Click for next page ( 108


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 107
Costs and Sources of Funding The focus of this chapter win be on the costs associated with studies that prospectively gather primary data on diagnostic technology. The aim of these studies may be to establish safety, efficacy, effectiveness, or cost- effectiveness, and several research designs may be used, including ran- domized controBed trials, nonrandomized comparative studies, or retro- spective studies. Our goals are to understand the factors that contribute to the costs of these studies and to examine briefly the question of who should pay for ~em. In He process, we will take a closer look at several studies of diagnostic technology. FACTORS THAT AFFECT STUDY COSTS Remarkably little information is available on costs of studies of diag- noshc technology. Clinical teals have received more attention, in part because of their seemingly large price tags. The Coronary Drug Project, He first major clinical trial sponsored by He National Institutes of Health, cost approximately $50 million (Levy and Sondik 1982). Several general issues warrant discussion. From the perspective of a funding agency, the important costs to consider are the incremental costs of doing a study. These represent expenses that are~beyond the costs of usual patient care. For example, in a study of a drug that is already used in a group of patients, the cost of me drug itself is not a cost of the study; the Snug win be prescribed whether or not the study is done. A similar situation occurs with respect to diagnostic tests. Often a study win be 107

OCR for page 107
108 ASSESSMENT OF DIAGNOSTIC TECHNOLOGY perfonned on a test that is already in clinical use. The incremental costs of the study win usually involve only the additional tests necessary for the study. There are, of course, exceptions: some projects may require funding for an tests or for activities that could be considered part of the usual patient care. In addition, He incremental costs should include any patient-related expenses brought about by adverse effects of the addi- tional testing. As a practical matter, however, these costs win often be impossible to predict. Another difficulty in demeaning the cost of a particular project arises when more than one question is studied simultaneously. A study of CI may simultaneously examine the effectiveness of different contrast agents. It may be difficult to allocate the costs of the study of contrast agents within the context of Be entire research project. In large trials ~ which many questions are addressed, this problem may be intractable. Many of the costs we discuss are typically included in the budget of a research proposal; some are not budgeted, however, and represent "hid- den" costs. For instance, the time and effort involved in protocol develop- ment may not be reimbursed, or, in some large projects, the budget may include investigator and consultant time for protocol development (see "Examples from Actual Studies" later in this chapters. Nonetheless, aD of the costs should be explicitly anticipated, whether or not they will appear in a submitted budget. The specific costs ova study depend on the technology under consid- eration and on Be coccal problem for which me technology is being used. Thus, rather than attempt to make detailed estimates, we win raise issues that warrant consideration by investigators and by the policymakers who must Bloc ate resources to support studies of diagnostic technology. We consider costs, including unreimbursed costs to patients, relevant to the funding of these studies. We have arbitrarily divided study costs into three categories: costs associated with planning and protocol develop- meet, costs associated with implementation, and costs associated with data interpretation. Meinert (1986) and Piantadosi (1987) discuss the cost of clinical trials; many of the issues that they consider are applicable to the assessment of diagnostic technology, and we draw on Weir work. Planning and Protocol Development Personnel Investigator time is a substantial cost during aU stages of Be study. Often it represents a hidden cost, because it may not appear in grant

OCR for page 107
COSTS AND SOURCES OF FUNDING 109 budgets; in many cases, Me institution underwrites all or part of the investigator's salary. Statisticians, research assistants, data entry personnel, data analysts, and administrative and secretarial staff are necessary for most projects. The costs associated with these personnel accrue at different times during the study but should be anticipated during the planning phase (McNeil 1979). Protocol Development An early step in most studies will be to specify the study protocolks) and to develop the fonns for data collection. It win usually be appropriate to consult with statisticians and data analysts at this stage. The associated costs may be trivial or substantial, depending on the scope of the project. The Veterans Administration Cooperative Studies program estimated the average planning costs of multi-institutional clinical trials to be $20,000 to $26,000, or about ~ to 2 percent of the total cost (Henderson 1980~. The Prospective Investigation of Pulmonary Embolic Diagnosis ff~IO- PED) study of diagnostic tests for pulmonary embolism, a $7.6 minion multi-institutional technology assessment sponsored by the NTH, required over a year to develop the necessary protocols (Vreim 1988~. Sample Size A critical activity during the planning phase is estimating the sample size needed to give the study adequate power to detect clinically mean~ng- fiA differences in test performance. - Sample size win strongly affect the costs of the study: the cost of diagnostic tests and the cost of foDow-up win directly depend on the number of subjects in the study. Administra- tive and personnel costs are likely to increase with larger sample sizes as well. Several factors affect He size of the sample needed for the study. The cost of the study may rise dramatically as the power of the study is increased or when the study is designed to detect smaller differences in outcomes (Detsky 1985~. The power of a study is the probability that the study win successfully detect a difference in outcome if the difference actuaRy exists. A study designed to reveal a 20 percent difference in clinical outcome at a given power may be many times more expensive than a study designed to show a 40 percent difference in clinical outcome at the same power. It may be difficult to show a difference in sensitivity and specificity of a new technology when the test perfonnance of the old

OCR for page 107
110 ASSESSMENT OF DIAGNOSTIC TECHNOLOGY technology is already quite good (Abrams and Hessel 1987). Thus, if an existing technology has a sensitivity and specificity that are nearly 1.0, the maximum possible difference in sensitivity and specificity between the new and old technology win be small; to detect this small difference will require a large sample size. In their study of cT, ultrasound, and gallium scans in the evaluation of patients with an undiagnosed cause of fever, McNeil et al. (1981) calculated that it might require as many as 500 patients to show a significant difference in Me receiver operating charac- tenstic (ROC) curves of the different tests. A possible strategy to reduce study costs without loss of power is to use comparison groups of uneven size (Meydrech 197S, Rosenberg 1983~. This approach can be useful when the costs associated wad one compar'- son group are less than the costs of the other. The methodology has been analyzed for case-control studies; whether it win be useful for technology assessment is not yet certain. Decision-Analytic Modeling We have suggested that technology assessment be model-dnven (see Chapter 3) in order to assure that all the relevant data are specified by the study protocol. This approach involves decision-analytic modeling of the relevant decision problems before actual data gathering begins. Thus, an individual with experience in decision-analy~c methods must be involved in the study from the outset. Our experience at Stanford University suggests that modeling the clinical problem, including literature review, may take three to six months of fur-time work, depending on the scope of the project. This modeling effort would represent a salary expense of approximately $15,000 to $30,000 to support a qualified investigator. Implementation Patient Accrual A number of costs, often unanticipated, relate to He process of recruit- ing and enrolling patients. (For a general model of predicting accrual costs in clinical trials see Piantadosi 1987.) One of the more common mistakes in the design of clinical trials is to underestimate He difficulty of enrolling the required number of patients (Hulley 1988~. A notable example has been discussed by McNeil et al. (19811: in a National Cancer Institute study of Or versus radionuclide studies in patients with sus-

OCR for page 107
COSTS AND SOURCES OF FUNDING 111 peeled intracranial disease, 3,000 patients were screened but only 156 were suitable for final data analysis. A variety of misadventures coT~rib- uted to this poor yield, many unrelated to accrual, but me study serves as an example of how unanticipated difficulties may arise. The study cost $2 million, or $16,441 for each patient analyzed. Several factors influence the difficulty of patient accrual. As patient eligibility criteria become more restrictive, larger numbers of patients will have to be screened. Once the investigators have found art eligible patient, the ease of enrollment will depend on the nature of the study. We can imagine that the enrollment effort in a study where the design requires doing a barium enema, flexible sigmoidoscopy, and colonoscopy in each subject might suffer because of Me distasteful nature of the tests. An additional consideration is attrition of patients. If attrition is likely to be a major problem, the investigators should plan to screen still larger numbers of patients. . . A research assistant may be vital to the recruitment effort. As we noted in Chapter 4, it will be impossible to enroll patients without the coopera- tion of refening physicians. The research assistant may play a role in garnering this cooperation. The help that he or she provides in explaining the protocol, gathering data, and providing follow-up information to referring physicians will facilitate enrollment. McNeil et al. (1981) noted the need for a fi~ll-t~me research assistant at each study site In their evaluations of Or and ultrasound. Data Collection Costs will increase as more data are collected because this effort will require more time fimm study personnel. In addition, procedures to ensure undone data collection and accurate data entry, and to provide quality control, win become more complex and costly as the size of the study increases. These problems may be particularly important for multi- institutiona] studies (see "Examples from Actual Studies," p. ~ 13~. Patient Follow-Up Depending on Me details of the research question, patient follow-up may be necessary. First, the determination of the true disease state of the patient may depend on clinical follow-up, particularly when there is not an acceptable and reliable gold-standard test. Likewise, if the gold- standard test is considered too dangerous to use in patients with a negative

OCR for page 107
112 ASSESSMENT OF DIAGNOSTIC TECHNOLOGY index test, clinical foRow-up win be necessary to determine Me patient's disease status. Second, in cost-effectiveness or cost-benefit studies, the investigator must determine patient outcomes, and this win usually in- volve clinical follow-up. Third, occasionally technology assessments win randomize patients to one technology or another, with coccal out- come being the measure of efficacy. The type of patient foBow-up used wiD clearly influence costs. A questionnaire win be more economical Wan chart review, which in turn win be more economical man patient ~ntenriews. Additional Medical Expenses A study patient may incur costs as an indirect result of the protocol. For example, additional hospital days may be required. Although these costs win usually be difficult to estimate (and Hey are urdikely to be budgeted), they may be substand al. For example; if a protocol added an average of one-half day to the hospital stay, this could amount to $300 for each patient, or up to $60,000 for a study with 200 patients. In some cases, these costs will be paid by third-party payers. With a DRG-based reimbursement plan, however, the institution win have to absorb these costs. Data Interpretation Costs of data storage and interpretation will depend on the scope of the project (see "Examples from Actual Studied. Microcomputers win often be sufficient for analysis, and their costs should be budgeted. Statis- tical consultation win be an impomnt element of the costs of data inter- pretation. UNREIMBURSED PATIENT COSTS Studies of diagnostic technology may involve costs to patients that are not reimbursed. Loss of wages due to time spent away from work is an example. Furler, a padent may suffer adverse effects from the tests. The more invasive or t~me-consuming the technology, the greater these unre- imbursed costs are likely to be. They will not directly affect the cost of performing a study, but they may affect patients' willingness to pa~ci- pate in the project.

OCR for page 107
COSTS AND SOURCES OF FUNDING EXAMPLES FROM ACTUAL STUDIES 113 We win now illustrate some of the costs described earlier in this chapter by examining past and current studies of diagnostic technology. We win look at studies of different scope and see differences In the magnitude of the relative cost components. Cost of Diagnostic Tests Reliable information about the Rue economic cost of diagnostic tests is difficult to obtain. Both direct and indirect costs should be considered CTravers and K~chmal 1988~. Direct costs include the cost of equipment and labor that can be directly related to a particular test. Indirect costs include labor that cannot be directly tied to a particular test (for example, supervisory and administrative personnel) and He costs associated with insurance, maintenance, depreciation, power, and the like. In the past, such information has been impossible to obtain, but cost accounting systems are now being developed that win make it possible to develop accurate estimates of test cost Cravers and Krochn~al 1988, Travers in press). Until these systems are widely utilized, however, we must esti- mate the cost of diagnostic tests from the amount charged for Me test. Charges are often a poor estimate of the true cost of a test because the charge reflects factors in addition to the cost of producing the service. Nonetheless, charges are usually the only available information about test cost. In 1978 Alderson and colleagues prospectively compared CI, ultra- sound, and technetium scans of the liver in patients with known breast or colon cancer (Alderson et al. 1983~. The authors studied 189 patients, 122 of whom had aD three studies. The aim of the study was to construct ROC curves for the three tests. The total of charges for the diagnostic tests, including professional fees, was at least $125,000 in 1978 (the year the study was performed). It would cost approximately $163,000 to perform the tests in 1988, given that year's charges for Me tests. If Me study paid the fills professional fees, triplicate readings of fihns (to assess interob- se~ver variation) would increase the 1978 total by $48,000, and the 1988 total by $91,000. In this study the cost of the diagnostic tests was borne by third-party payers. Under different circumstances, the funding agency might have to support the expense (see the description of the PIOPED project, below).

OCR for page 107
114 Personnel Costs ASSESSMEI~ OF DIAGNOSlIC TECHNOLOGY A current study of ~! provides an example of personnel costs. This study is designed to examine He cost-effec~veness of MRI and to develop new methodolog~c approaches to technology assessment (Mushlin 19X81. The study involves a substantial amount of methodolog~c research in the first year. Some of the first-year costs would thus not be applicable to other studies, but a significant component of the early effort involves decision-analytic modeling of the cI'n~cal problem as a means to guide the design of Me study protocol. Thus, the study is an example of the approach to technology assessment that we have advocated. The total budget for the project is about $1 million, including the indirect costs that cover overhead and over institutional expenses. Of the budgeted direct costs, personnel costs account for 95 percent. The person- nel include the investigators, research assistants, and a~ninistradve and support staff. Supply and travel costs each amount to ~ percent of the direct costs; data analysis accounts for about 2 percent. No money is budgeted for the tests or patient care, because third-party payers have agreed to fund the costs of the additional tests required for the compara- tive analysis. Multi-Institutional Studies The Prospective Investigation of Pulmonary Embolic Diagnosis proj- ect is one of the most ambitious diagnostic technology assessments ever attempted. The study compared ventilation-perfusion scanning and pul- monary angiography in the diagnosis of pulmonary embolism. The total cost to the NIH was $7.6 million, which included direct and indirect institutional costs (Vre~m 19881. The study involved six clinical centers and a data analysis center. The intervention group had 951 patients; the "usual care', group contained 568 patients. The data analysis center was budgeted for $~.8 million, or about 24 percent of the total NIH costs Clable 5.~. Personnel costs, including the cost of consultants, accounted for 60 percent of the total budget (76 percent of direct costs) of tile data analysis center, and for approximately 50 percent of the total budget (80 percent of direct costs) of the clinical centers CTable 5.2~. The second largest item in the budget of the clinical centers was the cost of angi- ogrmns in the intervention group (about $600,000 for the six clinical centers). The NIH did not pay for angiograms in the "usual care" group.

OCR for page 107
com AND SOURCE OF FUNDING TABLE 5.! Five-Year Budget of He PlOPED Data-Analys~s Center Cost Element Amount (dodders) Professional lab" Progranune2 1ab" Clencal labor Fnnge benefits Constants Travel Office equnpenent Computer services Building rent Office costs Indirect costs Total 230,000 321,000 241,000 140,000 140,000 57,000 35,000 125,000 66,000 54,000 415,000 1,824,000 _ . . TABLE S.2 Five-Year Budget of a Representative PlOPED Clinical Center (one of six centers) Cost Element Amount (dollars) Investigator labor Technician labor Clerical labor Fnnge benefits Consultants Travel Office equipment Angiograms Over direct costs Direct costs Total 142,000 174,000 78,000 117,000 6,000 17,000 3,000 98,000 10,000 351,000 996,000 115

OCR for page 107
116 ASSESSMENT OF DIAGNOSTIC TECHNOLOGY Thus, $7.6 minion represents an underesiim ate of the total cost of the study. The PIOPED study shows that as the scale of investigation increases, the relative size of Me cost components may change. Total data analysis costs win be larger at this grand scale. The cost per unit of data analyzed may increase or decrease, however, depending on the efficiency of the data processing. Much of the expense budgeted to the data analysis center was for personnel. In addition, development of the protocols took ap- proximately 15 months; clearly, In studies of this size, the planning and protocol development phase win involve substantial effort and expense. SOU1ICES OF FUNDING Funding for technology assessment and for clinical teals is likely to come from similar sources (Institute of Medicine [IOM] 1985, Meinert 1982~: government institutions, private foundations, industry, and third- party payers. Nonetheless, most observers feel that funding for technol- ogy assessment~has been inadequate (IOM 1985, Fineberg and Hiatt 1979~. Rapid dissemination of technology removes any incentive for manu- facturers to fund technology assessment. As we have noted, technology often becomes widely used before it has been adequately assessed. Fetal monitoring and MRT are weHA-known examples. Under these conditions, manufacturers have no incentive to fund technology assessment, because the results may only serve to decrease Me use of the manufacturer's product. Meinert (1986) analyzes this issue in the pharmaceutical indus- try. Antitrust laws may impede funding of technology assessment by third- party payers (IOM 1985, Rose and Leibenlutt 1986~. Generally speaking, antitrust law is applicable to situations in which there is agreement or concerted action between two entities that leads to an unreasonable re- straint of trade. Thus a technology assessment financed by third-party payers could, In principle, be subject to challenge under antitrust legisla- tion if the results of the assessment serve to reduce or foreclose the use of a device or procedure. As Rose and Leibeniutt- ~1986) observe, "even conduct that may ultimately be considered legal may nevertheless be the subject of very costly and lengthy litigation" (p. 1492~. Because the results of technology assessment will be public informa- tion, there is lime incentive for industry or third-party payers to provide funding when it is likely Mat some other party win. This reasoning has

OCR for page 107
COSTS AND SOURCES OF FUNDING 117 led to a variety of proposals for involving insurers and industry in the technology assessment effort. Proposals for Funding Various suggestions have been made about how technology assess- ment should be funded TOM 1985~. In 1980, Reknan caned for a "major new national program of support for the evaluation of medical procedures of all kinds" (p. 154~. He suggested that the work be done primarily in the pnv ate sector and that large-scale funding, $200 minion to $300 minion annually, would be necessary. The proposal recommended that an alloca- tion of 0.2 percent of the Heath Care Financing Administration budget (about $100 minion in 1980) and a proportion of We budgets of private Bird pardes be earmarked for technology assessment. In an analysis of the effects of reimbursement on biomedical innova- tion, Bunker et al. (1982) suggest Hat insurance coverage of new ~era- pies be contingent on their adequate assessment. TOM reports recom- mend establishing a publi~pnvate sector consortium for technology as- sessment (IOM 1983, IOM 1985~. The consortium would begin with start-up funds from Congress and then be supported from an endowment to be raised by pooling the funds of payers, foundations, professional associations, and other users of the assessments. As noted in the Introduc- tion, the Council on Heals Care Technology, a nongovernmental arm of the TOM, is an indirect outgrowth of the IOM reports. The council does not, however, play a direct role in He financing of technology assessment. The proposals share the common theme that private insurers, industry, and government should share the financial burden of technology assess- ment. A formal mechanism for ensuring such cooperation has not been established. SUMMARY AND CONCLUSIONS This chapter has explored the costs of studies of diagnostic technology. The major expenses in these studies win be for personnel and diagnostic tests. The investigator should anticipate expenses associated with the following: protocol development decision-analytic modeling costs associated with patient accrual

OCR for page 107
118 ASSESSMENT OF DIAGNOSTIC TECHNOLOGY data collection and analysis clinical follow-up. The sample size of the study will strongly influence the total cost; ques- ffons of power will therefore warrant careful consideration. Funding of technology assessment has been insufficient. Most propos- als on the subject recommend that the public and the private sectors share fiscal responsibility. The long-term cost of poorly perfonned technology assessment is likely to outweigh by far He more immediate cost of well- designed studies of diagnostic technology. REFERENCES Abrams, Ho., and Hessel, S. Health technology assessment: Problems and challenges. American Joumal of Roentgenology 149:1127-1128, 1987. Alderson, P.O., Adams, D.F., McNeil, B.J., et al. Computed tomography, ultrasound, and scintigraphy of the liver in patients with colon or breast carcinoma: A prospective comparison. Radiology 149:225- 230, 1983. Bunicer, I.P., Fowles, I., and Schaffarzick, I. Evaluation of medical- technology strategies: Effects of coverage and reimbursement. Parts I and IT. New England Journal of Medicine 306:620-624, 687-692, 1982. Detsky, A.S. Using economic analysis to determine the resource conse- quences of choices made in planning clinical trials. Joumal of Chronic Diseases 38:753-765, 1985. Fineberg, H.V., and Hiatt, H.H. Evaluation of medical practices: The case for technology assessment. New England Journal of Medicine 301:1086-1091, 1979. Henderson, W.G. Some operational aspects of the Veterans Administra- tion cooperative studies program from 1972 to 1979. Controlled Clinical Trials 1:209-226, 1980. Hulley, S.B., and Cummings, S.R., eds. Designing Clinical Research. Baltimore, Williams and WiLkins, 1988. Institute of Medicine. Medical Technology Assessment: A Plan for a Public/Private Sector Consortium. Washington, D.C., National Acad- emy Press, 1983. Institute of Medicine. Assessing Medical Technologies. Washington D.C., National Academy Press, 1985.

OCR for page 107
COSTS AND SOURCES OF FUNDING 119 Levy, Ret., and Sondik, E.~. Large-scale clinical trials: Are Hey worm ache cost? Annals of the New York Academy of Sciences 382:411-422, 1982. McNeil, B.~. Pitfalls in and requirements for evaluations in diagnostic technologies. In Wagner, I., ea., Proceedings of a Conference on Medical Technologies. DREW Pub. No (PHS) 79-3254. Washington D.C., U.S. Govemment Pdnting Office, 1979:33-39. McNeil, B.~., Sanders, R., Alderson, P.O., et al. A prospective study of computed tomography, ultrasound, and gallium imaging in padents with fever. Radiology 139:647-653, 1981. Meinert, C.~. Funding for clinical teals. Controlled Clinical Trials 3:165- 171, 1982. Meine~t, Cab. Clinical Tnals: Design, Conduct and Analysis. New York, Oxford University Press, 1986. Meydrech, E.F., and Kupper, Lid. Cost considerations and sample size requirements in cohort and case-control studies. American Joumal of Epidemiology 107:201-205, 1978. Mushlin, A.~. Personal communication, 1988. Piantadosi, S., and Patterson, B. A method for predicting accrual, cost, and paper flow in clinical Dials. Controlled Clerical Tnals 8:202-215, 1987. Reman. A.S. Assessment of medical practice A simple proposal. New England Joumal of Medicine 303:153-154, 1980. a-- r - Rose, M., and Leibenlutt, R.F. Antitrust implications of medical technol- ogy assessment. New England Journal of Medicine 314:149~1493, 1986. Rosenberg, M.J. Cost efficiency in study planning and completion. Amer~- can Joumal of Medicine 75:833-838, 1983. Travers, E.M., and K~ochm al, C.F. A new method for determining test cost per instrument. Medical Laboratory Observer 20:24-29, 1988. Travers, E.M. Managing costs in clinical laboratories. In Laboratory Microcost Anaysis: Developing Instrument and Test Costs. New Yori`, McGraw-Hill, in press. Vreim, C. Project officer, Prospective Investigation of Pulmonary Em- bolic Diagnosis project Q,IOPED). Personal communication, 1988.