information for the workshop participants, the steering committee commissioned a paper on currently used models for forecasting health care costs for the Medicare population, including their strengths and limitations, their methodological approaches to forecasting, and their applications (see Appendix A). The workshop, held on January 13, 2010, drew people from a wide variety of disciplines and perspectives, from federal agencies, academia, and nongovernmental organizations. The workshop agenda and a list of presenters appear in Appendix B, and biographical sketches of steering committee members appear in Appendix C. The slides used in support of the presentations are available at http://www7.nationalacademies.org/cnstat/workshop%20Cover%20Page.pdf.
PLAN OF THE REPORT
This report is a summary of the presentations and the discussions flowing from the presentations during the sessions outlined in the agenda (see
Appendix B). Following this introduction, Chapter 2 opens with an overview of the technical background paper, then discusses the relative merits and limitations of several current models. Chapter 3 focuses on modeling medical technology as a driver of Medicare health care spending, and Chapter 4 addresses such factors as obesity, socioeconomic status, chronic diseases, and disability that affect health status as drivers of Medicare health care spending. Chapter 5 focuses on the future in terms of research areas that may advance the current efforts from the perspective of the participants attending the workshop.
It is important to be specific about the nature of this report prepared by the workshop rapporteur, which is a factual summary of what transpired at the workshop. It is therefore limited to the views and opinions of those participating in the workshop. It reflects the concerns and areas of expertise of the workshop participants and is confined to the material presented by the workshop participants. The presentations and discussions were limited by the time available for the workshop. Neither the workshop nor this summary was intended as a comprehensive review of research relative to Medicare cost projections, nor was it designed to generate consensus conclusions or recommendations. The workshop focused instead on the identification of issues in understanding Medicare cost projections and themes and considerations for future improvements in data and models. Workshops such as this, even though they are not designed to produce consensus recommendations and conclusions, can be very helpful in documenting what is happening in the field and providing a sense of where the field needs to move forward.