DEPARTMENT OF HEALTH AND HUMAN SERVICES
National Institutes of Health
Office of the Director

DISEASE-SPECIFIC ESTIMATES OF
DIRECT AND INDIRECT COSTS OF ILLNESS
AND NIH SUPPORT

FISCAL YEAR 2000 UPDATE

Ruth Kirschstein, M.D.
Acting Director, NIH

February 2000

• Preface
  
• Executive Summary
  
• Introduction
  
• Background
  
  • History of Cost of Illness Estimation
    
  • Conceptual Methods and Data Issues
    
  • Causality and Attribution: Assigning Costs to Specific Diseases
    
  • COI Compared with Cost Benefit and Cost Effectiveness Analyses
    
• Methodology
  
• Presentation of the Data
  
• Discussion
  
• Conclusion
  
• References
  
• Tables (All tables are in pdf format - You must have Adobe Acrobat Reader installed on your computer to view documents in this format. It is available free for download here.):
  
  1. Costs of Illness and NIH Support for Selected Diseases and Conditions (pdf)
     
  2. NIH Support for Selected Diseases and Conditions (FY 1992 -FY 2001) (pdf)
     
  3. DHHS Support for Selected Diseases and Conditions (FY 1992-FY 2000) (pdf)
• Appendix: Information for Diseases and Conditions
  

PREFACE

This FY 2000 revision of the report "Disease-Specific Estimates of Direct and Indirect Costs of Illness and NIH Support" was prepared in response to a Congressional request that the National Institutes of Health (NIH) combine and update two previously requested reports entitled "Disease-Specific Estimates of Direct and Indirect Costs of Illness and NIH Support" and "HHS and National Costs for 13 Diseases and Conditions" (House Report 106-370).

The NIH submitted the first version of the report in 1995 and updated the numbers in 1997 and in early 1998. For the current revision, all of the NIH support figures have been updated to reflect the FY 1999 funding levels. Over the years, several, but not all, of the cost estimates have been revised. For this update, cost estimates were revised for alcohol abuse, allergic rhinitis, Alzheimer's disease and other dementia, asthma, atherosclerosis, dental, peptic ulcer, heart diseases, coronary heart disease, HIV/AIDS, lead poisoning, end stage renal diseases, mental disorders, neonatal respiratory distress syndrome, acute respiratory distress syndrome, smoking, and urinary incontinence.

The National Institute of Mental Health (NIMH) submitted a new estimate for the total costs of mental disorders, based on improved methodology and more recent data, which reflects the effects on treatment costs of the introduction of managed care. However, newer estimates for the costs of six mental disorder subcategories that were included in previous versions of this report are not available. Since the older subcategory estimates are now outdated and inconsistent with the new estimate for total costs of all mental disorders, the older estimates are not included in this updated report.

The report now contains 60 cost estimates: for 14 of the top 15 killers (no cost estimate is available for conditions originating in the perinatal period) and for 46 other diseases and conditions identified by the various Institutes within the NIH.

As with previous versions of this report, an appendix sheet for each disease includes more detailed documentation of the cost estimate and information on additional indicators of burden. Appendix pages have been updated for several of the conditions.

The original 1994 Congressional request asked NIH to include cost estimates and number of deaths for the 15 leading causes of death as identified by the Centers for Disease Control and Prevention (CDC). The list of leading causes in the original report was based on deaths for 1991. The current version reports deaths for 1998. The rank order by cause of death changes from year to year. In 1998, HIV/AIDS dropped off the list of the top 15 killers for the first time in several years. Alzheimer's disease was not among the top 15 in 1991, but was the twelfth leading killer in 1998.

The most substantial change for this edition is the addition of two tables. Table 2 (pdf) provides a history of NIH funding support for each disease area from FY 1992 through FY 2001. Support for research is distinguished from support for prevention/public education activities for each disease area in Table 2 (pdf). Table 3 (pdf) provides a similar history of Department of Health and Human Services funding for a subset of the Selected Diseases and Conditions included in Table 2 (pdf). Table 3 (pdf) was initially provided by NIH for only 13 diseases in 1998 as part of a Department of Health and Human Services report to Congress, "HHS and National Costs for 13 Diseases and Conditions.” That report was requested in the House Appropriations Committee Report on the FY 1998 Labor/HHS Appropriations Bill (HR 2264, House Report 105-205. p. 130). Costs estimates are not available for three of the diseases included in Tables 2 and 3: depression, kidney disease, and liver disease.

The narrative report has not changed substantively. The core message remains the same. Disease-specific cost of Illness estimates are uneven and essentially non-comparable for the reasons stated in the original report. Economic costs are an incomplete index of disease-specific burden and an insufficient criterion for priority setting.

EXECUTIVE SUMMARY

The original report was submitted by NIH in September 1995, in response to a request from the Senate Committee on Appropriations (Senate Report 103-318). The Senate Committee directed the NIH to develop a table showing estimates of the societal impact of the diseases on which NIH Institutes and Centers (ICs) conduct and support research, by disease and by IC, for the most recent years that estimates are available, and to adjust the estimates for possible double counting. NIH was also asked to include on the same chart, its annual spending on the diseases in question.

Subsequent discussion with Senate staff revealed that the Committee was particularly interested in cost estimates and NIH support for the following categories of disease: (1) the top 15 causes of mortality as identified by the Centers for Disease Control and Prevention; and (2) other diseases for which ICs have presented societal cost estimates in testimony before Congress or in official reports, publications, or speeches. Some ICs chose to submit available cost estimates for diseases identified as an important component of their research portfolio even though the estimates had not been previously quoted. NIH was directed not to generate new cost estimates in developing a response to this request.

The applicability of cost of illness (COI) estimates to policy and budgetary decisions related to life sciences research is limited for several reasons. There is considerable variability in the methods and data used to generate COI estimates. Estimates of the economic costs of illness do not capture some important aspects of the burden of illness such as reduced functioning, pain and suffering, and deterioration in other dimensions of health-related quality of life, including emotional and psychological impacts on families, friends, and co-workers. It is difficult to identify discrete research funding consistently with the disease categories that correspond to specific COI estimates since much of the research supported and conducted by NIH is basic in nature, and not disease-specific. There are also significant overlaps among different areas of research because of the interdependent nature of scientific knowledge. Finally, even if consistent and comprehensive estimates of the relative burdens of specific diseases were available, decisions regarding policy and budget would have to include other factors such as the importance of scientific advances and opportunities.

The report includes a background section to provide the reader with the context/framework within which to review the estimates presented in Table 1 (pdf). That section should be thoroughly read before attempting to make any inferences from the estimates. In compiling the requested Table, NIH found considerable variability in the methods and data used to generate COI estimates, resulting in incomparable estimates and in overlaps of costs allocated to different diseases. Since available documentation does not include sufficient detail to permit ex-post adjustments of cost estimates, the report focuses on providing information on how the estimates were developed. Likewise, the figures for NIH disease-specific support also include overlaps so that they correspond more closely to the cost estimates and to previous figures on program support provided by the NIH. The Appendix contains the documentation for each of the COI estimates in Table 1 (pdf) and provides other relevant information about each of the specific diseases/conditions.

Given the variation in approaches, COI estimates need appropriate documentation to establish their validity and to avoid misinterpretation and inappropriate comparisons among different diseases or conditions. COI estimates do not provide a simple formula for the allocation of research resources. They cannot substitute for the well-informed judgment required to synthesize information about the broader dimensions of disease burden with knowledge of scientific opportunities in developing strategies and budgets for research and development programs. However, COI estimates can provide order of magnitude indicators of the economic burden of particular diseases. While they should be interpreted with caution, COI estimates can help decision-makers in Congress and in the Administration anticipate and respond to public interests.

INTRODUCTION

As requested, this report is an update of the report originally submitted by the National Institutes of Health in response to the following directive issued by the Senate Appropriations Committee in its report on the FY 1995 budget:

Costs of disease study. The Committee has heard from NIH and others about the effect that particular diseases have on medical and other social costs. There are so many claims, however, that their individual relevance is hard to discern. The Committee, therefore, directs NIH to develop a table showing all of the estimates of the societal impact of the diseases on which NIH ICDs [Institutes, Centers, and Divisions] conduct research, by disease and by ICD, for the most recent years estimates are available. Medical costs should be shown separately from the cost of lost productivity or other indirect effects. The study should show the total of the estimates, accompanied by any discount NIH believes may need to be applied to reflect possible double-counting. NIH should indicate on the same chart fiscal year 1994 spending on the diseases in question (Senate Report 103-318, p.116).

Subsequent discussions with Senate Committee staff narrowed the scope of the request to the following categories of diseases: (1) the top 15 causes of mortality as defined by the Centers for Disease Control and Prevention (CDC); and (2) other diseases for which ICs have presented societal cost estimates in testimony before Congress or in official reports, publications, or speeches. Some ICs chose to submit available cost estimates for diseases identified as an important component of their research portfolio even though the estimates had not been previously quoted. NIH was specifically directed not to generate new cost estimates in responding to this request.

Cost of illness (COI) estimates provide order of magnitude indicators of the economic burdens imposed on society by various diseases and conditions. However, the applicability of COI estimates to policy and budgetary decisions related to life sciences research is limited for several reasons:

Variability in methods and data. There is considerable variability in the methods and data used to generate COI estimates. As a result, cost estimates for different diseases, or even for the same disease, may not be comparable. For example, one analysis may include only treatment costs, while another may include estimates of the value of lost production due to morbidity and mortality. Many patients have more than one disease/condition simultaneously, such as heart disease and diabetes, or have other underlying risk factors such as smoking and alcohol abuse. In such cases, costs may be allocated on the basis of primary diagnosis only, or some share may be attributed to comorbidities or underlying, contributory conditions. Furthermore, there is an inherent imprecision in estimating accurately items such as the costs of physician visits, hospital stays, and the value of lost productivity due to death and disabilities.

Burden of illness exceeds economic costs. Estimates of the economic costs of illness do not capture some important aspects of the burden of illness such as reduced functioning, pain and suffering, and deterioration in other dimensions of health-related quality of life including emotional and psychological impacts on families, friends, and co-workers.

Research categories do not coincide with disease categories. It is difficult to identify research funding consistently with the disease categories that correspond to specific COI estimates. Much of the research supported and conducted by NIH is basic in nature, and not disease-specific. In some cases, an important target of research is understanding normal developmental, physiological, or population-related processes to better understand, identify, prevent, and treat a range of problems. There are also significant overlaps among different areas of research because of the interdependent nature of scientific knowledge.

Importance of scientific advances and opportunities. Even if consistent and comprehensive estimates of the relative burdens of specific diseases were available, decisions regarding policy and budget would have to include other factors such as the importance of scientific advances and opportunities as well as the research tools available to address specific disease processes. A recent research advance often creates greater scientific opportunities for research and development in one disease area than in others. Conversely, a lack of knowledge regarding underlying pathophysiological processes can inhibit the development and evaluation of diagnostic and therapeutic technologies for other conditions, regardless of their social burden. Thus varying research support figures often reflect the level of scientific advances and opportunities in different areas, or the availability of research tools to pursue certain fields of research, as well as a judgment about the relative importance of different diseases.

The report is organized in six substantive sections. The Introduction provides the Congressional charge and briefly identifies several qualifiers related to using COI data. The Background section provides a brief history of COI estimation, a summary of conceptual issues regarding methods of analysis and data collection, a discussion of the multifactorial nature of diseases and its effect on the assignment of costs to specific diseases, and a comparison of COI with cost benefit and cost effectiveness analyses. The section includes a discussion of the evolution, rigor, and standards associated with COI estimation, and it highlights the variation in approaches and methods used to develop COI estimates to promote caution in the interpretation, comparison, and use of COI estimates to guide research policy decisions. The Background section also provides the reader with the context/framework within which to review the particular findings.

The Methodology section describes how the estimates and associated documentation were compiled. The Presentation of the Data section includes a description of the cost estimates and figures for NIH support for each disease/condition in Table 1 (pdf).

The last sections of the report, Discussion and Conclusion, explore the implications of the report for future studies, especially with regard to the expense and standardization of cost estimates and the need to pursue additional indicators for the burden of disease. Further issues associated with research resource allocation are also identified in these sections, including a discussion of the limitations of COI studies and comparisons with other cost estimation approaches.

Table 1 (Costs of Illness and NIH Support for Selected Diseases and Conditions) and the Appendix contain detailed documentation for each of the COI estimates and provides other relevant information about each of the specific diseases/conditions.

BACKGROUND

History of Cost of Illness Estimation

Cost of illness studies were pioneered in the late 1950s and early 1960s and have proliferated over the past 30 years.¹ Concern for the lack of consistency and comparability among studies led to the establishment in 1978 of a U.S. Public Health Service (PHS) task force on COI studies. The task force was charged with recommending methodological guidelines to promote conformity and enhance the comparability of the results of future studies. The recommended methodological guidelines were designed to be sufficiently flexible so that the objectives of individual COI studies could be achieved, while promoting conformity in ways that would enhance the comparability of studies. (Hodgson and Meiners, 1982, p. 430).

Many COI studies make use of national surveys of health status, utilization of health care, and health care expenditures funded and administered by Federal agencies² and address only one or a small subset of diseases at a time. Some employ a “top-down” approach in which all costs are allocated to one and only one category, based on primary diagnosis as defined in the International Classification of Diseases, 9th Revision, Clinical Modifications (ICD-9-CM), so there is no overlap or double counting.³ One disadvantage of this method is that it is applied only to broad diagnostic categories. The data sets used do not always support reliable estimates of the parameters required to estimate COI for the more narrowly-defined diseases or conditions, such as arthritis or Parkinson’s disease. Nor does the top-down approach fully capture the implications of a disease or condition, such as diabetes, which may contribute to increased costs as a secondary diagnosis, or as a risk factor for other diseases and conditions, such as end-stage renal disease, heart disease, and eye disorders.

For the above reasons, numerous studies conducted over recent years have focused on the economic costs of more narrowly defined diseases and conditions. Some of those studies attempt to account more comprehensively for the implications of a particular condition as a comorbidity or underlying cause of other conditions (Fox et al., 1993; Rice et al., 1990). Even when those studies conform with the broad guidelines established by the PHS task force (Hodgson and Meiners, 1982; Hodgson, 1983), the different methods and data employed restrict the comparability of estimates from different studies.

Conceptual Methods and Data Issues

The PHS guidelines addressed and clarified a number of conceptual and empirical issues inherent in estimating COI. These issues are briefly reviewed in this section.

1. Cost components

A major innovation of the early studies and the PHS guidelines was the specification and classification of costs included in COI estimates and the distinction between costs and transfer payments. COI studies measure the "economic" burden resulting from disease and illness across a defined population, including both direct and indirect costs. Direct costs are the value of resources used in the treatment, care, and rehabilitation of persons with the condition under study and are, therefore, unavailable to produce other goods and services. Indirect costs represent the value of economic resources lost because of disease-related work disability or premature mortality. It is important to distinguish costs -- net reductions in the value of economic resources available for other purposes -- from monetary transfer payments such as disability and welfare payments. Such payments represent a transfer of purchasing power to the recipients from others (the general taxpayers), but do not represent net increases in the use of resources.

COI estimation has progressed to the point where many studies distinguish between the core (health-related) and other related (non-health) costs of illness and disability. The components of direct and indirect costs are more precisely defined as follows:

• Core direct costs are those connected with the use of medical care in the prevention, diagnosis, and treatment of disease and in the continuing care, rehabilitation, or terminal care of patients. Examples include expenditures for hospitalization, outpatient clinical care, nursing home care, and home health care; services of primary physicians and specialists, dentists, and other health practitioners; drugs and drug sundries; and rehabilitation counseling and other rehabilitation costs, such as for prostheses, appliances, eyeglasses, hearing aids, and other devices to overcome impairments resulting from illness or disease. Collectively, these expenditures represent the personal health care component of the United States National Health Accounts.⁴
• Other related direct costs are borne by patients or other payers but are not included in the National Health Expenditures Accounts. Examples of such costs are expenditures for transportation to hospitals, to physicians' offices or to other health providers; certain household expenditures (e.g., help for cleaning, laundering, and cooking); special diets and clothing; relocation and moving expenses; and certain property losses, including the destruction of property resulting from alcohol abuse and criminal activity due to drug addiction.
• Core indirect costs measure the value of time that patients lose from employment or other productive activity due to mortality or morbidity. These costs also include reduced productivity once the patient returns to work, including unwanted job changes and loss of opportunities for promotion or education.
• Other related indirect costs include the value ascribed to time lost from work, housekeeping, etc., by family members or friends who transport, visit, and care for patients. Also included are indirect costs such as the value of productivity losses due to victimization by crime, incarceration, and criminal activities due to conditions such as alcoholism and drug abuse.

Because many diverse kinds of costs are attributable to disease, it is a complex task to estimate the full extent of the costs that result from a particular disorder. Where data are unavailable, some cost components may be left out, or proxies may be used for missing information, introducing differences in the resulting estimates.⁵

2. Method: Human Capital Versus Willingness to Pay Approach

Either the human capital approach or the willingness to pay approach may be used to value the lost years of life due to premature death and to value the lost activity days due to morbidity. The human capital approach is the most commonly used of the two methods and was the basis for all of the estimates reported in Table 1 (pdf). It attempts to value an individual’s contribution to national production and measures the indirect costs of illness in terms of the market valuation of lost wage earnings due to morbidity and mortality. Its basic economic principles can be traced back at least as early as the seventeenth century (Petty, 1699). Mushkin (1962) is credited with bringing the human capital method to the health field, but several others have contributed to its development and application.⁶ Most human capital studies follow the lead of Rice (1985) and use national data on population, life expectancy, labor participation rates and earnings to develop annual and lifetime earnings profiles by sex and age categories. More detailed analyses which distinguish variations in earnings related to race and educational attainment are feasible, but are seldom included in COI estimates. In more recent studies, the value of lost household services has also been included and is imputed on the basis of expected earnings of service workers such as maids and cooks combined with information on the average time spent on various household tasks by family members depending on their age, sex, and employment status.⁷

The "willingness-to-pay" (WTP) methodology attempts to determine the subjective valuation that individuals would place on being free of the disease under study. WTP is more aligned with conventional concepts in welfare economics, but it is difficult and expensive to implement, and there is no consensus on reliable and standardized survey instruments for doing so (Hu and Sandifer, 1981; Dickens, 1990; Viscusi, 1990). Because WTP estimates are generally higher than human capital estimates, the latter may be regarded as a lower bound estimate of indirect costs.

3. Prevalence-Based Versus Incidence-Based Costs

The COI studies reported in Table 1 (pdf) approach cost estimation from either of two perspectives. Most of them use the prevalence-based (or annual cost) approach that measures the costs which accrue during a base year due to all existing (or prevalent) cases of disease in that year. In estimating the economic burden resulting from the prevalence of disease, the present discounted value of future losses due to mortality is calculated. The conventional methodology attributes the future losses to the year in which the death occurred (Rice et al., 1990).

The incidence-based (or lifetime cost) approach measures the present value of the lifetime costs of the disease for all new (incident) cases with onset of disease during the given base year (Hodgson, 1983). Incidence-based costs require knowledge of the likely course of a disease and its duration, survival rates, onset and patterns of medical care, and the impact of disease on employment (Rice et al., 1990), so they are generally more difficult to estimate than prevalence-based estimates. However, the incidence-based approach is sometimes more useful for comparing the effects of alternative interventions to prevent, treat, or manage a particular disease. While most COI studies use the annual approach, there are some notable exceptions.⁸

4. Variability of Cost Estimates

The published literature on COI studies documents substantial variation in the methods and data used to estimate the overall costs of illness. The choices of what cost components to include, whether to use the human capital approach or WTP approach, and whether to use the prevalence or incidence approach each influence the value of the final estimate. For example, some studies generate estimates only for core direct costs and neglect indirect costs altogether. Other studies have shown that the indirect costs of lost productivity can amount to anywhere from 25 to 300 percent of the core direct costs for some diagnostic categories.⁹

Besides the variations in approaches to cost estimation, discussed in the sections above, additional factors that contribute to variations in COI estimates include the following:

• How narrowly or broadly the disease is defined, and how costs are attributed to secondary diagnoses and to underlying risk factors which cause, or complicate other conditions and disease processes;
• What methods and assumptions are used to adjust for incomplete data;
• What base year is used for data on prevalence, utilization, and costs, and what interest or discount rate is used to adjust future-year costs to a base-year value;
• Whether the estimate refers to the total resident U.S. population or a subset of the population defined by gender, age, ethnicity, or geography; and
• Whether costs are estimated from the perspective of the total cost to society or a narrower perspective, such as the cost to the Federal budget or to a private insurance company.¹⁰

The approach and method selected to develop a particular estimate are influenced by the characteristics of the disease process and its sequelae on costs, and the availability of data suitable for estimating the relevant cost elements. Locating or collecting data, investigating all components of costs, and documenting the effects of complex disease processes on costs all contribute to greater expense in generating estimates. Depending on time or financial constraints and how the estimate is to be used, a less comprehensive, less accurate estimation methodology may be chosen. Most COI studies employ “surrogate” measures, such as average charges per hospital day or physician visit, that are readily available through existing databases, in place of conceptually correct but unavailable information on unit costs.

The inherent imprecision in estimating unobservable quantities, such as the value of lost productivity due to death, also contributes to the variability in COI estimates.

Causality and Attribution: Assigning Costs to Specific Diseases

Many health problems develop as a result of a variety of genetic, environmental, behavioral, unknown, or even random factors. As a result, many patients present multiple conditions. The implications of the existence of comorbidities are an important, but often overlooked, aspect of the cost estimation process in the allocation of costs to specific diseases. A further complicating factor is determining whether the comorbidities are independent phenomena, or are somehow interdependent. For example, a patient admitted to a hospital for treatment of a heart condition may also suffer from Alzheimer's disease and cancer. The comorbidities may not be risk factors for heart disorders, but their presence may contribute to a higher than average length of stay and related treatment costs for the heart condition. Some portion of the extra cost could legitimately be attributed to Alzheimer's disease and cancer rather than solely to heart disease.

The attribution of cost differs if one or more of the comorbidities are identified as a risk factor, or probable cause of the primary diagnosis. It is known that smokers have an above average risk of developing heart disease and lung cancer. Persons who suffer from diabetes have a higher risk of developing heart disease, end stage renal disease, eye disorders and deterioration of other organs. Thus, some share of the total costs (not just the extra costs) of treating heart diseases for heavy-smoking diabetics could be attributed to diabetes, smoking, or both. However, the appropriate proportion is difficult to determine since a large percentage of smokers would have developed heart disease even if they had never smoked, and similarly for diabetes.

Consequently, a major consideration in establishing COI estimates involves the decision to adjust for overlaps or to allocate some costs to more than one disease or condition. For example, cirrhosis of the liver is one category of digestive disease, and any measure of the burden of digestive diseases that omitted liver cirrhosis would be seriously deficient. However, a large proportion of liver cirrhosis cases are the direct result of the long-term heavy consumption of alcohol that usually accompanies alcohol dependence. It would be equally misleading to estimate the burden of alcohol abuse and dependence without considering the proportion of liver cirrhosis cases that are attributable to alcohol.

The following Exhibits illustrate the implications of recognizing cause-and-effect relationships for classifying the leading causes of death. According to the taxonomy maintained by the CDC, the leading causes of death include the conditions listed in Exhibit 1 below. This framework emphasizes the primary pathophysiological conditions (listed causes) identified at the time of death. An alternative approach, identified by McGinnis and Foege (1993), emphasizes root causes -- a combination of inborn (largely genetic) and external factors. Exhibit 2 includes McGinnis and Foege's estimate of only external (non-genetic) root causes of death.

McGinnis and Foege (1993) illustrated the important principle that the cause of death can be viewed from two perspectives: the listed cause or the root cause. However, distinguishing and attributing costs to various listed and root causes is more difficult than their analysis might imply. McGinnis and Foege (1993) identify only nine root external (non-genetic) conditions, which account for about 50 percent of all deaths. Cost studies are often desired for more disaggregated disease categories (e.g., specific cancers, end-stage renal disease, or Alzheimer's disease) rather than just broad categories (e.g., diseases of the heart or malignant neoplasms). Further disaggregation, however, contributes to additional uncertainty and imprecision regarding attribution of costs to listed or root causes. As mentioned above, the relationship between comorbidities is not easily identifiable, conceptually or empirically.

However, the two-way classification implicit in the McGinnis approach still does not capture all causal links. The pathway from cause to effect may be quite indirect and involve one or more intermediate steps. For example, having a drug abuse disorder places an individual at increased risk of contracting AIDS, which in turn, often leads to treatment costs that can be ascribed to various cancers or infections. Similarly, one or a combination of such factors as unfavorable genetic constitution, poor maternal health, inadequate prenatal care, and adverse maternal behaviors (substance abuse and an unhealthy diet) may contribute to poor birth outcomes. At a minimum, conditions originating in the antenatal or perinatal period may cause slightly longer initial hospital stays or more frequent follow-up visits. In more serious instances, these factors may result in offspring with increased susceptibility to a variety of illnesses and diseases or lifelong physical, cognitive, and mental disabilities. While researchers are exploring both the pathophysiological and root causes of conditions classified as "originating in the perinatal period," it remains difficult to fully describe the risk factors leading to poor birth outcomes and to subsequently link either the factors or the outcomes to associated additional lifetime health care costs.

COI Compared with Cost Benefit and Cost Effectiveness Analyses

While COI studies have some influence in establishing "orders of magnitude," and were used in this report to respond to the Congressional directive, a number of economists and analysts have questioned the usefulness of the COI methodology when compared with cost benefit analysis (CBA) and cost effectiveness analysis (CEA), especially as a guide to allocation of resources for health care services.

CBA and CEA are methodologies used for evaluating the health outcomes and costs of health interventions (Drummond et al., 1987). Each methodology attempts to compare the incremental benefits with the incremental costs of a particular intervention relative to another intervention or the status quo. They are intended to help the decision maker make the best use of scarce resources and show the opportunity costs, or tradeoffs, involved in alternative decisions. They differ in the accounting for benefits. CBA attempts to value the consequences or benefits of each program in money terms so that costs and benefits are directly comparable. CEA measures the consequences of the program in the most appropriate natural or physical units such as "years of life gained" or "cases correctly diagnosed"; no attempt is made at valuing the outcome.

As currently advocated, CBA and CEA require some of the same types of economic data collection and analysis as COI, although usually at a more demanding level of specificity. In contrast to the CBA/CEA focus on comparing incremental benefits with incremental costs, COI measures the level of total economic costs during a specific period of time or for the lifetime of a group of new, incident cases. COI has been criticized (Shiell et al., 1987) for mixing measures of benefits and costs since it includes both an indicator of benefits (indirect costs) as well as the direct costs of interventions designed to prevent or treat disease.

While CEA is equivalent in intent to CBA (and incremental COI analysis), its use of natural or physical units as measures of program output means that it does not require that indirect costs be evaluated in monetary terms, thus dispensing the human capital/willingness-to-pay debate. The use of CEA, and its variant -- cost utility analysis (CUA), which incorporates quality of life considerations -- is an evaluation methodology which is currently under consideration by the U.S. Department of Health and Human Services for the evaluation of government health care programs.¹¹

Applying the principles of CBA and CEA to the allocation of research resources is necessarily speculative since analyses of costs and benefits must be performed before they are realized and before they can be measured with any degree of confidence. Experts are reduced to making best guesses, based on previous analogous experience regarding the probability and timing of successfully developed innovations, the expected incremental reduction in COI (or effect on alternative outcome measures) of the new intervention, and the additional research and development costs required to develop and implement the new intervention. In 1986, the Office of Technology Assessment released a report which concluded that the use of quantitative models such as CBA and CEA "while valid conceptually, does not provide a useful practical guide to improving Federal research decision making." It found that the factors to be taken into account are too complex and subjective, and that the payoffs are too diverse and incommensurable "to allow quantitative models to take the place of mature informed judgment."¹²

METHODOLOGY

The cost estimates in this report were provided by planning and evaluation staff in the individual ICs of NIH. No new estimates were prepared for this report. As reported in documentation included in the Appendix, a few of the estimates had been generated by IC staff, some were developed under contracts sponsored by an IC, and most were identified through a literature search. Analysts from the Office of Science Policy (OSP), in the Office of the Director (OD), developed a detailed "Assessment of Methods and Data" to elicit summary documentation for these cost estimates. The Office of Budget (OB), OD, compiled the data provided by ICs on NIH support for each disease.

PRESENTATION OF THE DATA

Table 1 (Costs of Illness and NIH Support for Selected Diseases and Conditions) (pdf) presents COI estimates and NIH spending for the 14 of the 15 leading causes of death in 1998, and for 46 other diseases and conditions for which one or more ICs have previously quoted a cost estimate (in testimony, speeches, or reports), or for which cost estimate was available and an IC considered the condition an important component of its research portfolio.

The ranking and annual deaths for the top 15 causes of mortality in 1998¹³ are listed in columns 2 and 3 respectively. In addition, estimated deaths for Alcohol Abuse and Dependence, for Drug Abuse, and for Smoking (for tobacco use) are listed and indicated with an asterisk (*) in column 3. These three conditions were listed among the top 9 major external (non-genetic) factors that contribute to death in the United States (McGinnis and Foege, 1993).

The documentation provided by the ICs confirmed suspected inconsistences in methods and data used, resulting in incomparable estimates and in overlapping costs included in estimates for different diseases. Unfortunately, available documentation does not include enough detail to permit ex-post adjustment of the cost estimates for such overlap and related inconsistencies. This report is limited to providing available information on how the estimates were developed and a number of cautions regarding interpretation and use of the estimates.¹⁴

No cost estimates were located by the NIH ICs for two of the top 15 causes of mortality -- Conditions Originating in the Perinatal Period; and, Nephritis, Nephrotic Syndromes and Nephrosis. A cost estimate for Kidney and Urologic Diseases (which include nephritis, nephrotic syndromes, and nephrosis) was substituted for the more specific Nephritis, Nephrotic Syndromes and Nephrosis category.¹⁵ The time and effort required to generate a new estimate for perinatal conditions was beyond the scope of the Senate’s request in this report (see Appendix).

Although only a single point estimate is provided for each disease/condition listed in Table 1 (pdf), a range of uncertainty should be ascribed to each estimate. The cost estimates are based on a series of parameters, each of which is, at best, estimated from survey data with implicit sampling error. Numerous judgments regarding selection and interpretation of proxies for missing or incomplete data influence the derivation of the final value and add further uncertainty.

Furthermore, any attempt to compare cost data across disease categories must consider in depth, the many conceptual and methodological issues that may lead to variations in cost estimates. More details, including related measures of disease burden and references to the underlying cost reports for each disease, are included in the documentation summaries provided in the Appendix.

Major considerations for interpreting and comparing the cost estimates in the Table (pdf) with each other and with estimates from other sources include the following:

• Reference Year (col. 4). The estimates are each expressed in dollars for a particular reference year. The reference years range from 1985 to 1999. To express all estimates in a common reference year, it would be necessary to adjust for changes in the disease burden over time, patterns of treatment and care, and the purchasing power of the dollar for health care services. No attempt was made to adjust the estimates for differences in reference year, because currently available price indices for health care expenditures cannot adjust for all these factors; the necessary information on subcomponents of the cost estimates required to make such adjustments is missing for several diseases, and even a successful adjustment for differences in reference year fails to compensate for other important differences in data and methods.
• 
  
  
• Cost components (cols. 5-13). The comprehensiveness of the estimates of direct and indirect costs differs across diseases because of the difficulty and cost required to estimate the non-health related costs, and the indirect costs related to reduced productivity after returning to the job and the value of services of unpaid care providers. Specialized studies providing information on reduced productivity of patients after returning to work, on services of unpaid care providers, and on other related costs such as training, reeducation, extra household expenses, property loss, costs of the judicial system, and administration of social welfare programs may be located or commissioned to provide more comprehensive and detailed cost data for a specific disease or related sets of diseases, but they are not generally available for all diseases.

  All but two of the 60 estimates (suicide and sudden-infant-death syndrome) include core direct costs (col. 6). 13 of the studies attempt to include other related non-health direct costs (col. 8). 41 of the studies provide estimates of indirect costs. Those estimates usually include the estimated value of lost workdays due to premature mortality (col. 9), due to morbidity (col. 10) or both. 14 studies include estimates of reduced productivity while on the job (col. 11); 3 estimate the value of services of unpaid care providers (col. 12); and 6 include estimates of other related indirect costs (col. 13)

• 
  
  
• Discount Rate (col. 14). The stream of lost earnings over future years is converted to base or reference year values using a discount rate intended to reflect the tradeoff between the values of a dollar received today versus one received next year. Choice of the appropriate discount rate remains controversial and may vary considerably between studies. Valid comparisons of estimates require adjustment of discount rates to a common value. The discount rates used in the studies reported in the Table fall within a fairly narrow range, 3 percent to 6 percent. Even so, this range substantially influences estimates of the indirect costs due to mortality. For example, the present value (in 1990 dollars) of the average remaining lifetime earnings for a 30- to 34-year-old male is estimated at $898,779 with a 3-percent rate but only $688,100 with a 6-percent rate.
• 
  
  
• Definition of disease (cols. 15-17). Because the interrelationships among disease categories or causal agents are complex (e.g., diabetes is a risk factor for cardiovascular disease; tobacco addiction is a risk factor for cancer) and patients often present more than one disease or condition, it is not always feasible or appropriate to construct mutually exclusive disease categories and associated cost estimates. Cost estimates depend on how narrowly or broadly the disease is defined, whether it includes related conditions beyond its narrowly defined or primary ICD-9-CM code (col. 15) (e.g., digestive diseases include colorectal cancer and, therefore overlap with the estimate for cancer); whether the estimate includes identifiable extra costs attributable when the disease is listed as a secondary diagnosis or comorbidity (col. 16); and whether the estimate includes costs attributable to the disease or condition as an underlying cause or risk factors of other diseases (col. 17).

  For example, the estimates for Diseases of the Heart, Malignant Neoplasms, Chronic Obstructive Pulmonary Disease, and Pneumonia and Influenza are based on primary diagnoses and include only core costs (direct and indirect). By contrast, the estimates for Alcohol Abuse, Diabetes, and Drug Abuse reflect the characteristics of these disorders by including the costs of the various health consequences for which these disorders are the underlying cause, as well as various other related costs that result from these diseases. This results in the double counting of some costs when the estimates of these conditions are compared with cost estimates for their manifestations such as end stage renal disease, some eye disorders, heart conditions, and digestive diseases, and injuries.

• 
  
  
• Method of Estimation. All of the COI estimates in this Table are based on the conventional human capital method rather than the less common willingness-to-pay approach (see Background section). Because the human capital approach values the indirect costs of illness in terms of market earnings, it yields very low values for the retired elderly and for children. It also ignores certain dimensions of illness and death, such as pain and suffering.
• 
  
  
• Approach of Estimation. Two approaches can be used to estimate COI: prevalence-based (annual) cost and incidence-based (lifetime) cost (see Background section). All of the COI estimates in this Table use the prevalence-based approach, except accidents and adverse effects; homicide and legal intervention; several injury subcategories; and an eye subcategory - diabetic retinopathy.
• 
  
  
• Scope and Perspective of Estimation (col. 18). Most estimates address the total U.S. resident population; they are not specific to particular geographic areas or ethnic groups. However, a few are limited by the age of the patients. Similarly, most of the studies estimate costs to the total society, regardless of who bears the costs (not indicated in the Table). The only exceptions in this Table are end-stage renal disease, which is limited to Federal payments for Medicare claims, and infertility, which includes only payments by insurance companies.
• 
  
  
• Institute Identifying the Cost Data (cols. 19-20). The estimates for each disease or condition in the Table were provided by the IC identified in column 19. As reported in documentation included in the Appendix, a few of the estimates had been generated by IC staff, some were developed under contracts sponsored by an IC, and most were identified through a literature search. Column 20 indicates whether the estimate has been quoted by the NIH in testimony, speeches, or reports.
• 
  
  
• NIH Support (col. 21). This column contains the NIH spending on each disease compiled by OB, OD, from data provided by ICs. The support figures contain overlap. For example research that addresses cancer of the pancreas may be reported under both Cancer and Digestive Diseases. This corresponds to the standard reports to Congress provided in previous years. Since overlap in cost estimates could not be removed, no attempt was made to eliminate overlaps in the corresponding NIH spending figures.

DISCUSSION

The costliness and variability of methods used to generate COI estimates have several implications. First, COI estimates are simply not available for most conditions; in addition, older published estimates may be obsolete due to changes in diagnostic capabilities, treatment, prevalence, and the course of a disease, as well as changes in the costs of health care. Second, cost studies generated with different methods are not comparable, and in most cases, they cannot be adjusted sufficiently to eliminate disparities. Third, COI estimates reflect some degree of inherent imprecision that results from the estimation of fundamentally unobservable quantities, such as the value of output that was not produced as a result of the disease in question. Finally, COI estimates inevitably provide incomplete measures of the burden of disease because they cannot thoroughly explore and evaluate the effects of the disease/condition on all potential cost components. The emotional and psychological impacts of illness and disability often extend to family, friends, and others who render care. Because of the difficulties inherent in quantifying these attendant burdens of disease, the costs associated with these burdens may not be linked to specific diagnostic categories or systematically included in COI estimates.

There is no simple, inexpensive solution to achieving fully consistent, fully comparable COI estimates. The art of cost estimation could be enhanced, and the variability of estimation procedures reduced, by collecting more detailed and comprehensive data on the costs and utilization of health care classified by diagnosis. Efforts to improve data should explore options to assess the impact of underlying causes and comorbidities on the costs of illness. However, comprehensive data collection is expensive. A set of consistent estimates could be developed, but significant effort would be required to standardize methods and data for each estimate.

In addition, strict standardization is unlikely to serve all purposes and may generate its own set of controversies. Insistence on standardization might limit the scope of the cost components included in the studies, prohibit estimates for important but narrowly defined diseases (because of small sample size in national surveys), and preclude the use of data sets that would improve estimates for one class of diseases but do not provide figures for other diseases.

Finally, the standards or guidelines used in preparing estimates will influence the allocation of costs across different diseases. For example, a study which allocates utilization and associated costs based on primary diagnosis and proximate cause provides a much different picture than a study which attempts to consider the contributory effects of comorbidities or one that attributes costs to underlying risk factors. If the resulting cost estimates are perceived to affect public policy decisions regarding the allocation of health care or research resources, advocacy groups can be expected to note the implications of proposed guidelines and will support or challenge the estimates’ validity pursuant to their specific interests.

COI estimates address only the economic burden of illness. Assessments of disease, health service programs and research programs, would benefit from efforts to improve availability of indicators of other dimensions of disease burden, e.g., statistics on incidence, prevalence, death rates, and morbidity that are routinely available for some, but not all diseases. Efforts should also include the development and wider application of indicators of the impact of disease and treatment on functioning and other dimensions of health-related quality of life. The Public Health Service, including the NIH, has sponsored research in those areas for many years.¹⁶ Such measures are especially important as more of the health services budgets are absorbed by the treatment of chronic conditions. For such disorders, important research program outcomes are methods of prevention and treatments that improve functioning and quality of life rather than merely extending life.

Related to these efforts to measure health-related quality of life are attempts to gauge or rank the intensity of preference for different health states according to indices variously referred to as quality-adjusted life years, quality of well-being scale, or years of healthy life. The unifying objective of these measures is to combine dimensions of quality of life, length of life, and patients’ preferences into a single measure.¹⁷ Such efforts supplement COI estimates by gauging the burden of disease, the cost effectiveness of new therapies and prevention programs, and the contribution of biomedical research to the health of the nation’s residents.

CONCLUSION

The inherent variability in methods of estimation documented in this report, make comparisons of COI for different diseases problematic. In addition, economic cost estimates measure only a portion of the full burden of disease and illness to society. They neglect other equally important but difficult to measure dimensions--such as deterioration in function and heath-related quality of life, including the emotional and psychological impacts on families, friends, co-workers, and others who render care.

NIH priorities are influenced by a confluence of important factors, which include medical and public health needs, the state of scientific knowledge in a given field, and advances which make feasible new investigations. The NIH relies heavily on non-directed, investigator-initiated research to generate the most creative ideas from the scientific community. Because much of the research supported and conducted by NIH is basic in nature, tangible results of scientific inquiry cannot always be defined or planned in advance. In some cases, research is undertaken to learn more about the normal developmental, physiological, or even population-related processes. As a result, it is difficult to identify discrete research funding consistently with the disease categories that correspond to specific measures of burden.

In addition, due to the interdependent nature of scientific knowledge, there are significant overlaps among different areas of research, even when they are targeted toward apparently unrelated areas. For example, progress in such fields as rational drug design and bioengineered vaccines have resulted from fundamental support in fields of genetics, immunology, cell and molecular biology, biophysics, and biochemistry. These burgeoning technologies may supplant present interventions that address symptoms rather than cure or prevent disease, and may have a profound impact on the economic COI in the United States.

For the reasons outlined above, COI estimates do not provide a simple formula for the allocation of research resources. They cannot substitute for the well-informed judgment required to synthesize information about the broader dimensions of disease burden with knowledge of scientific opportunities in developing strategies and budgets for research and development programs.

However, COI estimates can provide order of magnitude indicators of the economic burden of particular diseases. As such, they are part of the information policy makers find useful to inform their decision making. While they should be interpreted with caution, COI can help Congress and the Administration to anticipate and to respond to public interests. Given the variation in approaches documented in this report, users of COI estimates would be well advised to insist on detailed documentation to establish the validity and underlying assumptions of COI studies. This would avoid misinterpretation of data and misleading comparisons of diseases.

REFERENCES

American Diabetes Association, 1993. Direct and Indirect Costs of Diabetes in the United States in 1992 (Alexandria, VA. ISBN 0_945448_32_5)

American Public Health Association, 1992. "Advances in Health Status Assessment: Proceedings of a Conference," Medical Care, Vol. 30, No. 5.

American Public Health Association, 1994. "Conducting Medical Effectiveness Research: A Report From the Inter-PORT Work Groups," Medical Care, Vol. 32, No. 7.

Becker, G. 1967. Human Capital and the Personal Distribution of Income (Ann Arbor, MI: University of Michigan).

Brody, W.H. 1975. "Economic Value of a Housewife," Research and Statistics. Note, No. 9 (Washington, DC: U.S. Social Security Administration).

Centers for Disease Control and Prevention/National Center for Health Statistics. 1996. Monthly Vital Statistics Report Vol. 45, No. 3, Supplement, September 30, 1996 (Hyattsville, MD: U. S. Department of Health and Human Services).

Cooper, B.S., and Rice, D.P. 1976. "The Economic Cost of Illness Revisited," Social Security Bulletin, February: 21-36.

Dickens, W.T. 1990. "Assuming the Can Opener: Hedonic Wage Estimates and the Value of Life", Journal of Forensic Economics, 3:51-60.

Drummonds, M.F., Stoddart, G.L., and Torrance, G.W. 1987. Methods for the Economic Evaluation of Health Care Programmes (Oxford University Press: New York).

Fein, R. 1958. Economics of Mental Health (New York: Basic Books).

Fox Ray, N., Will, S., and Tamer, M. 1993. Direct and Indirect Cost of Diabetes in the United States in 1992 (Alexandria, Va.: American Diabetes Association, Inc.).

Gold, Marthe R., Siegel, Joanna E., Russell, Louise, Weinstein, Milton C. 1996. Cost Effectiveness in Health and Medicine, (Oxford University Press: New York).

Hartunian, N.S., Smart, C.N., and Thompson, M.S. 1981. The Incidence and Economic Costs of Major Health Impairments (Lexington, MA: D.C. Heath and Company).

Hodgson, T.A. and Meiners, M.R. 1982. "Cost of Illness Methodology: a Guide to Current Practices and Procedures," Milbank Memorial Fund Quarterly, 60:429-462.

Hodgson, T.A. 1983. "The State of the Art of Cost-of-Illness Estimates, " Advances in Health Economics and Health Services Research, 4:129-164.

Hodgson, T.A. 1988. "Annual costs of illness versus lifetime costs of illness and implications of structural change," Drug Information Journal, 22:323-341.

Hu, T. and Sandifer, F.H. 1981. Synthesis of Cost of Illness Methodology. National Center for Health Services Research Contract No. 233-79-3010 (Washington D.C.: Public Services Laboratory, Georgetown University).

Jarvinen D., Rice, D.P., Kelman, S. 1988. Cost of Illness Studies: An Annotated Bibliography 1988. or Administrative Use Only. U.S. Department of Health and Human Services, Public Health Service, Alcohol, Drug Abuse, and Mental Health Administration.

Kamlet, M.S. 1992. "The Comparative Benefits Modeling Project: A Framework for Cost-Utility Analysis of Government Health Care Programs," Office of Disease Prevention and Health Promotion (Washington, D.C.: U.S. Department of Health and Human Services and The Foundation for Health Services Research).

Letsch, S.W. et al. 1992. "National Health Expenditures 1991," Health Care Financing Review, 14(2):1-29.

Lipscomb, J. 1986. Human Capital, Willingness-to-pay, and Costs-effectiveness Analysis of Screening for Birth Defects in North Carolina. Presented at the Annual Meeting of the American Economics Association, New Orleans, December 28, 1986.

McGinnis, J.M. and Foege, W. H. 1983. "Actual Causes of Death in the United States," JAMA, Nov. 10, Vol 270, No. 18.

Miller, T. et al. 1994. "Medical Care Spending--United States," Morbidity and Mortality Weekly Report, Vol. 43/No.32, August 19, pp. 581-585.

Mushkin, S.J. 1962. "Health as an Investment." Journal of Political Economy, 70(5):129-57.

Mushkin S.J. and Collings, E.A. 1959. "Economic Costs of Disease and Injury," Public Health Reports, 74:795-809.

National Institutes of Health. 1990. Quality of Life Assessment: Practice, Problems, and Promise. Proceedings of a Workshop, Office of the Director, NIH.

Office of Technology Assessment. 1986. Research Funding as an Investment: Can We Measure the Returns?: A Technical Memorandum. (Washington, D.C.: Congress of the United States, Office of Technology Assessment).

Office of Technology Assessment. 1994. Identifying Health Technologies That Work: Searching for Evidence (Washington, D.C.: Congress of the United States, Office of Technology Assessment).

Patrick, D. and Erickson, P. 1993. Health Status and Health Policy: Allocating Resources to Health Care (New York: Oxford University Press).

Peskin, J. 1984. "The Value of Household Work in the 1980s," American Statistical Association, 1983 Proceedings of the Social Statistics Section (Washington, D.C.: American Statistical Association).

Petty, W. 1699. Political Arithmetik, or a Discourse Concerning the Extent and Value of Lands, People, Buildings, Etc. (London: Robert Caluel).

Rice, D.P. 1966. Estimating the Cost of Illness: Health Economics Series No. 6. PHS Pub. No. 947-6 (Washington, D.C.: U.S. Government Printing Office).

Rice D. and Cooper, B. 1967. "The Economic Value of Human Life," American Journal of Public Health, 57(11): 1954-66.

Rice D.P., Hodgson T.A., and Kopstein, A.N. 1985. "The Economic Costs of Illness: A Replication and Update," Health Care Financing Review, 7:61-80.

Rice, D.P., Kelman, S., Miller, L.S., Dunmeyer, S. 1990. The Economic Costs of Alcohol and Drug Abuse and Mental Illness (Rockville, MD: National Institute on Drug Abuse).

Robinson J.C. 1986. "Philosophical Origins of the Economic Valuation of Life," Milbank Quarterly, 64:133-155.

Rubin, R.J., Altman W.M., and Mendelson, D.N.1994. "Health Care Expenditures for People with Diabetes Mellitus, 1992," Journal of Clinical Endocrinology and Metabolism, 78: 809A-809F.

Shiell, A., Gerard, K., and Donaldson, C. 1987, "Cost of Illness Studies: An Aid to Decision-making?" Health Policy, 8:317-323.

Viscusi, W.K. 1990. "The Econometric Basis for Estimates of the Value of Life," Journal of Forensic Economics, 3:61-70.

Waitzman, N.J., Romano, P.S., Scheffler, R.M. 1994. "Estimates of the Economic Costs of Birth Defects," Inquiry, 33:188-205.

Walker, K.E. and Gauger, W.H. 1980. "The Dollar Value of Household Work," Information Bulletin, 69 (Ithaca, NY:New York State College of Human Ecology, Cornell University).

Weisbrod, B.A. 1961. Economics of Public Health (Philadelphia, PA: University of Pennsylvania).

FOOTNOTES