The Feasibility of Economic Evaluation ofDiagnostic Procedures: The Case of CT

Scanning

April 1981

NTIS order #PB81-209777

C A S E S T U D Y # 2

THE IMPLICATIONS OF

COST-EFFECTIVENESSANALYSIS OF

MEDICAL TECHNOlogy

APRIL 1981

BACKGROUND PAPER #2: CASE STUDIES OFMEDICAL TECHNOLOGIES

C A S E S T U D Y #2: T H E F E A S I B I L I T Y O F E C O N O M I C E V A L U A T I O N

O F D I A G N O S T I C P R O C E D U R E S : T H E C A S E O F C T S C A N N I N G

Judith L, Wagner, Ph. D.Senior Research Asociate, The Urban Institute

Wshington, D.C.

I 1OTA Background Papers are documents that contain information believed to beuseful to various parties. The informtion under-girds formal OTA assessments or isa n outcome of interna exploratory planning and evaluation. The material is usuallynot of immediate policy intereest such as is contained in an OTA Report or TechnicalMemorandum, nor does it present options for Congress to consider.

1 J

Library of Congress Catalog Card Number 80-600161

For sale by the Superintendent of Documents,U.S. Government Printing Office, Washington, D.C. 20402

Foreword

This case study is one of 17 studies comprising Background Paper #2 for OTA’sassessment, The Implications of Cost-Effectiveness Analysis of Medical Technology.That assessment analyzes the feasibility, implications, and value of using cost-effec-tiveness and cost-benefit analysis (CEA/CBA) in health care decisionmaking. The ma-jor, policy-oriented report of the assessment was published in August 1980. In additionto Background Paper #2, there are four other background papers being published inconjunction with the assessment: 1) a document which addresses methodologicalissues and reviews the CEA/CBA literature, published in September 1980; 2) a casestudy of the efficacy and cost-effectiveness of psychotherapy, published in October1980; 3) a case study of four common diagnostic X-ray procedures, to be published insummer 1981; and 4) a review of international experience in managing medical tech-nology, published in October 1980. Another related report was published inSeptember of 1979: A Review of Selected Federal Vaccine and Immunization Policies.

The case studies in Background Paper #2; Case Studies of Medical Technologiesare being published individually. They were commissioned by OTA both to provideinformation on the specific technologies and to gain lessons t-hatthe broader policy aspects of the use of CEA/CBA. Several of thecally requested by the Senate Committee on Finance,

Drafts of each case study were reviewed by OTA staff; by

could be applied tostudies were specifi-

members of the ad-visory panel to the overall assessment, chaired by Dr. John Hogness; by members ofthe Health Program Advisory Committee, chaired by Dr. Frederick Robbins; and bynumerous other experts in clinical medicine, health policy, Government, and econom-ics. We are grateful for their assistance. However, responsibility for the case studies re-mains with the authors.

Director

Advisory Panel on The Implications ofCost= Effectiveness Analysis of Medical Technology

John R. Hogness, Panel ChairmanPresident, Association of Academic Health Centers

Stuart H. Altman Sheldon LeonardDean ManagerFlorence Heller School Regulatory AffairsBrandeis University General Electric Co.

James L. BenningtonChairmanDepartment of Anatomic Pathology and

Clinical LaboratoriesChildren’s Hospital of San Francisco

John D. ChaseAssociate Dean for Clinical AffairsUniversity of Washington School of Medicine

Joseph FletcherVisiting ScholarMedical EthicsSchool of MedicineUniversity of Virginia

Clark C. HavighurstProfessor of LawSchool of LawDuke University

Barbara J. McNeilDepartment of RadiologyPeter Bent Brigham Hospital

Robert H. MoserExecutive Vice PresidentAmericans College of Physicians

Frederick MostellerChairmanDepartment of BiostatisticsHarvard University

Robert M. SigmondAdvisor on Hospital AffairsBlue Cross and Blue Shield Associations

Jane Sisk WillemsVA ScholarVeterans Administration

OTA Staff for Background Paper #2

Joyce C. Lashof, Assistant Director, OTAHealth and Life Sciences Division

H. David Banta, Health Program Manager

Clyde J. Behney, Project Director

Kerry Britten Kemp, * EditorVirginia Cwalina, Research Assistant

Shirley Ann Gayheart, SecretaryNancy L. Kenney, Secretary

Martha Finney, * Assistant Editor

Other Contributing Staff

Bryan R. Luce Lawrence Miike Michael A. RiddioughLeonard Saxe Chester Strobel’

OTA Publishing Staff

John C. Holmes, Publishing Officer

John Bergling* Kathie S. Boss Debra M. Datcher

Patricia A. Dyson* Mary Harvey* Joe Henson

Preface

This case study is one of 17 topics being is-sued that comprise Background Paper #2 to theOTA project on the Implications of Cost-Effec-tiveness Analysis of Medical Technolology. * Theoverall project was requested by the SenateCommittee on Labor and Human Resources. Inall, 19 case studies of technological applicationswere commissioned as part of that project.Three of the 19 were specifically requested bythe Senate Committee on Finance: psychother-apy, which was issued separately as Back-ground Paper #3; diagnostic X-ray, which willbe issued as Background Paper #5; and respira-tory therapies, which will be included as part ofthis series. The other 16 case studies were se-lected by OTA staff.

In order to select those 16 case studies, OTA,in consultation with the advisory panel to theoverall project, developed a set of selectioncriteria. Those criteria were designed to ensurethat as a group the case studies would provide:

●

●

●

●

●

●

●

examples of types of technologies by func-tion (preventive, diagnostic, therapeutic,and rehabilitative);examples of types of technologies by physi-cal nature (drugs, devices, and procedures);examples of technologies in different stagesof development and diffusion (new, emerg-ing, and established);examples from different areas of medicine(such as general medical practice, pedi-atrics, radiology, and surgery);examples addressing medical problems thatare important because of their high fre-quency or significant impacts (such ascost );examples of technologies with associatedhigh costs either because of high volume(for low-cost technologies) or high individ-ual costs;examples that could provide informativematerial relating to the broader policy andmethodological issues of cost-effectivenessor cost-benefit analysis (CEA/CBA); and

. examples with sufficient evaluable litera-ture.

On the basis of these criteria and recommen-dations by panel members and other experts,OTA staff selected the other case studies. These16 plus the respiratory therapy case study re-quested by the Finance Committee make up the17 studies in this background paper.

All case studies were commissioned by OTAand performed under contract by experts in aca-demia. They are authored studies. OTA sub-jected each case study to an extensive reviewprocess. Initial drafts of cases were reviewed byOTA staff and by members of the advisorypanel to the project. Comments were providedto authors, along with OTA’s suggestions forrevisions. Subsequent drafts were sent by OTAto numerous experts for review and comment.Each case was seen by at least 20, and some by40 or more, outside reviewers. These reviewerswere from relevant Government agencies, pro-fessional societies, consumer and public interestgroups, medicaI practice, and academic med-icine. Academicians such as economists and de-cision analysts also reviewed the cases. In all,over 400 separate individuals or organizationsreviewed one or more case studies. Although allthese reviewers cannot be acknowledged indi-vidually, OTA is very grateful for their com-ments and advice. In addition, the authors ofthe case studies themselves often sent drafts toreviewers and incorporated their comments.

These case studies are authored workscommissioned by OTA. The authors are re-sponsible for the conclusions of their spe-cific case study. These cases are not state-ments of official OTA position. OTA doesnot make recommendations or endorse par-ticular technologies. During the variousstages of the review and revision process,therefore, OTA encouraged the authors topresent balanced information and to recog-nize divergent points of view. In two cases,OTA decided that in order to more fullypresent divergent views on particular tech-nologies a commentary should be added tothe case study. Thus, following the case

studies on gastrointestinal endoscopy andon the Keyes technique for periodontal dis-ease, commentaries from experts in the ap-propriate health care specialty have beenincluded, followed by responses from theauthors.

The case studies were selected and designed tofulfill two functions. The first, and primary,purpose was to provide OTA with specific in-formation that could be used in formulatinggeneral conclusions regarding the feasibility andimplications of applying CEA/CBA in healthcare. By examining the 19 cases as a group andlooking for common problems or strengths inthe techniques of CEA/CBA, OTA was able tobetter analyze the potential contribution thatthese techniques might make to the managementof medical technologies and health care costsand quality. The second function of the caseswas to provide useful information on the spe-cific technologies covered. However, this wasnot the major intent of the cases, and theyshould not be regarded as complete and defini-tive studies of the individual technologies, Inmany instances the case studies do represent ex-cellent reviews of the literature pertaining to thespecific technologies and as such can stand ontheir own as a useful contribution to the field. Ingeneral, though, the design and the fundinglevels of these case studies was such that theyshould be read primarily in the context of theoverall OTA project on CEA/CBA in healthcare.

Some of the case studies are formal CEAS orCBAS; most are not. Some are primarily con-cerned with analysis of costs; others are moreconcerned with analysis of efficacy or effec-tiveness. Some, such as the study on end-stagerenal disease, examine the role that formalanalysis of costs and benefits can play in policyformulation. Others, such as the one on breastcancer surgery, illustrate how influences otherthan costs can determine the patterns of use of atechnology. In other words, each looks at eval-uation of the costs and the benefits of medicaltechnologies from a slightly different perspec-

tive. The reader is encouraged to read this studyin the context of the overall assessment’s objec-tives in order to gain a feeling for the potentialrole that CEA/CBA can or cannot play in healthcare and to better understand the difficulties andcomplexities involved in applying CEA/CBA tospecific medical technologies.

The 17 case studies comprising BackgroundPaper #2 short titles and their authors are:

Artificial Heart: Deborah P. Lubeck and John P.Bunker

Automated Multichannel Chemistry Analyzers:Milton C. Weinstein and Laurie A. Pearlman

Bone Marrow Transplants: Stuart O. Schweitz-er and C. C. Scalzi

Breast Cancer Surgery: Karen Schachter andDuncan Neuhauser

Cardiac Radionuclide Imaging: William B.Stason and Eric Fortess

Cervical Cancer Screening: Bryan R. LuceCimetidine and Peptic Ulcer Disease: Harvey V.

Fineberg and Laurie A. PearlmanColon Cancer Screening: David M. EddyCT Scanning: Judith L. WagnerElective Hysterectomy: Carol Korenbrot, Ann

B, Flood, Michael Higgins, Noralou Roos,and John P. Bunker

End-Stage Renal Disease: Richard A. RettigGastrointestinal Endoscopy: Jonathan A. Show-

stack and Steven A. SchroederNeonatal Intensive Care: Peter Budetti, Peggy

McManus, Nancy Barrand, and Lu AnnHeinen

Nurse Practitioners: Lauren LeRoy and SharonSolkowitz

Orthopedic Joint Prosthetic Implants: Judith D.Bentkover and Philip G. Drew

Periodontal Disease Interventions: Richard M.Scheffler and Sheldon Rovin

Selected Respiratory Therapies: Richard M.Scheffler and Morgan Delaney

These studies will be available for sale by theSuperintendent of Documents, U.S. Govern-ment Printing Office, Washington, D. C. 20402.Call OTA’s Publishing Office (224-8996) foravailability and ordering information.

Case Study #2

The Feasibility of EconomicEvaluation of Diagnostic Procedures:

The Case of CT Scanning

Judith L. Wagner, Ph. D.Senior Research Associate

The Urban InstituteWashington, D.C.

Contents

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

LIST OF FIGURES

Figure Ale. Pagel. A Model for Economic Evaluation of Diagnostic Procedures . . . . . . . . . . . . . . . 62. Disease Detection Rates for Two Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193. Case-Finding Costs for 100 Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Case Study #2:

The Feasibility of EconomicEvaluation of Diagnostic Procedures:

The Case of CT ScanningJudith L. Wagner, Ph. D.

Senior Research AssociateThe Urban Institute

Washington, D.C.

INTRODUCTION

In the past, the evaluation of medical proce-dures—when it has occurred at all—has focusedon the assessment of the procedures’ effective-ness in altering the status of patients’ health orimproving the accuracy of diagnosis. Economiccriteria, such as the immediate or induced costsof a procedure, have seldom been included inthe evaluation.

Recently, however, as the Nation’s ability topay for new advances in medical procedures hasbeen increasingly taxed, interest in economicevaluation has been rising, and the applicationof economic criteria has gained acceptance. Onemanifestation of the increasing legitimacy ofeconomic evaluation of medical technologies isthe passage of the Health Services Research,Health Statistics, and Health Care TechnologyAct of 1978 (Public Law 95-623), establishinga National Center for Health Care Technologywhich is mandated to consider, among otherthings, the cost effectiveness of medical tech-nologies.

In spite of this new interest, the actual appli-cation of economic evaluation to medical proce-dures remains limited. Except for preventiveservices such as immunization and asymptomat-ic screening (11,31,32,33,34,40), little effort hasbeen devoted to performing this type of evalua-tion, Although some recent literature representsa start in this direction (42), the territory ofeconomic evaluation of medical procedures re-mains largely uncharted.

Nowhere do questions of the feasibility andusefulness of economic evaluation arise morethan in the area of evaluating diagnostic proce-dures. Diagnostic testing has been widely char-acterized as subject to uncontrolled and explo-sive growth (16), and the existence of tradeoffsbetween the information obtained and the costof such testing is often noted (22). But just howsuch tradeoffs should be measured is an issuethat has not been pursued at length.

Computed tomographic (CT) scanning wasintroduced in 1973. Initially limited to the head,CT scanning can now be used to detect diseasesin other parts of the body. The use of this diag-nostic technology has initiated a controversy ofunprecedented proportions. This controversyhas been cast in terms of the tradeoff betweenthe benefits and costs of CT scanning. Conse-quently, a great deal of literature extendingbeyond the boundaries of traditional clinicalevaluation has emerged. Numerous articles andreports have been published that describe thehistorical pattern of diffusion of CT scanners inhospitals and private practices throughout theUnited States (5,10,35), the operating costs ofCT scanning units (13,14), patterns of utiliza-tion of CT scanners in particular institutions(2,25,29), and most important here, the cost ef-fectiveness of CT scanning (15,17,23,24,41,46).Many of these studies have been reported inclinical journals, demonstrating the intense in-terest of the medical community not only in the

.3

clinical uses of CT scanning, but also in the im-pact of this new technology on the practice ofmedicine. Interestingly, almost all of the studiesof the economics of CT scanning have been au-thored by physicians or scientists without majorcollaboration with economists.

Judging by the quantity of published re-search, one would surmise that the state of theart of economic evaluation of diagnostic proce-dures—CT scanning, in particular—is well ad-vanced. In fact, the opposite is true. The studiesof the cost effectiveness or economic benefit ofCT scanning reveal major conceptual and meth-odological weaknesses that not only mitigatethese particular studies’ value, but also castdoubt on the overall feasibility of conductingeconomic evaluations of diagnostic procedures.

The purpose of this case study is to explorethe feasibility of economic evaluation of diag-nostic procedures. Several questions are ad-dressed. To what extent can cost-benefit anal-ysis (CBA) and its derivative, cost-effectivenessanalysis (CEA)—the cornerstones of economicevaluation methodology—be applied to diag-nostic procedures? What, if any, methodologi-cal and- conceptual obstacles toanalyses are there? How usefulsuch analyses?

CBA and CEA are methodseating scarce resources among

performing suchare the results of

to assist in allo-alternative uses,

These methods were developed primarily toevaluate large public-sector investments such ashighways, dams, and airports. When applied todiagnostic procedures, they are intended to pro-vide information on two related questions: 1)Under what circumstances should the procedurebe performed? and 2) How much investment incapacity to perform the procedure is justified?The answer to the second question rests on thor-

ough study of the first, for only by knowingwhen a procedure should be performed can oneassess how much investment in capacity isjustified.

The fundamental argument of this case studyis that the conceptual, methodological, and dataproblems inherent in answering the first ques-tion (i.e., when should a given diagnostic pro-cedure be used?) often create such inaccurateand unreliable results that CEA/CBA is general-ly not worthwhile. In only a few situations (tobe enumerated later) is such economic evalua-tion likely to be particularly illuminating andtherefore useful to medical decisionmakers.

Such a pessimistic prognosis for the applica-tion of CBA and CEA needs to be explained andsupported with evidence. In this case study,therefore, we begin with an idealized model ofeconomic evaluation of diagnostic proceduresand then use this model to explore the implica-tions of problems that occur in attempts to ap-ply it. Some typical second best approaches, de-signed to circumvent these problems, are thendescribed, with particular emphasis on the useof these approaches in the evaluation of CTscanning.

The economic evaluations of CT scanning in-clude three general analytic approaches: 1) goodfaith but limited attempts to apply the principlesof economic evaluation, 2) estimates of the im-pact of CT scanning technology on health careexpenditures, and 3) analyses of the amount andlocation of CT scanning capacity required tomeet a predetermined level of demand. Each ofthese approaches is assessed with respect to itslimitations as valid economic evaluation and theparticularquestionsusefulness

methodological weaknesses that raiseas to the ultimate feasibility andof economic evaluation.

CBA AND CEA: THE THEORETICAL IDEAL

The purpose of CBA and CEA is to help deci-sionmakers determine the best allocation of re-sources among possible alternative uses. CBA isused to measure the difference between the val-

ue of all benefits resulting from an activity, bothat present and in the future, and the costs. If thisdifference, or net social benefit, is positive, theactivity would be said to be worth its costs, and

the decision should be made to allocate scarceresources (labor, capital, etc. ) to producing thebenefits.

Straightforward as this decisionmaking toolis, the benefits accruing from an activity, par-ticularly those undertaken in the public sector,are often difficult even to identify, much less tovalue. Because CBA requires enumeration andvaluation of all significant benefits, analystshave derived the seemingly simpler technique ofCEA. Here, a single measure of effectiveness isdesignated, and the ratio of cost to effectivenessguides resource allocation decisions amongcompeting alternatives.

Applying these principles to medical proce-dures, CBA would enumerate and place a valueon all benefits (both positive and negative) de-rived from performing a procedure on patientswith a specified set of conditions and wouldcompare those benefits to the cost of performingthe procedure. The resulting net social benefitwould indicate whether the procedure should beperformed under the specified conditions. InCEA, a measure of procedure effectivenesswould be designated, and the ratio of that singlemeasure to cost would be the critical item forresource allocation. Lives saved, life-yearssaved, quality-adjusted life-years saved, disabil-ity-days saved, and age-adjusted disability dayssaved are measures of effectiveness often chosenin studies of health care programs (43).

Although CEA eliminates the problem ofplacing a monetary value on benefits, its scopeis limited. While the scope of CBA is the wholesociety, CEA’S is limited to the comparison ofstrategies with similar impacts. Since diagnosticprocedures may have widely varying effects—some procedures may affect mortality whileothers may affect only the quality of life—CEAcan only be used to compare a narrow set ofprocedures with similar purposes.

The application of CBA and CEA to diagnos-tic procedures can be described with reference tofigure 1. A set of patients with specific present-ing signs and symptoms faces two or more alter-native diagnostic pathways, each representing aspecific combination of diagnostic tests. These

alternative pathways may represent differentdiagnostic tests or even different testing se-quences. For example, pathway A might repre-sent the administration of a radionuclide (RN)brain scan followed by a CT head scan if the RNscan is positive or equivocal, while pathway Bmight represent a CT scan only. Each of thesetwo diagnostic pathways carries with it certainresource costs and has specific diagnostic re-sults. Four possible diagnostic results are shownfor each pathway: true positive, true negative,false positive, and false negative. Each of theseoutcomes implies a course of therapy (with neg-ative findings implying no therapy), each in-volving its own resource costs. Finally, the ther-apies result in health outcomes measured byrelevant indicators. In CEA, a single indicatoror an index of indicators is chosen, and the im-pact of the pathway on this indicator is com-pared to the total resource costs (both therapeu-tic and diagnostic) associated with the pathway.The diagnostic pathway with the lowest ratio ofcost to effectiveness is preferred. In CBA, allrelevant outcomes, both good and bad, wouldbe measured and their value assessed.

Notice that the purpose of economic evalua-tion is to decide under what conditions, if any, adiagnostic procedure should be used. Typically,a procedure will be of undisputed and obviousvalue for some groups of patients. The impor-tant question for resource allocation is notwhether the diagnostic procedure or the equip-ment producing it is justified or whether theprocedure raises or lowers health care costs, buthow the procedures should be used in the prac-tice of medicine. Only by knowing the costs andeffects of treating each group of patients withthe procedure is it possible to know whetherthere is too much or too little capacity to per-form the procedure in any community.

As seemingly obvious as this point is, it hasnot been recognized by the authors of manystudies purporting to be analyses of the cost ef-fectiveness of CT scanning. Too many of thesestudies, whose titles include the term “cost effec-tiveness, ” are simply historical descriptions ofthe aggregate impact of CT scanning on health

Figure 1 .—A Model for Economic Evaluation of Diagnostic Procedures

I

I Dx(A) +

I

❑Dx(A)

Pat ientswith

specificsigns,

symptoms, orother

demographiccharacteristics

Dx(B)

I

II False -A A

care costs in a hospital or community, These proved by increasing, decreasing, or rearrang-studies are plagued with measurement prob- ing CT scanning capacity in the United States. 1-

lems, but even without those problems, theI Similarly, estimating the impact of a new procedure on medi-studies would offer no guidance on whether the care or public program expenditures, while useful in Federal budg-

existing allocation of resources could be im- eting, is not CEA.

PROBLEMS IN APPLYING THE IDEAL MODEL

The ideal model of economic evaluation is however, arise in or are exacerbated by the at-straightforward enough, but in attempts to ap- tempt to apply the model to diagnostic proce-ply it to diagnostic procedures, serious prob- dures. Problems in the following areas merit alems arise. Some of the problems are common closer look:to all economic evaluations. For example, the

●

question of how one should compare costs andeffects accruing at different times has not yetbeen fully resolved. z Several of the problems, ●

identifying homogeneous patient groupsfor analysis,

specifying appropriate diagnostic path-ways,

measuring diagnostic accuracy,

Q measuring diagnostic and therapeutic costs,and

● specifying outcomes of the diagnostic/ ther-apeutic process.

Each of these areas is discussed below.

Identifying HomogeneousPatient Groups

Application of the ideal model of economicevaluation requires dividing the universe ofpatients into reasonably homogeneous groupsbased on their presenting conditions and per-sonal characteristics. The patients in eachgroup, with a common set of signs and symp-toms and perhaps other important characteris-tics, such as age or sex, are treated uniformlywith respect to the diagnostic pathways applied.

The outcome of a CEA can be extremely sen-sitive to the criteria (signs and symptoms) usedto identify a patient group. These criteria can beso general that they lump together patients withvery different problems, or they can be so de-tailed that they essentially create a unique groupfor each individual patient. If, for exampIe, ananalysis of CT head scanning is based on a uni-verse of patients with all types of neurologicalcomplaints, the CEA would average togetherthe costs and outcomes associated with very dif-ferent kinds of patients. Patients with recentstrokes would be analyzed together with pa-tients complaining of persistent headaches. Atthis extreme, the usefulness of economic anal-ysis in discriminating between the appropriateand inappropriate uses of CT scanning is ques-tionable. Conversely, if the criteria are so de-tailed that few patients fall into each group, theanalysis would need to be repeated for each ofthe many groups of patients and would be pro-hibitively expensive. Moreover, such detailedcriteria are impractical for use in diagnosticdecisionmaking because they are likely to in-clude subjective signs that are difficult to incor-porate into a diagnostic protocol.

The problem of patient grouping is frequentlynoted in connection with clinicaI trials (38). Asthe number of groups of patients increases, thesample size required to show statistically signifi-cant effects also increases, possibly exhausting

the supply of patients available for entry into aclinicaI trial. At the same time, aggregating mayobscure significant benefits available to particu-lar subgroups. But the problem is even moreacute in economic evaluation, which must esti-mate costs associated with diagnosing and treat-ing patients. Cost data are rarely available on adisaggregated basis, and allocation of aggregatecosts to individual patient groups involves a lossof accuracy in cost measurement that can bequite serious.

Specifying AppropriateDiagnostic Pathways

The outcome of economic evaluation varieswith the number of diagnostic pathways stud-ied. Omission of significant alternatives canseriously compromise the results. If a new diag-nostic test is developed for a previously undiag-nosable disease, then the pathways in conten-tion are easy to specify: the new test versus notest. But when a new diagnostic procedure addsinformation to that available from an existingarsenal of procedures, then each diagnosticpathway comprises a procedural sequence inwhich the performance of later procedures isconditioned on the achievement of certain re-sults from earlier procedures. Moreover, two se-quences that differ onIy in their rules for con-tinuing the testing process must be consideredseparate diagnostic pathways, since each se-quence will have yields and costs that are differ-ent from those of the other sequence. Hence, thenumber of pathways to be considered may bequite large, and the best pathways may inad-vertently be excluded from analysis.

The importance of comparing alternative test-ing sequences places economic evaluationsquarely at odds with principles for evaluatingthe accuracy of diagnostic procedures. Some ofthe best clinical studies of diagnostic procedureshave measured the diagnostic accuracy of agiven procedure independent of the outcome ofthe tests that precede or follow it. Indeed, a car-dinal principle in measuring diagnostic accuracyof interpretive tests is that the reader or analystshould not be aware of the outcomes of previoustests (1). Yet, if knowledge of such outcomeswould improve the accuracy or timeliness of di-

agnosis in a real clinical setting, and hence thepotential outcomes of therapy, then that effectshould be considered in the economic evalua-tion. Conversely, if knowledge of previous re-sults tends to bias the findings in a particulardirection, and hence leads to more misdiagnosisand inappropriate and costly treatment, thatshould also be considered in the economic eval-uation. Unfortunately, the best efficacy studiesspecifically eschew the measurement of diagnos-tic accuracy in such biased but realistic settings.It is not surprising, then, that estimates of therelative frequency of outcomes of different diag-nostic pathways are often lacking.

Any aspect of diagnosis or treatment thatchanges the parameters of the diagnostic path-way—either diagnostic accuracy, therapeuticeffectiveness, or resource costs—requires defini-tion of a new diagostic pathway. Moreover, aparticular diagnostic pathway assumes a partic-ular mix of inpatient and outpatient treatment.If two treatment protocols are identical exceptfor the setting of care, then the protocols mustbe considered as separate pathways, since theircosts may differ widely.

Technological change, which may improvethe accuracy of a diagnostic test or the effec-tiveness with which the diagnosed disease canbe treated, also requires the specification of newpathways. One accepted way of dealing withuncertainty about or changes in such param-eters without constructing additional pathwaysis use of “sensitivity analysis. ” Sensitivityanalysis systematically assesses the effect ofparameter changes on the solution to the eco-nomic analysis. The question is posed as fol-lows: What percentage change in the diagnosticaccuracy of test X would be needed to induce achange in the analytic findings? The approach ispowerful when only one or two parameters aresubject to change or uncertainty, but it breaksdown under conditions of general uncertaintyabout the appropriate value of many param-eters. Unfortunately, most economic evalua-tions of diagnostic procedures face such condi-tions,

It should now be clear that the need to specifyand compare a large number of alternative path-

ways to one another taxes both available dataand analytical resources.

Specifying Diagnostic Accuracy

The model pictured in figure 1 is constructedas if each alternative diagnostic pathway yieldsonly two diagnostic findings: positive and nega-tive. Often, a positive finding is associated withseveral diseases or different stages of a disease.Hence, estimates of the total sensitivity (i.e.,false-positive rates) and specificity (i.e., truenegative rates)3 of a diagnostic pathway are gen-erally inadequate for economic evaluation, Eachtype of positive finding must be linked to theparticular therapeutic pathway it evokes, sothat the cost of that pathway can be estimated.Thus, for any patient group with specific pre-senting signs and symptoms, accurate economicevaluation requires data not only on the totalsensitivity of a diagnostic pathway, but also onthe rates of detection of the full spectrum of dis-ease states that the diagnostic pathway mightidentify. Such detailed estimates are seldomavailable.

Measuring Diagnostic andTherapeutic Costs

Economic evaluation requires measurementof the diagnostic and therapeutic costs incurredwhen patients embark on a particular pathway.A CBA should measure the cost of all presentand future health care stemming from the healthproblem that precipitated entry into the path-way in the first place. The cost of future healthcare made necessary by a false-negative finding,for example, should be included. In CEA, thecriterion for inclusion of health care costsdepends on the chosen indicator of effective-ness. The use of disability days as the effec-tiveness indicator would obviate the need formeasurement of health care costs associatedwith treament of disability itself. Similarly, acost-effectiveness study using the 5-year sur-vival rate as the outcome measure of interestshould count all costs up to the day of outcome

measurement. The cost of a false negativewould be relevant only if it induced the deliveryof additional medical care prior to the point atwhich effectiveness is measured. Thus, withCEA, there is a natural stopping point to costmeasurement that is dictated by the time hori-zon implicit in the effectiveness measure.

Most of the cost-effectiveness studies of CTscanning have treated the measurement of diag-nostic and therapeutic costs cavalierly, ignoringthe difficult conceptual and measurement prob-lems and taking ad hoc approaches withoutacknowledging their limitations. The authors ofthese studies either do not appreciate the dif-ficulties inherent in cost measurement, or theyrecognize and are overwhelmed by the prob-lems, and therefore ignore them. Nevertheless,the problems are great and may defy solution.

Cost is a measure of value—in this case, thevalue of productive resources (labor, equip-ment, supplies, etc. ) that must be forgone toother uses when a particular diagnostic/ ther-apeutic pathway is chosen for the patient groupunder study. The appropriate economic conceptis that of incremental or avoidable costs, thoseincurred as a direct result of committing a givenincrement of patients to a diagnostic therapeu-tic pathway or that can be avoided if the samepatients do not take the pathway. Avoidablecosts are treated as “variable, ” while unavoid-able costs are assumed to be “fixed .“

The definition of what is avoidable or vari-able depends on the time perspective of the deci-sion and the size of the increment. Generally,the longer the time perspective and the largerthe increment, the more costs are avoidable.Suppose, for example, that the decision iswhether to perform CT scanning on patientswith suspected stroke in a hospital with a scan-ner currently used at .50-percent capacity forother types of patients. In the very short run,the only truly avoidable costs are the materials,supplies, and extra wear and tear on equipmentcaused by serving the stroke patients. Mostlabor costs may be invariant with serving thenew patients, because technicians must be paidfor full workshifts whether they are busy or not.Consequently, the appropriate measure of costwould exclude these fixed operating costs. Also,because the equipment exists to serve other

assume that the service will be provided usingthe most efficient technology currently availableto produce the expected diagnostic and thera-peutic results. Therefore, the appropriate meas-ure of capacity cost is not the original cost of theequipment and plant, but the current cost of re-placing it.

Depending on the rates of technologicalchange and general inflation, replacement costmay be quite different from the original capitalcost of the equipment in any institution. Butwhereas the original cost of existing equipmentis readily available from accounting records,estimation of current costs of replacing capacityis fraught with difficulty. Product improve-ments in CT scanning, for example, have in-creased equipment costs and improved the flex-ibility of CT scanning and the quality of re-sulting images. If these quality improvementsare unnecessary to produce the level of diagnos-tic accuracy specified for the pathway, themeasured costs of CT scanning should not bebased on these new models. Whole-body scan-ners, for example, are increasingly used instudies of the head. Since the purchase price andoperating costs of head scanners are less thanthose of body scanners, it would be incorrect toallocate to head scanning patients a portion ofthe extra costs associated with a body scanner.But how will one estimate the cost of replacinghead scanners when they are no longer manu-factured? The measurement of original capitalcost, while inappropriate, is at least precise.

Whichever concept of capacity cost is used—original capital cost or current replacementcost—it must be allocated between the patientgroup under study and all other patient groups.A common rule of thumb is to allocate thesecosts in proportion to the share of total volumeaccounted for by the patient group under study.Underlying this rule is the implicit assumptionthat the costs of any excess capacity should beborne equally by each patient group using theservice. In fact, this assumption may bias theanalysis against certain patient groups. Head-ache patients, for example, may be able to waitweeks for a CT scan without negative effects onoutcome, but may not have to do so because ex-cess CT scanning capacity has been made avail-

able to serve trauma patients on an emergencybasis. CT scanning costs allocated to the head-ache patients should rightly be calculated as ifthe equipment were used to full capacity, whilethe trauma patient group should bear the entirecost of the excess capacity. Such differential costallocation has not been attempted in the CTscanning papers reviewed by this author.

Several economic studies of CT scanninghave used hospital procedure charges to esti-mate the cost associated with the use of CTscanning. This approach is unacceptable, be-cause procedure charges bear no relation to theincremental costs of treating particular patientgroups. Even if each charge were set to equatethe revenues from a procedure with its associ-ated costs, which is seldom the case, the unitcharge so constructed would be unlikely to re-flect optimal utilization of capacity or appropri-ate allocation of indirect costs and overhead. Itshould be noted, however, that in most Amer-ican hospitals charges for specific proceduresare often completely unrelated to their costs.Profits from one procedure typically subsidizelosses from another. Indeed, the hallmark of agood hospital administrator is that charges havebeen manipulated to maximize revenues. Conse-quently, the use of procedure charges as a proxyfor incremental cost seriously compromises ananalysis. The extensive reliance on procedurecharges in economic studies of CT scanning iscurious in light of the relatively detailed in-formation available on the operating costs ofCT scanning units (13,14). Although the obsta-cles are formidable, improved estimates of thecost of some procedures are within reach.

Another common practice in economic stud-ies is to use physicians’ fees as estimates of thevalue of physicians’ inputs into the diagnosticand therapeutic processes. Although such feesrepresent actual payments, they are not accu-rate reflections of the value of physician inputs.Third-party payer policies have created anoma-lies in fee structures, so that some diagnosticprocedures offer higher returns on the physi-cian’s investment of time and skill than others.In particular, fees for new procedures are oftenset higher than those for established procedures.A better approach to valuing physicians’ serv-

ices would be to use carefully constructed rel-ative value scales based on the time and skill re-quirements of individual procedures. Construc-tion of good relative value scales is expensive,however, and is unlikely to be undertaken aspart of the evaluation of medical procedures.Some recent tentative evidence suggests that theCalifornia Relative Value Scale may be reason-ably well correlated with the level of difficultyof surgical procedures i(19), but more inves-tigation of this question is needed before itwould be possible to recommend reliance onsuch a system.

To summarize, the obstacles to accuratemeasurement of the resource costs associatedwith a diagnostic/therapeutic pathway aregreat, but substantial improvement in cost esti-mates would be possible if researchers would at-tempt to approximate long-run incrementalcost. At present, most cost analyses are grosslyinaccurate, so much so, in fact, that wrong in-ferences have been drawn about the usefulnessof CT scanning.

Specifying Outcomes of theDiagnostic/Therapeutic Process

way. CEA assumes that a single measure or in-dex can capture the essential nature of the out-comes. The former type of analysis challengesour present capacity to objectively quantify alltypes of effects, both good and bad, while thelatter runs the risk of oversimplifying the deci-sionmaking problem.

The problem of specifying and measuringoutcomes is common to all evaluation and istherefore beyond the scope of this case study.One type of outcome, however, is unique toeconomic evaluation of diagnostic procedures:the benefits derived from negative findings.Negative findings are certainly valued by pa-tients, for such findings can have dramatic im-pacts on their peace of mind. The reassurancevalue of negative findings is generally ignored inCEA, because satisfactory methods for estimat-ing changes both in patient satisfaction and inthe value that patients place on those changesare lacking at present (42). Consequently, CEAis likely to be biased against diagnostic/ther-apeutic pathways offering high specificity (i. e.,high true-negative rates) relative to those path-ways with high sensitivity (i. e., high true-posi-tive rates).

CBA requires the specification of all signifi-cant kinds of outcomes of each diagnostic path-

ECONOMIC EVALUATIONS OF CT SCANNING

Only two studies of CT scanning with whichthe author is familiar are good-faith efforts toimplement the ideal model of economic evalua-tion in a comprehensive way. One such study,by Knaus and Wagner (24), is successful; a sec-ond study, by Baker and Way (4), represents afailure to understand and carry out the prin-ciples of economic evaluation.

Knaus and Wagner (24) analyzed the effec-tiveness and costs of two alternative diagnosticpathways for patients presenting with suddensevere headaches but no focal signs; such pa-tients are generally viewed with skepticism ascandidates for CT scanning. The analysts in thisstudy compared the outcomes of two diagnosticpathways: 1) performing a CT scan on all such

patients referred; and 2) performing a CT scanonly on patients presenting additional neurolog-ical signs (i. e., depriving the patient of a C Tscan unless and until his or her condition deteri-orates). Only one possible positive finding wasconsidered: a ruptured intracranial aneurysmresulting in subarachnoid hemorrhage. Becauseearly diagnosis of this disease significantly im-proves the prognosis of surgical intervention,the authors concluded that the cost of each lifesaved by early detection is about $511,000, or$23,000 per life-year saved.

Knaus and Wagner observed that the analysisdepended on the attainment of a diagnostic yieldat least as high as that obtained at a largeteaching hospital: about one aneurysm per 125

toms, and other factors that might improve thecost-effectiveness ratio for headache patients.

It is instructive to ask why Knaus’ study metso many criteria for good economic evaluation.A critical determinant appears to be the avail-ability of empirical estimates of the outcome ofsurgery, medical therapy, and no interventionin patients in various symptomatic stages ofcerebral aneurysms. Without such probabilityestimates, the study would have foundered. Theavailability of data from clinical trials madepossible the decision analytic approach for anadmittedly narrow set of diagnostic findings forpatients with headaches.

The study by Baker and Way (4) of CT scan-ning in the abdomen is a manifest demonstra-tion of what can go wrong when clinicians at-tempt to apply the principles of economic eval-uation. The authors randomly selected the casehistories of approximately 200 hospitalized pa-tients who had received body scans in 1976 or1977. Each patient was classified according tothe organ system studied (e. g., biliary, liver,etc. ). Baker and Way then constructed an arbi-trary scale of effectiveness, where a value of 1was assigned to a patient for whom the CT scanhad a lifesaving effect, a value of 18 was as-signed to a patient if the CT scan actually pre-cipitated a death, and a value of 10 indicated nonet value. The costs of CT scans and other re-lated diagnostic tests were presumably esti-mated from the procedure charges, although theauthors did not specify their methods. A ratio ofcost to effectiveness was constructed for eachorgan system, with the effectiveness scale valueas the denominator. The authors concluded that“the differences between the categories [organsystems] are insignificant, and the average resultsuggests that, in general, the value of the scansto medical care in relation to the cost of scanswas low, ”

Apart from the obvious question of the reli-ability of subjective assessments of effectivenessand the cost measurement techniques, two criti-cisms of Baker and Way’s study are relevanthere. First, it is utterly meaningless to comparethe cost-effectiveness ratio of CT scanningacross patient groups. The appropriate compar-ison is between CT scanning and other diagnos-

of CT scanning on the cost of diagnosis is alogical starting point but is generally an insuffi-cient approach to economic evaluation.

The fundamental weakness of the studies re-maining to be discussed, therefore, is not theirfailure to consider therapies and outcomes.Rather, it is their choice of diagnostic pathways.Each one of these studies compares actual pat-terns of diagnosis with and without CT scan-ning. Most are pre-CT/post-CT comparisons ofdiagnostic costs in an institution or country.Given the system of medical practice in force to-day, however, there is no uniformity fromphysician to physician in the use of diagnosticprocedures on similar patients. Since one physi-cian may order a battery of tests simultaneous-ly, while another may move cautiously fromone to another, each diagnostic pathway com-prises many approaches to diagnosis. Thus, toshow that the introduction of CT scanning hashistorically raised or reduced diagnostic costs isto show that and nothing more. That findingholds no implications for resource allocation,because the pathways are too amorphous to de-termine how CT scanning should fit into thediagnostic process.

Homogeneous patient groups, however, arelikely to undergo diagnostic strategies that arereasonably uniform. Therefore, empirical stud-ies of the impact of CT scanning on diagnosticcost are more likely to be useful as resourceallocation tools when these studies are concen-trated on a narrow set of presenting signs andsymptoms.

Three studies of specific patient groups withrelatively unambiguous and homogeneous signsand symptoms were reported by Larson and col-leagues at the University of Washington. Theyanalyzed the impact of CT scanning on the utili-zation and cost of diagnostic services for pa-tients with presenting conditions suggestive ofbrain tumors (26), cerebrovascular disease (27),and hydrocephalus (30). Admission forms wereabstracted for all inpatients admitted to the uni-versity-affiliated hospital with signs and symp-toms suggesting one of the three diseases duringthe year prior to and 2 years after the introduc-tion of a CT head scanner. Patterns of service

delivery and medical care in the pre-CT andpost-CT patient groups were compared, includ-ing the utilization of and charges for neurodiag-nostic procedures; the length of hospital stay;the time required to reach final diagnosis; thelevel of detail of the diagnosis; and the type oftherapy employed, CT scanning had a majorimpact on the configuration of diagnostic stud-ies used for each of the three types of patients,but the relative benefits resulting from thesechanges varied among the three.

In the study of patients with suspected braintumors (26), CT scanning significantly reducedthe use of RN brain scanning (from 85 to 38 per-cent of patients), angiography (from 65 to 50percent of patients), and pneumoencephalogra-phy (PEG) (from 15 to 4 percent of patients).The length of hospitalization, speed of diagnos-tic workup, and timing and type of therapy forthese patients, though, did not change signifi-cantly. The overall detectability of brain tumorsincreased markedly after CT scanning becameavailable. Total per-patient hospital charges forneurodiagnostic procedures did not change sig-nificantly. The authors therefore concluded thatCT scanning as it was integrated into the neuro-diagnostic process at the study hospital was costeffective, because it permitted reductions in theuse of invasive and somewhat risky proceduresbut did not raise hospital charges or otherwiseaffect prognosis,

In the case of patients with suspected cerebro-vascular disease (27), the introduction of CTscanning significantly reduced the number oflumbar punctures and RN brain scans, but thetotal per-capita hospital charge for neurodiag-nostic procedures increased about 33 percent.The length of hospitalization and speed of diag-nostic workup did not change, and although thelevel of detail of the discharge diagnosis wasgreatly increased, such detail was unnecessaryto plan therapy. In the judgment of the authors,CT scan results were not essential to establish-ing a diagnosis in these patients but played aconfirming role. In patients with suspected cere-brovascular disease, therefore, CT scanning wasfound to be of questionable value.

The introduction of CT scanning in the diag-nostic workup of children with suspected hydro-

cephalus (30) led to major reductions in RNbrain scans and PEGs. CT scanning also re-duced the total charges for neurodiagnostic pro-cedures. Speed of diagnosis, length of hospitali-zation, and kind of therapy did not change.Thus, CT scanning was found to be particularlyappropriate for such patients.

All three Larson studies depended on the useof procedure charges as a proxy for the econom-ic costs involved in performing the neurodiag-nostic procedures. The resulting distortion ofestimated cost impacts is instructive. In thestudy of suspected hydrocephalus (30), totalestimated diagnostic charges dropped precipi-tously with the advent of CT scanning, largelybecause of the high charge for a PEG examina-tion ($492 at the study hospital). The highcharge at least partly reflects that procedure’srelative rarity and the existence of excess capaci-ty. The long-run incremental cost of a PEG ex-amination is probably much lower. Were suchexaminations not painful and risky, andtherefore to be avoided wherever possible, asubstantial change in the estimated cost of thisprocedure could have reversed the findingsregarding the benefit of CT scanning in patientswith suspected hydrocephalus. In fact, a PEGprocedure charge of $2OO would have resulted inno difference between pre-CT and post-CT totalper-capita charge for neurodiagnostic proce-dures, and the analysis would have rested on theimplications of substituting CT for PEG exam-inations for patient morbidity and risk.

The investigation did not question the appro-priateness of inpatient versus outpatient work-ups. Some neurodiagnostic studies can probablybe performed on an outpatient basis. By accept-ing the existing settings of care as the basis forcomparison, the studies may have overlookedparticularly efficient alternatives.

The three Larson studies do show, however,that with good hospital data systems and effortand persistence on the part of the investigators,it is possible to identify patients with reasonablyhomogeneous presenting signs and symptoms.The ability to obtain medical records catego-rized by admission problem rather than dis-charge diagnosis is critical for estimating theoutcomes of different diagnostic pathways, and

although the Larson studies do not specify path-ways with sufficient clarity, their patient iden-tification methodology would support suchanalyses.

A more recent study by the same authors in-vestigated the impact of outpatient CT scans onheadache patients (28). Three cohorts of pa-tients were selected for study: one cohort con-taining patients seen before installation of a CTscanner, a second cohort containing patientsseen shortly after installation, and a thirdcohort containing patients seen 1 year after in-stallation of the CT scanner, The patients ineach of these cohorts were selected from a re-view of electroencephalogram (EEG) records,where headache was the reason for the EEG.Thus, the patient sample excluded those whomay have had CT scans without EEGs. The costof diagnostic tests was computed using proce-dure charges; the diagnostic testing charges perpatient increased 16 percent after the introduc-tion of CT scanning. The neuroradiologic diag-nostic yield of patients with normal neurologi-cal workups was low, and, in the case of CT,was negligible. This finding led the authors tosuggest that abnormal findings on neurologicalexamination should be a screening criterion forperforming CT scans and other neuroradiologi-cal procedures for patients with headaches.

The fact that diagnostic charges increasedafter the introduction of CT scanning says verylittle about the appropriate allocation of re-sources; the more powerful evidence presentedby the authors is the apparent absence of realclinical benefit, even with respect to diagnosticcapability, deriving from CT scanning—and,indeed, from most neurodiagnostic procedures.If radiological and other technology-bound pro-cedures are to be performed on patients with nofocal findings, then this study has not adequate-ly demonstrated that CT scanning cannot lowerthe cost of diagnosis if it is appropriately sub-stituted for other tests. A clear cost-reducing op-tion, however, appears to ensue from stoppingthe diagnostic workup of the patient when theclinical examination reveals no abnormalities.In light of the importance of such a conclusion,it is unfortunate that the sample of patients inthis study was chosen by excluding those who

may have had CT scans without EEGs. Moreneeds to be done to refine the study of diagnos-tic procedures for patients with headaches.

Ambrose and his colleagues (2) in Britainstudied the impact of CT scanning on the diag-nosis of head injuries, but they did not attemptto attach cost estimates to the observed changesin utilization of diagnostic procedures and ex-ploratory surgery. The introduction of CT scan-ning at the study hospital dramatically reducedthe use of arteriography, PEG, and exploratorycraniotomy without changing the mortalityrate; however, once CT examinations wereavailable, they were ordered so frequently thatnet costs may have increased. Nevertheless, therealized effect is not the relevant question for re-source allocation. CT scanning can undoubted-ly reduce the cost of diagnosis for some patientsby substantially reducing the high cost, morbid-ity, and mortality associated with the diagnosticprocedures it replaces. The real questionthough, is how to achieve the best results. Likeprevious studies, the Ambrose study failed toexplore how CT scanning might optimally beintegrated into diagnostic and management pro-tocols.

In a recent study of CT scanning in acute headtrauma at a U.S. teaching hospital, Zimmer-man, et al. (47) noted a progressive decrease inutilization of skull X-rays following the intro-duction of CT scanning, but they also observedthat “a certain proportion of the present skullradiographic examinations are done because ofthe referring physician’s adherence to tradi-tional ways of evaluating trauma patients. ” Acritical question that economic analysis mightaddress is exactly when, if at all, skull X-raysshould be employed in the evaluation of patientswith head trauma.

Bahr and Hodges (3) investigated the impactof CT scanning on the total cost of hospital care(measured by billed charges) for neurologicaland neurosurgical inpatients at a universityhospital. The total average hospital bill for thesepatients decreased about 10 percent, after ad-justing for inflation, but the decrease was notstatistically significant. The study populationwas disaggregated into three diagnostic groups:

1) patients with extracerebral collections, 2) pa-tients with proved intracranial tumors, and 3)patients with vascular disease. The total infla-tion-adjusted hospital bill for patients in the firsttwo categories decreased by 23 and 35 percentrespectively, while that of the third group in-creased by 30 percent.

It is difficult to interpret such findings becauseof the possibility that the patient mix changedover the period, especially if the availability ofCT scanning enabled testing of some as outpa-tients. In addition, the study population is am-biguous, consisting of patients with all types ofneurological complaints. The disaggregated costcomparisons are based on final diagnosis, noton the patients’ presenting signs and symptoms.Consequently, the cost or savings by diagnosisare irrelevant to the use of CT scanning underspecified presenting conditions. The difficulty ofinterpreting cost impacts based on procedurecharges has already been noted.

In one of the earliest economic impact studies,Wortzman, Holgate, and Morgan (45) reviewedthe records of over 200 inpatients who weregiven CT head scans at a Canadian hospital. Ineach case, the authors judged whether an angio-gram or PEG had been prevented or promptedby the CT scan. The hospital costs (measured bycharges) saved or induced by use of the CT scanwere calculated on the assumption that the hos-pital stay for an angiogram is 1 day and that fora PEG examination, 3 days. Net savings result-ing from the availability of CT scanning wereestimated at $300 per patient. In a second partof the study, the authors reviewed all outpatientCT scans at the hospital and, with the help of asurvey of referring physicians, estimated that 58percent of the outpatients would have beenhospitalized had the CT scan been unavailable.Assuming that these patients would have stayedthe average length of stay in neurology (14days), the authors estimated that CT scanningsaved $1,490 per outpatient scan.

The vulnerability of the savings estimate tochanges in assumptions is obvious and needs nofurther comment. Nor does the use of charges toreflect cost. The method for determining thedegree of substitution of CT for other pro-cedures is judgmental and therefore subject to

investigator and respondent bias. Most impor-tant, however, the aggregated nature of thepatient group under study mitigates the value ofthe findings. It is certain that savings, howevermeasured, could be further increased by moredetailed analysis of presenting signs andsymptoms.

In a recent reappraisal of their early study,Wortzman and Holgate (45) revealed discrepan-cies between their estimates of potential savingsand realized effects in subsequent years. Hos-pital stays and admissions did not decline aspredicted in the study. Although the number ofangiograms and PEGs did decline, the cost ofthese procedures, as measured by charges, rose,thus wiping out much of the estimated savings.The procedure charge increases were necessi-tated by decreases in utilization without con-comitant capacity reduction. But the authorswrongly assumed that estimated cost savingsshould be altered to reflect these changes. Theprinciple of long-run incremental cost requiresthe assumption that capacity will ultimately ad-just to reduced demand.

A study by Thomson (41) at Frenchay Hospi-tal in Great Britain traces changes in the use ofparticular procedures and services in the 2 yearsfollowing the installation of a CT scanner.There, as in most other institutions, the numberof arteriograms, RN brain scans, and PEGs de-clined after the introduction of CT scanning; thelength of stay of neurosurgical inpatients de-clined as well, and admissions were avoided.The cost of operating a CT scanner at the studyhospital was meticulously calculated using amodified concept of capacity replacement costand assuming operation at full capacity. Theeconomic cost of other neurodiagnostic pro-cedures made unnecessary by CT scanning werenot measured so carefully, but were assignedrelatively low values in order to bias the studyagainst CT scanning. Hospital stay savings wereestimated at the average daily cost of a NationalHealth Service hospital bed. CT generated netsavings for the hospital, and, by inference, forthe National Health Service, although the sav-ings were minimal,

Since Thompson made no attempt to follow aspecific patient group but merely reported on in-

stitutionwide changes in utilization of services,it is impossible to know the kinds of patients forwhom CT scanning is cost effective. Other au-thors have documented such institutionwidechanges (26). This kind of information, evenwhen linked to good cost estimates, ignores thecentral problem for medical decisionmaking andhealth facilities planning: To what uses shouldCT scanning capability be applied?

Two studies have attempted to estimate theimpact of CT scanning on the national cost ofdiagnosis in the United States. Willems, et al.(44) estimated net changes in total U.S. expendi-tures for diagnosis due to the introduction of CTscanning. The performance of a CT scan gen-erated a technical charge, professional fee, andassociated hospital room and board charges, allof which totaled an amount between $295 mil-lion and $428 million in 1976. Savings resultingfrom reduction in PEGs, RN brain scans, andarteriograms were deducted from this total,leaving a net estimate of $181 million to $ 3 9 0million in increased expenditures due to CTscanning in 1976.

Like analysts in most of the other studiespreviously described, WilIems used procedurecharges to estimate the cost of most procedures.Although charges are an inadequate surrogatefor resource costs, they do reflect real transfersfrom payers to providers of health care; in thatsense, the total expenditure burden imposed byCT scanning is of interest. It is unknown, how-ever, whether the expenditures were used tosubsidize the provision of other services thathospitals provide at a loss or, perhaps, whetherthe expenditures were used to inflate physicianor other professional incomes.

A study by researchers at Arthur D. Little,Inc. (17) represents the only effort to date toestimate the national impact of both head andbody scanning on the cost of diagnosis. The na-tional impact on the utilization of competingdiagnostic procedures and exploratory surgeryin 1977 and 1980 was estimated by a variety ofsources, including consultations with experts inthe field. The technical cost of providing eachtype of procedure was estimated assumingequipment at full utilization and current costs ofreplacement of equipment. (Except for explor-

atory surgery, charges were not used to estimateprocedure costs. ) To arrive at a unit cost foreach of the various diagnostic procedures, thenational average hospitalcharge was applied to thehospitalization required bycedure.

It was predicted in the A.

room and boardadditional days ofeach type of pro-

D. Little study thatthe availability of CT head scanning would raisediagnostic costs by about $29 million in 1977and $31 million in 1980. These low estimates ofnet impact contrast with the results of theWillems study but can be explained by differentassumptions about the effect of CT scanning onexploratory surgery and skull X-rays.

The predicted impact of CT body scanning issubject to greater speculation since, at the t imeof the study, body scanning had only begun tofind its way into the diagnostic process. Theauthors predicted that in 1980, as a result of theuse of CT scanning, the estimated cost of diag-nosing abdominal and mediastinal disorderswill have increased by $152 million. This esti-mate is based on the assumption that the availa-bility of CT body scanning will reduce the num-ber of exploratory surgeries by 50 percent andthat no other diagnostic procedures will be af-fected by CT scanning.

Although the ultimate significance of theseestimates for decisionmaking is most certainlylow, the A. D. Little study represents the best ef-fort to date at estimating the resources costsassociated with performing diagnostic proce-dures. Capacity costs were based on currentequipment replacement prices and full capacity.However, indirect costs (administrative ex-penses in departments outside of the CT scan-ning unit) were arbitrarily assigned a value of 50percent of direct costs, a level that probablybears no relation to the actual incremental costsassociated with the existence of the CT scanningcapacity. The technical cost of surgery androutine costs of hospital care were based oncharges rather than costs. With the high cost ofhospital construction today, hospital room andboard charges reflecting original capital costsprobably understate replacement costs, In spiteof these shortcomings, the method for estimat-

ing procedure costs is a model for economicevaluation of CT scanning.

Studies of the Cost of Case Finding

A commonly advocated approach to econom-ic evaluation is the calculation of the cost of casefinding, which is variously defined as “the costper positive finding, “ “the cost per correct find-ing, ” or “the cost per true positive” (6). The costof case finding is a concept borrowed from theeconomic evaluation of screening procedures. Itis particularly useful in determining which oftwo or more mutually exclusive screening pro-cedures is more cost effective in detecting a par-ticular disease in an asymptomatic population.A procedure (or sequence of procedures) thatcosts the least per true positive finding is pre-ferred to all others.

Unfortunately, the concept does not travelwell. When applied to analysis of CT scanning,which is used on symptomatic patients in con-junction with other tests, the cost of case findingis meaningless. The results of studies using theconcept illustrate its limitations.

Carrera, et al. (9) calculated the cost per ab-normal CT scan for three types of cases:1) headaches without neurological findings,2) headaches with neurological findings, and3) suspected temporal lobe epilepsy. Because ofdifferences in the diagnostic yield of CT scan-ning among these three types of cases, the costper abnormal CT scan examination was $4,000;$1,300; and $2,000 respectively. 7 In a separatestudy, Knaus and Davis (23) calculated the case-finding cost of CT scanning for 25 leading indi-cations for CT head scanning. The case-findingcost at the study hospital ranged from $411 forpatients in coma to $3,500 for patients withheadache. Larson, et al. (28) estimated that thecost of case finding for headache patientswithout abnormal neurological findings was$11,901. The authors of all three papers ques-tioned the cost effectiveness of CT scanning inpatients with headaches without neurological

—7 A CT procedure charge of $240 was used as a surrogate for

cost .

findings. Any conclusion about cost effective-ness is premature, however, for the cost of casefinding can only be compared to the benefits offinding abnormalities in the patient group.

Evens and Jest (12) have used the concept ofcase-finding cost to compare CT head scanningand RN brain scanning on patients with neuro-logical complaints. Using estimates of sensiviti-ty and specificity of CT and RN scanning, theauthors calculated the cost per correct finding:CT scanning cost $141 per correct finding; RNscanning cost $73 per correct finding. Although,on the surface, this finding would imply that RNscanning is the procedure of choice, Evens andJest observed that CT is a more sensitive test.Hence, in a large number of cases, the combina-tion of the two tests would be ordered, with ajoint cost per correct finding of $214. If an initialnegative RN scan generates a confirmatory CTexamination in at least 60 percent of neurolog-ical cases, then an initial CT scan costs less percorrect finding than an initial RN examination.

Notice that Evens and Jest used a variant ofcase-finding cost. Whereas Carrera (9) andKnaus and Davis (23) calculated the cost per ab-normal or positive finding, Evens and Jest cal-culated the cost per correct finding. A third ap-proach, the cost per true positive finding, hasalso been used. g The relative merit of these alter-native measures of case-finding cost depends onthe errors in diagnosis and implications for ther-apy of each alternative. In fact, the most ap-propriate measure can be selected properly onlyif the costs and benefits of pursuing diagnosticpathways to their ultimate outcomes areknown.

This point can be illustrated algebraically. LetTP, FP, TN, and FN denote the proportion oftrue positives, false positives, true negatives,and false negatives expected from a given diag-nostic pathway in a specified population of pa-tients. Let C be the total cost of performing

‘See, e.g., B. J. McNeil and S J. Adelstein, “Measures of Clin-ical Efficacy. The Value of Case Finding in Hypertensive Reno-vascular Disease, ” 1975 (32).

diagnostic procedures on these patients. Thethree case-finding cost options are:

Cost per true positive = cTP (100)

Cost per abnormal finding = c

(Tp + Fp) (1OO)

Cost per correct finding = c(TP + TN) (100)

Suppose that two diagnostic procedures, A andB, have the disease detection rates shown infigure 2 when performed on a patient groupwith a 0.50 disease prevalence rate. Assume also

Figure 2.—Disease Detection Rates for Two Tests

+

T e s t

r e s u l t s

—

D i a g n o s t i c

test

A

B

that both procedures cost the same to admin-ister, $10 per patient. Figure 3 gives the case-finding costs of each procedure for each 100 pa-tients as measured by the three case-finding op-tions. In figure 3, diagnostic procedure B wouldbe preferred if the option “cost per correct find-ing” were used, and diagnostic procedure Awould be preferred under the options “cost pertrue positive” and “cost per abnormal finding. ”

Suppose, however, that the implications of afalse negative are major (e.g., a fatal but com-pletely treatable disease goes undetected), whilethose of a false positive are minor (e. g., inex-pensive additional tests are required). Clearly,diagnostic procedure A would be preferred, anduse of the cost per correct finding would be un-acceptable. Similarly, if the implications of afalse positive are major (e.g., application of in-appropriate, expensive, and risky therapy),while those of a false negative are also great,then the use of cost per correct finding is su-perior. Thus, to accurately determine whichmeasure of case-finding cost is preferred, it isnecessary to know something about the subse-quent benefits and costs resulting from a par-ticular diagnostic procedure. The more that isknown, the more accurate can be the weightsput on the different kinds of errors. Thus, case-finding costs do not bypass the need for in-formation concerning the full implications ofembarking on a diagnostic pathway.

Figure 3.—Case-Finding Costs for 100 Patients

Cost per C o s t p e r C o s t p e rt r u e p o s i t i v e correct f ind ing abnormal finding(C/TP) (C/TP + T N ) ( C / T P + F ? )

least 500 angiograms and 200 PEGs could beeliminated each year, the CT scanner wouldsave money.

The cost analysis rightly assumed that allradiographic equipment is operated at capacityboth before and after the introduction of CT;this avoids ascribing to CT scanning savingsthat would otherwise be obtainable from con-solidating radiographic services in larger hospi-tals, In fact, the whole question of CT scannerlocation appears to be inextricably linked to theoptimal distribution of neuroradiologic capacityas a whole, Costs could be reduced by regional-izing such capacity, but these savings wouldhave to be compared to the added cost of trans-portation to regional facilities. For emergencypatients, the cost savings would also have to beweighed against reduced access to care. Thus,the Swedish study, though excellent in its cost-ing methodology, is still limited as a planningapproach, since it analyzes the effect of intro-ducing CT scanning to an already existing sys-tem in which the present allocation of radiologi-cal resources may very well be inappropriate.

Bartlett, et al. (7) reported on a comparison ofalternative options for the location of CT scan-ning capacity in a particular region of England.The study assumed that the demand for CTscans would be 3.5 per thousand residents peryear. Alternative configurations for meeting(and exceeding) that demand were then com-pared. The critical question was whether CTscanning capacity should be installed in allhospitals with neurosurgery, neurology, and ac-cident centers. The not so startling conclusionwas that more scanners would cost more moneyand the added cost of installing a CT scanner inevery accident center (far exceeding the requiredscanner-to-population ratio) must be comparedto the lifesaving potential of the scanner in headtrauma. We are thus led full circle back to thebasic question of the conditions for which CTscanning is economically justified.

An American study by Greenwald, et al. (18)is an application of a mathematical optimizationtechnique to the question of scanner capacityand location. The authors formulated the prob-lem thus: How does one choose the number andlocation of CT scanners such that the sum of CT

operating and patient travel costs is minimized?The solution must meet two constraints: 1) thatthe region’s demand for CT scans is met, and 2)that each scanner is assigned a patient load notexceeding its capacity. Other considerations,such as the desirability of placing CT scanningcapability in centers of radiological excellence,were not included in the model, but easily couldbe if the potential locations of CT scanners werelimited to those sites meeting certain criteria.

The authors showed that when patient travelcosts are taken into account, more scanners canactually reduce the total cost of meeting a givenlevel of demand. As demand for CT services in-creases or decreases, however, the optimal num-ber of facilities also changes. Hence, the modelbegs the question of the optimal number of scan-ners by assuming that the answer to the priorquestion—under what conditions is CT scan-ning worth its costs — has a 1 read y been an-swered.

CONCLUSIONS

Taken together, the literature on the eco-nomics of CT scanning reveals a plethora ofstudies that either ask the wrong questions, or,in asking the right ones, must make heroic andunsupportable assumptions. Few studies haveaddressed the fundamental question of econom-ic evalution: For what kinds of patients is CTscanning worth its costs?

The energy devoted to documenting historicalcost savings resulting from CT scanning is par-ticularly misplaced. This literature offers no realinsight into the question facing health care deci-sionmakers today. Studies of the cost of casefinding, which appear on the surface to be moreuseful in guiding resource allocation decisions,offer such a partial view of costs and effects ofany diagnostic pathway that their results aremore obfuscating than enlightening. Finally,those methods designed to assist directly in thebig resource allocation decision—the placementof CT scanning capacity—must ignore the un-answered question of conditions of cost-effec-tive use.

The mathematical programing approach tofacility location has a rich history,’” especiallywith health care facilities. It is worth noting thatif CT scanning is considered a necessary diag-nostic tool for head trauma and other emergen-cy situations, then the formulation suggested byGreenwald, et al. (18) may not be appropriate.Instead, a preferred formulation might be tominimize the number of scanners required to in-sure that all points in the region lie within somemaximum time or distance criterion.11 The re-sulting configuration of CT scanning facilitieswould very likely be quite different from thatobtained in the Greenwald, et al, study.

Although the methodological and conceptualobstacles to conducting useful economic evalua-tions are great, the literature shows that somecan be overcome. Unfortunately, the obstacleshave been dealt with in pieces—each study hasoffered a solution to a particular problem, butno study is completely satisfactory in all areas.In particular, the proper estimation of resourcecosts is an area that appears to be a good candi-date for major improvements with relatively lit-tIe effort, although some conceptual problemswill continue to exist. Certainly the use ofcharges as a surrogate for cost can be elimin-ated. The identification of patients by specificsigns and symptoms also appears to be feasible,although the cost of research based on specifiedpatient groups will depend on the existence ofgood hospital data collection systems. Theproblem of identifying patients by signs andsymptoms will be intensified with the use of CTbody scanning because of the ambiguity of signsand symptoms related to diseases of the ab-domen.

When the data needed to perform valid eco-nomic evaluations are considered, it is not sur-prising that so few appropriately constructedstudies exist. The cost of analyzing the signs andsymptoms of the many different kinds of pa-tients who are reasonable candidates for CTscanning would be great in itself; to imagineconducting such analysis for all types of patientsand diagnostic procedures is unrealistic in theextreme. At best, a few controversial indica-tions for expensive diagnostic procedures mightbe the subjects of economic evaluation. Re-search to date on CT scanning suggests that pa-tients with headaches and suspected cerebrovas-cular disease would be excellent candidates foradditional analysis with more appropriatemethods of economic evaluation.

If one cannot expect a comprehensive analysisof the appropriate conditions of use of a diag-

nostic procedure, how can those responsible fordecisions about the level and location of diag-nostic capacity make rational choices? The taskis impossible, as most health planning agencypersonnel have suspected all along. Instead ofattempting to find analytic approaches to sup-port what is essentially a political decision,health planning agencies might do better to setfairly arbitrary regional capacity ceilings anduse economic evaluation to explore the cost andaccess implications of alternative locationalchoices.

Even better, the incentives might be improvedfor those who are most knowledgeable aboutthe diagnostic alternatives: the providers them-selves. If we had greater confidence that their in-centives were consistent with the goals of publicpolicy, the limitations of formal evaluationwould be far less troublesome.

1.

2.

3.

4.

5.

6.

7.

8.

9.

Abrams, H. L., and McNeil, B, J., “Medical Im-plications of Computed Tomography (’CATScanning’ ),” N. Eng. J, Med. 298:255, 1978.Ambrose, J., et al., “EMI Scan in the Manage-ment of Head In juries,” Lancet 1:847, 1976.Bahr, A. L., and Hodges, F. J., “Efficacy ofComputed Tomography of the Head in Chang-ing Patient Care and Health Costs: A Retrospec-tive Study, ” Am, ]. Roent. 131:45, 1978.Baker, C., and Way, L. W., “Clinical Utility ofCAT Body Scans, ” Am. ). Surg. 136:37, July1978.Banta, D., “Number and Distribution of CTScanners, 1978, ” report to the Bureau of HealthPlanning and to State and local health planningagencies (Hyattsville, Md.: National Center forHealth Services Research, August 1978).Banta, D., and McNeil, B. J., “The Costs ofMedical Diagnosis: The Case of the CT Scan-ner, ” Health Care Mgt. Rev, 3:7, 1978.Bartlett, J. R., et al,, “Evaluating Cost-Effective-ness of Diagnostic Equipment: The Brain Scan-ner Case, ” Br. Med. ]. 2:815, 1978.Bunker, J. P., et al, (eds. ), Costs, Risks, a n dBenefits of Surgery (New York: Oxford Univer-sity Press, 1977).Carrera, G. F., et al., “Computerized Tomogra-phy of the Brain in Patients With Headache and

10.

11.

12.

13.

14.

15.

16.

17.

With Temporal LobeCost Effectiveness, ” /,1978.

Epilepsy: Findings andComp. Tomog, 2:108,

Creditor, M. C., and Garrett, J. B., “The In-formation Base for Diffusion of Technology:Computed Tomography Scanning, ” N. Eng. J,Med. 297:49, 1977.Dales, L. G., et al., “Evaluation of a PeriodicMultiphasic Health Checkup, ” Meth. Inform.Med. 13:140, 1974.Evens, R, G., and Jest, R. G., “The Clinical Effi-cacy and Cost Analysis of Cranial ComputedTomography and the Radionuclide Brain Scan, ”Sem. Nuc. Med. 7:129, 1977,

“Economic Analysis of Body ComputedTomography Units—Including Data on Utiliza-tion, ” Radio, 127:151, 1978.

“Economic Analysis of Computed To-mography Units, “ Am. ], Roent. 127:191, 1976.Evens, R. G,, et al., “The Utilization, Reliabili-ty, and Cost Effectiveness of Computer CranialTomography in Evaluating Pseudotumor Cere-bri, ” Am, J. Radio. 129:263, 1977.Freeborn, D. K,, et al., “Determinants of Medi-cal Care Utilization: Physicians’ Use of Labora-tory Services, ” Am. J. Public Health 62:846,1972.Gempel, P. A,, and Harris, G. H. “Comparative

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Cost Analysis: Computed Tomography vs. Al-ternative Diagnostic Procedures, 1977-1980”(Cambridge, Mass.: Arthur D. Little, Inc.,1977).Greenwald, H. P., et al., “Transportation or CTScanners: A Theory and Method of Health Re-sources Allocation, ” Health Serv. Res. 14:207,1979.Hsiao, W. C., and Stason, W, B,, “Toward De-veloping a Relative Value Scale for Medical andSurgical Services, ” Health Care Fin. Rev. 1;23,1979.Institute of Medicine, Computed TomographicScanning: A Policy Statement (Washington,D. C.: National Academy of Sciences, 1977).Jonsson, E., and Marke, L. A., “CAT Scanners:The Swedish Experience, ” Health Care Mgt.Rev, 2;37, 1977.Kassirir, J. P., and Pauker, S. G., “Should Diag-nostic Testing Be Regulated?” N. Eng. ]. Med.299:947, 1978.Knaus, W. A., and Davis, D. O., “Utilizationand Cost-Effectiveness of Cranial Computed To-mography at a University Hospital, ” ]. Comp.Assist. Tomog. 2:209, 1978.Knaus, W. A., and Wagner, D., “CT ScanningPatients With Headache: A Cost-Benefit Calcu-lation, ” Am. ]. Neuro. (in press).Knaus, W. A., et al., “Impact of New Technol-ogy: The CT Scanner, ” Med. Care 15:533, July1977.Larson, E. B., and Omenn, G., “The Impact ofComputed Tomography on the Care of PatientsWith Suspected Brain Tumor, ” Med. Care15:543, 1977.

“Computed Tomography in PatientsWith C’erebrovascular Disease: Impact of a NewTechnology on Patient Care, ” Anl. ]. Roent.131:35, 1978.

“Diagnostic Evaluation of Headache:Impact of Computerized Tomography and Cost-Effectiveness, “ ]. A.M.A. 243:359, 1980.

“Effect of Computed Tomography onUtilization of Cerebral Angiography, ” Am. ].Roent, 129:1, 1977.

“Impact of Computed Tomography onthe Care of Patients With Suspected Hydroceph-alus, ” Am. ]. Roent. 131 :41, 1978.McNeil, B. J., and Adelstein, S. J., “Determiningthe Value of Diagnostic and Screening Tests, ” ].Nucl. Med. 17:439, 1976.

, “Measures of Clinical Efficacy: The Val-ue of Case Finding in Hypertensive Renovascu-lar Disease, ” N. Eng. ]. Med. 293:221, 1975.McNeil, B. J., et al., “Measures of Clinical Ef -

34.

35.

36.

37.

38.

39,

40.

41.

42.

43.

44.

ficacy: Cost-Effectiveness Calculations in theDiagnosis and Treatment of Hypertensive Reno-vascular Disease, ” N. Eng. ]. Med. 293:216,1975.Neuhauser, D., and Lewicki, A. J., “What DoWe Gain From the Sixth Stool Guaiac?” N. Eng.j. Med. 293:266, 1975.Office of Technology Assessment, U.S. Con-gress, Policy Implications of the Computed To-Mography (CT) Scanner, GPO stock No.052-003 -00565-4 (Washington, D. C.: U.S. Gov-ernment Printing Office, 1978).

The Implications of Cost-Effectiveness) lessAnalysis of Medical Technolology, GPO stockNo. 052-003 -00765-7 (Washington, D. C.: U.S.Government Printing Office, August 1980).Orkand Corp., Planning Approach, Criteria,and Standards for Specialized Trauma Services:Optimal, Intermediate, and Minimal Levels ofCare (Silver Spring, Md.: Orkand Corp., Sept.30, 1976). (Also available as publication No.HRP-0015437 from Department of Commerce,National Technical Information Service, Spring-field, Va. )Pete, R., et al., “Design and Analysis of Ran-domized Clinical Trials Requiring ProlongedObservation of Each Patient, ” Br. ]. Cancer34:585, 1976.ReVelle, C., et al., “Facility Location: A Reviewof Context-Free and EMS Model s,” Health Serv.Res. 12:129, 1977.Schoenbaum, S. C., et al., “Benefit Cost Analy-sis of Rubella Vaccination Policy, ” N. Eng. ].Med. 24:306, 1976.

Thomson, J. L. G., “Cost-Effectiveness of anEMI Brain Scanner: A Review of a Two-Year Ex-perience, ” Health Trends 9:16, 1977.Weinstein, M. C., “Economic Evaluation ofMedical Procedures and Technologies, ” in Medi-cal Technology, J. Wagner (ed. ), Urban InstituteConference, NCHSR Research Proceedings Se-ries, DHEW publication No. (PHS) 79-3254(Washington, D. C.: National Center for HealthServices Research, Office of Health Research,Statistics, and Technology, PHS, DHEW, 1979).Weinstein, M. C., and Stason, W. B., “Founda-tions of Cost-Effectiveness Analysis for Healthand Medical Practices, ” N. Eng. ]. Med.296:716, 1977.Willems, J., et al., “The Computed Tomography(CT) Scanner, ” in Medical Technology: TheCulprit Behind Health Care Costs? proceedingsof the 1977 Sun Valley Forum on NationalHealth, S. H. Altman and R. Blendon (eds. ),

DHEW publication No. (PHS) 79-3216 (Wash- mography: An Evaluation of Cost-Effective-ington, D. C.: DHEW, 1979). ness, ” Radio. 117:75, 1975.

45. Wortzman, G., and Holgate, R. C., “Reapprais- 47. Zimmerman, R. A., et al., “Cranial Computedal of the Cost Effectiveness of Computed To- Tomography in Diagnosis and Management ofmography in a Government-Sponsored Health Acute Head Trauma, ” Am. ], Roernt. 131:27,Care System, “ Radio. 130:257, 1979. 1978.

46. Wortzman, G,, et al., “Cranial Computed To-

u. $. OOVERl'MENT PRINTING OFFICl; liel