cost effectiveness of automated multichannel chemistry

Case Study #4: Cost Effectiveness of Automated Multichannel Chemistry Analyzers ● 5

tiveness criterion. Thus, paradoxically, thecost effectiveness of the multichannel ana-lyzer may rest on whether we, as a society,are willing to pay enough for diminishingincremental health benefits to justify suffi-cient volumes of testing to permit us to af-ford the reduction in unit testing costs thatmultichannel analyzers offer.

Cardiac Enzymes

15.

16.

Three cardiac enzymes—creatine phospho-kinase (CPK), lactic dehydrogenase (LDH),and aspartate aminotransferase, also calledserum glutamic oxaloacetic transferase(SGOT)—and their isoenzymes were se-lected as examples of chemistry tests thatare frequently ordered on multichannelequipment. CPK, LDH, and SGOT testsare available on all major multichannelanalyzers; isoenzymes of CPK and LDH arebecoming available on the equipment of atleast one manufacturer. Their major use isto diagnose (“rule in/rule out”) myocardialinfarction (’[heart attack”). In the emergen-cy room, they are used to decide whether toadmit a patient to the hospital; in the cor-onary care unit, they are used to monitor apatient’s condition and to decide when todischarge the patient to the general ward.

The diagnostic value of cardiac enzymes isdifficult to evaluate, because the levels ofthese enzymes are themselves used to definethe diagnosis of myocardial infarction. Nopublished study has examined the in-cremental diagnostic information value ofany single enzyme test, given the othertwo; of two, given the other one; of the en-zymes, given the isoenzymes; or of the iso-enzymes, given the enzymes. Such eval-uations could lead to cost savings if some ofthe tests were found to provide information

INTRODUCTION. . . In recent years, the economic and tech-

nologic development of laboratory operationsseem to have had a greater influence than thephysician on trends in laboratory utilization(17).

17.

18.

largely supplied by other tests in themanagement of myocardial infarction.

The clinical efficacy of tests measuringcardiac enzymes and isoenzymes dependsultimately on the risks associated with fail-ing to hospitalize a patient with myocardialinfarction, or on the incremental risk asso-ciated with complications suffered outsidea coronary care unit. Even if these tests im-proved the accuracy of diagnosis, or evenprognosis, the value of this clinical in-formation would be nil unless treatmentmade a difference in health outcome. Theefficacy of coronary care units and of hos-pitalization of patients with myocardial in-farction is a matter of much controversyand is not likely to be resolved soon. There-fore, the health benefits to be derived froma more sensitive or specific diagnosis ofmyocardial infarction remain uncertain.

As long as coronary care and hospitaliza-tion are standard practice, the greatest val-ue of cardiac enzyme and isoenzyme meas-urement may be induced economic savings.To the degree that these tests improve thesensitivity of the diagnosis, physicians maybe more willing to rule out myocardial in-farction on the basis of negative findings.To the degree that they improve specificityof diagnosis, they are less likely to lead tohospital admission of patients withoutmyocardial infarction. An economic eval-uation of the impact of enzyme and iso-enzyme measurements is feasible and maybe sufficient to demonstrate their cost effec-tiveness given actual practice, although it isdifficult to assess the cost effectiveness ofsuch measurements in ideal practice be-cause of the uncertainty surrounding thebenefits of intervention.

The Problem

Since 1963, when Technicon Corp. began tomarket the first automated chemistry analyzerthat could perform more than one test on a

6 ● Background Paper #i?: Case Studies of Medical Technologies

single sample of serum, this technology has hada profound economic impact on the U.S. healthcare system. The question still being debatedamong industry officials, health services re-searchers, hospital administrators, laboratorydirectors, and clinical chemists and pathologistsis whether this technology has been, on balance,an economic dream or nightmare.

Evidence is strong on both sides of the ques-tion. On the one hand, largely as a result ofautomation and especially as a result of multi-ple-channel technology, the cost per chemistrydetermination has fallen dramatically during thepast decade. This cost reduction has resulted, inlarge part, from increased productivity of lab-oratory personnel as measured by the numberof determinations per labor-hour. One Britishstudy showed a tripling of laboratory outputper technician-hour between 1955 and 1966 in ahospital that had begun to use a 12-channelanalyzer during that interval (55). A U.S. studyshowed a doubling of laboratory output perlabor-hour from 1961 to 1970, during whichtime multichannel analyzers were purchased bythree of the five hospitals in the study (30).

The importance of such savings in labor costsis underscored by studies that indicate thatdirect labor costs constitute 50 to 65 percent ofthe total direct costs of hospital-based, clinicalchemistry laboratories (30,35,53,55). Compari-sons of different studies indicate that the aver-age direct cost of a chemistry determination wasapproximately $1.25 in 1969 (in hospitals thathad not adopted a 6- or 12-channel analyzer)(30), $0.70 in 1971 (35), and $0.30 in 1974 (32).This trend has persisted despite a 35-percent in-crease in price levels from 1969 through 1974.On the basis of the continued decline in prices ofreagents and consumables per test (10,23,46)and the increased labor productivity made pos-sible by the availability of 20- and 30-channelanalyzers, one can conclude that the real costper test has continued to decline.

The problem is that, despite the dramatic re-ductions in unit costs, total expenditures onclinical chemistry tests have risen dramatically.Between 1971 and 1975, the number of chemis-try determinations performed in the UnitedStates increased by 67 percent—from 2.9 billion

to 5 billion. During that period, charges (nottrue costs) for these tests increased from $5.6billion to $15 billion (34). These figures do notreflect the concomitant increase in costs of tests(other than chemistry) induced by unexpectedresults and required to rule out or confirm adiagnosis.

Underlying this alarming trend in laboratorycosts is the increased frequency with which phy-sicians order multiple-test profiles to be per-formed on automated equipment, rather thanjust individual tests needed for the immediatediagnosis. If, at one extreme, all of the tests hadclinical value equal, on average, to the tests thatwould have been ordered without the availabili-ty of test panels, then the cost per unit of healthbenefit would have clearly decreased as a resultof multichannel technology. If, at the other ex-treme, most of the additional determinationshad no clinical value, then the cost per usefuldetermination—hence, the cost per unit ofhealth benefit—would have increased as a resultof multichannel technology.

The truth, no doubt, lies between these ex-tremes. Therein lies the dilemma in evaluatingthe cost effectiveness of multichannel chemistryanalyzers. In essence, their cost effectivenessdepends on the way they are used in clinicalpractice. It depends not only on their ability toreduce costs or improve efficacy of tests thatwould have been performed by other methodsin their absence, but also on the benefits of new,high-volume, uses of the chemistry lab (e.g.,hospital admission screening) induced by hav-ing more tests at low incremental, but highfixed, cost.

Scope of the Case Study

The remainder of this case study is in fourparts. In the first part, by way of background,we offer a brief review of the history of multi-channel clinical chemistry technology; we alsodescribe several analyzers that exemplify thetechnology as currently marketed in the UnitedStates. Next, we present an analytic frameworkfor evaluating the cost effectiveness of the multi-channel analyzer, pointing out in the process themethodological problems inherent in evaluatinga diagnostic technology that consists of many


elements, each of which may have many clinicaluses. Within this analytic framework, wesurvey the evidence that bears on the questionof cost effectiveness. Third, we review theavailable data concerning the costs of multi-channel chemistry analysis. Finally, we use theanalytic framework to examine the evidenceconcerning the cost effectiveness of using thecardiac enzymes and isoenzymes (for whichtests are becoming more commonly available onmultichannel analyzers) in the diagnosis ofmyocardial infarction. The review of the dataon cardiac enzymes has more general implica-tions for evaluation of not only the cost-effec-tive use of the individual tests that can be per-formed with multichannel analyzers, but alsothe cost effectiveness of developing multichan-nel capability to perform such tests.

This case study touches on only some of theaspects of technology assessment that might beundertaken with respect to multichannel chem-istry analyzers. Thus, for example, we pay onlypassing attention to the question of test preci-

sion and accuracy, one that has received con-siderable attention in the literature (5,28) and inthe regulatory activities of the Food and DrugAdministration. Moreover, we do not considerthe impact of multichannel chemistry analyzerson the organization of laboratory services orhealth care institutions. Rather, we concentrateon the issues of cost and cost effectiveness ofmultichannel chemistry analyzers.

For purposes, of this case study, we consideronly instruments designed to yield more thanone chemical determination on a single bloodsample. Thus, for example, the study excludesthe centrifugal fast analyzer, which (despite itsability to process from 15 to 39 samples at atime at very high speed) is technically a single-channel analyzer., Evaluation of the centrifugalfast analyzer technology as an alternative to themultichannel analyzer technology would be ofgreat value at this early stage of its diffusion,but is beyond the scope of the present study.

DESCRIPTION OF MULTICHANNEL CHEMISTRY TECHNOLOGY

Overview of the Industry

Development of the first continuous-flowanalyzer, begun by Leonard T. Skeggs in 1950,led to the introduction of Technicon’s Auto-Analyzer to the U.S. market in 1957. At aboutthe same time, the first discrete-sample analyz-ers, the Robot Chemist and the AutoChemist,were marketed in the United States by AmericanOptical and AGA, respectively (l). Continuous-flow and discrete-sample remain the two typesof multichannel chemistry analyzers availabletoday, although considerable sophistication andversatility have evolved.

The selling feature of a multichannel analyzeris its ability to per-form numerous tests simul-taneously on one blood sample. This featuredistinguishes multichannel analyzers fromdedicated machines, which can run only one testat a time, and from centrifugal fast analyzers,which currently perform only one test at a time,although at very high speeds.

Although there is a large and growing marketfor automated multichannel analyzers (sales in1978 were estimated at $170 million) (23), thefield is dominated by a small number of firms.The undisputed market leader is Technicon;probably one-third to one-half of all automatedmultichannel analyzers in use are Technicon in-struments. 1 Other firms that manufacture auto-mated multichannel analyzers include AbbottLaboratories, American Instrument, AmericanMonitor, Beckman Instruments, Chemetrics,Coulter Electronics, E. I. du Pont de Nemours,Inc. (Du Pont), Gilford, Hycel, MicromedicSystems, Ortho Diagnostics, Perkin-Elmer,Union Carbide, and Vickers.

Y., . ..!

I Figures on market shares are not available to the public. Marketresearch tirms pr{wide this kind of information for substantialtees, but the results are available only to their clients, Our esti-mates are based (m conversations with individuals familiar withthe industry, on Hycel’s 1978 report to the Security ExchangeC{}mmlssion (23), and on Schwartz’s discussion of data from theCenter trom Dwease Control (42).

. .

8 ● Background Paper #2: Case Studies of Medical Technologies

This study concentrates on the products ofthree firms: Technicon, Hycel, and Du Pont.These firms differ both in size and in the im-portance of automated chemistry analyzers intheir overall sales pictures. Although they donot exhaust the market, their products exhibitenough variety to give a good sense of the mar-ket. Basic product information for the instru-ments reviewed below is shown in table 1.

Continuous-Flow Analyzers

General Description

Details of the continuous-flow process are de-scribed clearly by Schwartz (42). The processused today is a modification developed by Leon-ard Skeggs in the 1950’s. In brief, one serumsample is split into numerous aliquots (parts)separated by air bubbles. The partitioned sam-ple passes through separate incubation and de-tecting modules. Samples and reagents aredrawn through the system by a pump. The flow-ing reactants are mixed as they pass throughglass tubes in the form of concentric helixes.

In addition to dividing the sample, the airbubbles regulate the flow and clean the tubesbetween sample portions. Reagents are added inthe required sequence, and everything flowsalong to the module where the appropriatemeasurements are taken at the endpoints of thereactions. The results are recorded on a stripchart recorder and may a I so be sent to a com-puter. Every chemical test is run every time a

sample passes through the system; hence, re-agents are consumed whether or not a test is re-quested or reported. It is possible, however, todeactivate one or more channels on any particu-lar day, thus reducing the number of tests run.

Technicon Corp.

Technicon is the only manufacturer of con-tinuous-flow analyzers. It purchased the origi-nal rights to the process from Leonard Skeggs in1954. Since the introduction of its AutoAna-lyzer in 1957, Technicon has held patents thatpreclude the use of the continuous-flow tech-nique by other manufacturers. That technique isused in all of Technicon’s multichannel ana-lyzers, which are all variants of the SequentialMultichannel Analyzer (SMA), introduced in1965.

Four of Technicon’s multichannel instru-ments, the SMA 6/60, SMA 12/60, SMA II, andSequential Multichannel Analyzer with Com-puter (SMAC), are described below.2 Purchaseprices and other specifications are given in table1. In all, 23 tests are available on any of Tech-nicon’s machines (see table 2).

Technicon’s primary business is the develop-ment, production, marketing, and servicing ofautomated analytical systems, including re-agents. Technicon is generally agreed to be the

‘Technicon’s Auto Analyzer 11 is not discussed in this report.That analyzer is available in two- and three-channel models, aswell as a basic single-channel model.

Table 1 .—Multichannel Analyzers Produced by Three Major Manufacturers

Number of Number of tests Number of samplesManufacturer Model channels available processed per hour List priceTechnicon Corp. . . . . . . . . . . . . . . . SMA 6/60 6 23 60 $ 52.000

Hycel, Inc. . . . . . . . . . . . . . . . . . . . .

E. 1. du Pont de Nemours & Co. . . .

SMA 12/60SMA IISMA IISMAC

Super-17SKS-60Hycel-M

ACA-IACA-IIACA-IIACA-III

12121820171730303060

625 b

2323232318152429383838

609090

1506060

12097a97a97a97a

99,500138,600173,250271,000

75,000120,000225,000

49,00069,00089,000

120,000

afq~mbgr of tjgtgrminations per hour.bRefers to software capabditles.

SOURCE: Personal communications and marketing materials from the manufacturers (10,24,46).


Table 2.—Chemical Determinations Available on Technicon, Hycel, and Du Pont Multichannel Analyzers

Technicon(continuous-

f low) Hycel (discrete-sample) Du Pont (discrete-sample)Test All models Super-17 SKS-60 Hycel-M ACA-I ACA-II ACA-IIIAcid phosphatase . . . . . . . . . . . . . . . . . . . x x xAlanine aminotransferase (SGPT) . . . . . . X x x x x xAlbumin . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xAlkaline phosphatase . . . . . . . . . . . . . . . . X x x x x x xAlpha-hydroxybutyrate dehydrogenase . x x x xAmmonia . . . . . . . . . . . . . . . . . . . . . . . . . . x xAmylase . . . . . . . . . . . . . . . . . . . . . . . . . . . x x xAspartate aminotransferase (SGOT) . . . . X x x x x x xBilirubin, total . . . . . . . . . . . . . . . . . . . . . . X x x x x x xBilirubin, direct . . . . . . . . . . . . . . . . . . . . . X x x x xBilirubin, neonatal. . . . . . . . . . . . . . . . . . . x xCalcium . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xCarbon dioxide . . . . . . . . . . . . . . . . . . . . . . X x x x xCerebrospinal fluid protein. . . . . . . . . . . . x x xChloride . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x xCholesterol . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xCreating phosphokinase(CPK) . . . . . . . . X x x x x xCreatine phosphokinase-MB (CPK-MB). . x xCreatinine . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xEthanol . . . . . . . . . . . . . . . . . . . . . . . . . . . . x x xGamma-glutamyl transferase . . . . . . . . . . X x x xGlucose . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xIron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x xLactic acid . . . . . . . . . . . . . . . . . . . . . . . . . x x xLactic dehydrogenase(LDH) . . . . . . . . . . X x x x x x xLactic dehydrogenase, liver . . . . . . . . . . . x x xLipase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x xMagnesium. . . . . . . . . . . . . . . . . . . . . . . . . x x xPhenobarbital . . . . . . . . . . . . . . . . . . . . . . x xPhenytoin . . . . . . . . . . . . . . . . . . . . . . . . . . x xPhosphorus, inorganic . . . . . . . . . . . . . . . X x x x x xPotassium . . . . . . . . . . . . . . . . . . . . . . . . . X x xPrimidone. . . . . . . . . . . . . . . . . . . . . . . . . . x xProtein, total . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xPseudocholinesterase . . . . . . . . . . . . . . . x x xSalicylate . . . . . . . . . . . . . . . . . . . . . . . . . . x x xSodium . . . . . . . . . . . . . . . . . . . . . . . . . . . . X x xTriglyceride . . . . . . . . . . . . . . . . . . . . . . . . X x x x x xUrea nitrogen(BUN) . . . . . . . . . . . . . . . . . X 1 x x x x xUric acid . . . . . . . . . . . . . . . . . . . . . . . . . . . X x x x x x xSOURCE: Personal communications and marketing materials from the manufacturers(10,24,46)

market leader, responsible for perhaps 40 to 50percent of all sales of automated multichannelchemistry analyzers in the United States.“Clinical systems” represented 57 percent of itstotal 1978 revenues of more than $275 million.Another 27 percent of Technicon’s 1978 reve-nues were contributed by sales of reagents andc o n s u m a b l e s .

SMA 6/60 and SMA 12/60.—These noncom-puterized analyzers are descended directly fromthe SMA 12 “Hospital Model” that was intro-duced in 1965. Their names indicate that they

have 6 and 12 channels, respectively, and thateach is capable of processing 60 samples perhour. Both analyzers are equipped with aflamephotometer for analyzing serum electrolytes.The manufacturer estimates that at least 2,000SMA 6/60s and 5,000 12/60s (first sold in 1969)are currently in place (46).

SMA II.–The SMA II was first sold in 1977,and the manufacturer estimates that severalhundred SMA IIs are now in place. Two dif-ferent models are available, one with 12 chan-nels (SMA II-12) and the other with 18 (SMA


11-18). The SMA 11 has a computer that can beused for extensive report production. A flamephotometer is used for analyzing serum elec-trolytes. Both models process 90 samples perhour. The SMA II requires smaller sample sizesthan do the SMA 6/60 and 12/60. In the SMA11-18, which is, in effect, composed of a 6-channel and a 12-channel module, reagent econ-omy is improved somewhat, since only themodules that contain tests included in the testorder are activated.

SMAC.—SMAC is the largest of Technicon’sautomated multichannel analyzers. It has 20channels and can process 150 samples per hour.SMAC is computerized, allowing for extensivereport preparation, computerized verificationthat all operations are occurring in their appro-priate sequence, and selective reporting of ab-normal results, among other functions. Sincethe first sale in 1974, more than 1,000 SMACshave been sold. This analyzer offers more spe-cific chemical methods than previous Techniconinstruments. Electrolytes are analyzed by anion-selective electron method instead of by thetraditional flame photometer.

Discrete-Sample Analyzers

General Description

Discrete-sample analyzers, in effect, replicatethe manual process of testing. These devicesperform a number of tests sequentially on onesample. Opinion is divided on whether thespeed, accuracy, and precision of discrete-sample analyzers are comparable to those of thecontinuous-flow analyzers (42,43). Increasingsales indicate, however, that these analyzers fillwhat many clinical laboratory directors per-ceive to be a need.

A discrete-sample processor is a collection ofrelatively independent, general purpose chan-nels that are run in parallel. The channels aretied together at three points: 1) sample presenta-tion, made in sequence to successive channels;2) sample transport, at several reaction tem-peratures; and 3) data acquisition, accom-plished through a computerized system thatreads dedicated detectors in each channel. Usersmay select only those tests whose results areneeded; other tests will not be run,

The prototypes for the discrete sample ana-lyzers were American Optical’s Robot Chemistand AGA’s AutoChemist. Today, there aremany models available. The models of twomanufacturers, Hycel and Du Pent, are dis-cussed below.

Hycel, Inc.

Hycel is primarily a manufacturer of clinicalchemistry analyzers. This company derives alarge share of its revenues from sales of auto-mated analyzers and related reagents. In 1978,37 percent of Hycel’s total revenues of almost$39 million were from the sale of clinical sys-tems and 39 percent were from sales of reagentsto be used with automated analyzers (23).

Among Hycel’s products are three multichan-nel chemistry analyzers: the Super-Seventeen,the SKS-60, and the Hycel-M. These analyzersrange in price from $75,000 to $225,000 (seetable 1). The prices of the machines differ de-pending on the number of channels and testsavailable, output as measured in test results perhour, and the test methods used.

Super-Seventeen. — Introduced in 1975, theSuper-Seventeen is a computerized, program-mable multichannel analyzer with 17 channelsthat is capable of performing 18 tests (see table2). The Super-Seventeen uses a flame photom-eter to measure electrolytes. It is capable ofprocessing 60 samples per hour, yielding 1,020determinations. Patient data are stored oncassettes.

SKS-60.—First sold in 1978, the SKS-60 has17 channels and is capable of performing 15tests (see table 2). In place of the traditionalcalorimetric methods used by most analyzers,the SKS-60 employs newer, more specific en-zymatic test methods that are thought to pro-vide more reliable results because of reduced in-terference. The following functions are undercomputer control: reagent dispense timing, in-cubation, calculation of chemistry results, andsystem calibration. If the appropriate tests areselected, globulin, AG ratio, and BUN/creati-nine ratio are calculated and printed outautomatically. The SKS-60 can process. 60samples per hour, yielding 900 determinations.


Hycel-M.—The Hycel-M is Hycel’s newestautomated analyzer at this writing. This compu-terized analyzer has 30 channels and is capableof processing 120 samples per hour. Twenty-four tests are now available (see table 2). Anycombination of individual tests, panels, or pro-files may be selected by the operator. Themachine is equipped with a microprocessor thatprocesses samples and another processor thatmanages data.

E. I. du Pont de Nemours & Co.

Du Pont is a large firm with many interestsand sources of revenue, among which auto-mated chemistry analyzers and reagent sales arerelatively much less important than they are toTechnicon and Hycel. Nonetheless, Du Pont’sAutomatic Clinical Analyzer (ACA) line is be-ing expanded and refined, and sales are consid-erable in terms of market penetration. The com-pany estimates that 30 to 40 percent of hospitalswit h more than 100 beds have an ACA in theclinical laboratory; about 70 percent of theACAs are thought to exist alongside othermanufacturers’ equipment. Du Pont emphasizessmall-batch processing, so its equipment is oftenused to complement the larger processors. TheACA was first sold in 1970; ACA sales recentlypassed the 2,000 mark (10).

Du Pont manufactures three multichannelanalyzers: the ACA-I, ACA-II, and ACA-III.The number of channels and tests, outputspeeds, and purchase prices are shown in table1. All three machines are mechanically similar.Each operates with an individual test pack foreach analysis. The analyzer is loaded with anumber of samples and, for each sample, thecorrespondin g test packs containing the ap-propriate reagents, The test pack serves as bothreaction chamber and cuvette for photometricanalysis. Either standard or enzyme methodscan be incorporated into the test packs. Undercomputer control, the proper amounts of sam-ple and diluting agent are injected into the testpacks. Temporary seals around the reagents inthe packs are broken with hydrostatic pressure.The analyzer then mixes the reagents, waits apredetermined amount of time, and measuresthe outcome of the reaction. One test pack is ac-

cepted every 37 seconds, so a total of 97 deter-minations can be made per hour.3

The number of channels refers not to physicalapparatus, but to the maximum number of teststhat can be programed into the machine at anyone time. The ACA works best in a lab that runsonly a few profiles or many individual tests eachday, or in a lab needing fast, accurate emergen-cy or small-batch test processing. It would notcompete with a multichannel instrument in a labwhere dozens of multiple-test profiles were runeach day.

ACA-I.—This analyzer has 30 channels andcan perform 29 different clinical tests (see table2). It produces a report slip that has two parts: aphotographic reproduction of the handwrittensample identification card, and a computer-printed list of the test names and numericalresults.

ACA-II.—The ACA-II is available in a 30- ora 60-channel version. Thirty-eight test methodscan be selected by the purchaser (see table 2).One of these, the CPK-MB, has been “technical-ly released” (i.e., it is not yet universallyavailable). Results are provided by the ACA-IIin the same format as results provided by theACA-I.

ACA-III.—This machine is similar in manyrespects to the other ACAs. The main differenceis that it contains a microprocessor, whichmakes it programmable and thus very versatile.The ACA-III can be programed to run tests onas many as 625 channels in sequence. At pres-

‘Alth{mgh t h e ACA I S, s t r i c t l y s p e a k i n g , a multlchdnne]

analyzer , because I t autt~matlcally wi thdraws the apprc>pr]ate

am(wnt (d sample tr{}m a sample cup to perform many tests, ItIuncti(ms in many respects like a single-channel analyzer. Dllterenttests are fully c(mtalned within the consumable test packs andpassed through the machine In a c(~nvey[~r-belt manner. The tlrsttest pack enters the tlrst preheat area within the temperature cham-ber t(dlowin~ It\ associated sample cup. Thirty-seven secondslater, the “conveyor’’” IS Indexed tiwward, and the Ilrst test pack

m(~ves to the sec(md preheat area wlthi n the temperature chamber,while the second test pack m(wes into the tlrst preheat area.An(]ther 37 sec(mds later, the system indexes torward, m(}vlng the

tlrst test pack t{} the next stati{m, the sec{>nd pack til the secondpreheat area, and the third test pack to the tlrst preheat area. Thepr[)ce~s c[~ntlnues In thi~ way, with the tirst test pack eventual ly

reaching the ph~~t(~meter, where the results are react. In another 37wct)nds, that test resu I t IS printed (w t and the second test packenters the ph(lt(~meter area. Thus, a determination IS made every

37 sec(~nds.


ent, only 38 chemistry tests are available (seetable 2). But since Du Pont has been introducingnew tests (or methods) at a rate of four or fiveper year, the enormous expansion capability ofthe ACA-III may eventually be useful. Themicroprocessor can also reprogram the machineto accept new methods for performing existingtests; provide the flexibility to run adult andpediatric sample sizes simultaneously; and pro-duce a more extensive printout of results, in-cluding an indication that a particular result isabnormal, along with the range of normalvalues for the test, the units of measure for thetest results, and the time of day the test was run.

Centrifugal Fast Analyzers

Centrifugal fast analyzers were developed atthe Oak Ridge National Laboratory. The firstcommercial machine was sold in 1970. Theseanalyzers are designed to run a single test on alarge number of samples—the reverse of themultichannel concept of running a large numberof tests on a single sample. The commerciallyavailable centrifugal fast analyzers today in-clude the GEMSAEC (Electronucleonics, Inc. ),the CentrifiChem (Union Carbide), and theRotochem II (AMINCO). These analyzers donot compete directly with the high-speed, multi-channel analyzers such as SMAC and Hycel-Mat present, but they may within the next fewyears.

Design Issues and Tradeoffs

Accuracy and Precision

For both continuous-flow and discrete-sampleanalyzers, there is a tradeoff between speed andreduced variable cost, on the one hand, and ac-curacy and precision, on the other. The tradeoffhas been most clearly recognized in the design ofcontinuous-flow analyzers; the problem of “car-ryover” from one sample to the next in the com-mon tubing of these analyzers has been studiedat length (s). There are also problems of samplecross contamination in discrete-sample analyz-ers. Companies have directed their product im-provement efforts (e.g., in Technicon’s SMA IIand SMAC) to minimizing the carryover prob-lem; despite this problem, however, the accu-racy and precision of multichannel instruments

have long been recognized to be superior tothose of manual methods (54).

Another theoretical problem with continu-ous-flow analyzers and certain discrete-sampleanalyzers has to do with the timing of reactions.If reactions are all required to be synchronized,then suboptimal accuracy must be tolerated forthe slower reactions, or a reduction in theoverall processing rate must be made to accom-modate the slowest reactions, or additionalchannels must be created to split the fastest reac-tions into two or more parts.

Speed

The ability of automated equipment to ana-lyze specimens rapidly is increasing to the pointwhere the speed of analysis is no longer the rate-limiting step in the overall testing process (28).Limits in the ability to speed up the specimen in-put and coding system and to record and reportresults make further increases in equipmentspeed of little value. This is reflected in therelatively modest sales in the United States ofthe Vickers M300, which can analyze 300 sam-

ples per hour on 20 channels, yielding 6,000determinations per hour. Concerns for futuredevelopment of automated equipment will shiftto improved versatility and reliability, asevidenced by Technicon’s development of themore flexible, intermediate-sized SMA II andDu Pont’s development of increased test selec-tion in the ACA-II and ACA-III.

Selectivity y

Discrete-sample analyzers have an apparentcost advantage over continuous-flow analyzers,because they have the capability of performingonly those tests desired for each sample, thusnot only saving reagents, disposable, and chan-nel maintenance, but also minimizing the num-ber of unsolicited tests. On the other hand, thereagent cost per test is so low with the availablecontinuous-flow instruments that this advan-tage may be more theoretical than real.4

4See review of cost data below in the part of this case study onthe economics of the multichannel analyzer.

Case Study #4: Cost Effectiveness of Automated Multichannel Chemistry Analyzer-s ● 13

Future Trends

Instruments

The rate of product turnover in the market formultichannel chemistry analyzers is rapid. Man-ufacturers, in their user cost estimates, suggestthat machines be depreciated over 4 to 7 years,although the machines’ useful life could be muchlonger (1,22). High corporate expenditures onproduct development -e.g., in 1978, Techniconand Hycel spent 7.4 and 8.7 percent, respective-ly, of their sales revenues on research and devel-opment (23,45) — undoubtedly contribute to therapid rate of turnover. That Technicon spent $7million to develop SMAC (34), which must berecovered in the purchase price of equipment,further suggests possible inefficiencies in theequipment replacement rate.

The rapid growth of centrifugal fast analyzersin this market cannot be ignored. Despite their

single-channel capability, such analyzers arealready competing successfully in the samemarket as the multichannel instruments.

Chemistry Tests

A number of chemistry tests that have hereto-fore been rather expensive and had to be per-formed manually will be available in the nearfuture on multichannel analyzers, especially ondiscrete-sample systems. Examples include iso-enzymes of creatine phosphokinase (CPK-MB),high-density lipoproteins, renin, and vanillyl-mandelic acid (VMA). If added to multichannelequipment, each of these will undoubtedly in-crease in use, and the possible implications forclinical practice in the areas in which thesetests are used (e.g., diagnosis of myocardial in-farction and coronary disease, diagnosis andtreatment of hypertension) should be examinedcarefully.

FRAMEWORK FOR COST= EFFECTIVENESS EVALUATION

Efficiency and Cost Effectiveness

There are at least two levels of questions onemight ask concerning the cost effectiveness ofthe automated multichannel analyzer. The ques-tion at the simpler level is whether, for a givenpattern of test ordering (i. e., for a specifiednumber of each type of test ordered per unit oftime), a Particular class of automated multi-channel analyzers can produce results at lowercost than (and with accuracy and precision atleast equal to) alternative methods. This may becalled the question of economic efficiency. Anevaluation of the economic efficiency of multi-channel analyzers for any particular pattern ofutilization would be straightforward, given dataon the various components of testing costs. s

Even with information on the most efficientmeans of producing a specified set of test resultsover time, however, it would remain to be de-termined whether any particular pattern of uti-lization, and therefore any particular analyzer,is cost effective. This determination would re-

These are revlevveci below In the part ot this case study on theecon(lmic~ 01 the mllltlchannel a ndlvzer,

quire answers to a more complex set of ques-tions, which concern not only the costs oftesting, but also the induced costs and healthbenefits attributable to the many availablechemistry tests in their various clinical uses.

One important limitation of the analyticframework described below ought to be men-tioned. The analytic approach is designed topermit evaluations of the cost effectiveness ofdiscrete clinical chemistry technologies (e.g.,automated analyzers v. manual, multichannelv. single channel)—not of the cost effectivenessof configurations of equipment in a clinicallaboratory. Although evaluation of the latterwould surely be the more realistic basis forpolicy, we offer this more microlevel approachas a necessary step toward that complex under-taking.

Economic Efficiency

Available data concerning the question ofeconomic efficiency in the production of testresults are reviewed in the part of this case studyon the economics of the multichannel analyzer.


Here we offer a conceptual framework foranalysis of the efficiency question.

Fixed and Variable Costs

Both economic theory and cost-accountingprinciples offer methods of classifying costs forpurposes of analysis. One distinction that is cen-tral to the evaluation of multichannel analyzersis that between costs that do not vary with thevolume of testing (fixed costs) and costs thatvary in direct proportion with the volume (vari-able costs). A major economic effect of auto-mation in the laboratory, generally, and ofmultichannel analyzers, in particular, has beento transform costs that had previously beenvariable into fixed ones (4,35). The result is thatincreasing automation becomes economicallymore attractive at higher volumes of use.

Among the costs that are typically consideredto be fixed, that is, independent of the numberof samples analyzed, are equipment costs, in-surance, maintenance, laboratory supervision,and overhead (including administration, space,utilities, etc. ). Variable costs, that is, thoseroughly proportional to the number of samplesanalyzed, include reagents, and consumablesand supplies. Automation reduces the variablecosts of reagents and supplies per test, while in-creasing the fixed costs of equipment, insurance,and maintenance.

Direct labor, the most important economicelement, is not yet reflected in this typology ofcosts. Labor costs for technician and technolo-gist time do not fall neatly into either category.Some aspects such as sample collection, codingand preparation, and reporting of results aremore variable than fixed. Other aspects, espe-cially sample processing and analysis, are morenearly variable if tests are performed manually,but more nearly fixed if automated equipment isused. An additional fixed labor cost with auto-mated equipment is training of technicians.Thus, a major impact of automation has been totransform many labor costs from variable tofixed (35).

One common misperception is that automa-tion reduces the need for highly skilled profes-sional labor and thus reduces unit labor costs.

Commenting on this, Mather points out that thetraining and skill required for adequate qualitycontrol and supervision are no less withsophisticated instruments than with manualmethods (28).

The validity of all cost estimates is, of necessi-ty, compromised by market imperfections inboth the production process and the hospital.As in virtually all benefit-cost and cost-effec-tiveness analyses, one must settle for proxies for“true” costs, recognizing that biases may existand endeavoring to adjust for the most impor-tant and obvious of these,

The Cost Envelope

The problem of selecting the most efficienttechnology for producing a specified number oftests has been formulated by several authors(4,11,35). If the assumption is made that deter-minations are undifferentiated and that the onlyvariable affecting cost is the number of deter-minations performed, the picture may be repre-sented as in figure 1. This figure displays totaldirect costs, as a function of test volume, forthree prototypical technologies: manual, first-generation automated, and advanced auto-

Figure I.—Cost Envelope for Three AlternativeClinical Chemistry Technologies

Totaldirectcost

N, N2

Number of determinations


mated. The slope of each cost line represents thevariable cost per test; the intercept on the ver-tical axis represents the fixed cost. In reality, ofcourse, the fixed cost for manual determinationsis not zero, although it is much smaller than forautomated technologies. Note that the more ad-vanced technologies exhibit greater fixed costsbut lower variable costs.

For any specified number of tests, the most ef-ficient technology is the one with the lowesttotal cost at that test volume. Thus, for testvolumes less than N1, the manual method ismost efficient; for volumes between N1 and N2,first-generation automation is most efficient;and for volumes above N 2, the advancedtechnology is most efficient. This is a well-known result: Automated chemistry technol-ogies are economically most advantageous athigher test volumes. ’

The cost envelope in figure 1 is characterizedby decreasing incremental costs per test as thetest volume increases. This reflects the costcharacteristics of the underlying technologies.The technologies with lower variable costs (buthigher fixed costs) become relatively more effi-cient as the number of tests increases.

Multichannel Efficiency Issues

The questions of economic efficiency becomemore complex when one takes into account themultichannel aspect of the technology. For onething, the simple distinction between fixed andvariable cost no longer suffices. There are somecosts that depend on the number of samplestested, but not on which determinations are runon each sample; and there are others that de-pend both on the number of samples and thenumber and identity of determinations per-formed.

Although the distinction may be unimportantfor continuous-flow instruments once they havebeen designed and purchased (since all tests onthe machine are run on every sample), it is veryimportant when equipment design decisions aremade. Specifically, two questions have impor-tant efficiency implications: 1) how to select the

“Data c(~ntlrmlng this p(~lnt are rev]ewed below In the part Otthis case study on the econ[)rnlcs of the multichannel analyzer.

particular tests to include in the 12, 20, or 3 0channels; and 2) how to ascertain the most effi-cient number of channels. At this point, we arestill assuming that the number and distributionof tests ordered have been specified; hence, theeconomic analysis depends on the test-orderingpattern.

Optimal Selection of Channels.—With theobjective of minimizing costs, the selection ofchannels in a multichannel analyzer ought to de-pend on the frequency with which determina-tions are requested, individually and in groups.In the extreme case, if a particular test werealways ordered alone, it would probably not beefficient to include it among the channels of amultichannel continuous-flow analyzer. At theother extreme, if certain profiles were always re-quested together, they would be more logicalchoices for automation.

Taylor contends that the key economic con-sideration in deciding which tests to automate isthe number of aliquots of sample to be proc-essed by the lab technician (44). He reports that,prior to the introduction of an SMA 12/60 in hishospital, the average number of aliquots re-quired per specimen was 1.98. (Ideally, theaverage would be 1.00, the number that wouldobtain if all multiple-determination specimenscould be accommodated by the multichannelanalyzer. ) With the SMA 12/60, the optimalcombination of tests yielded an aliquot rate of1.42 per specimen. The 20-channel SMAC, op-timally configured, would reduce this rate to1.21 per specimen, given the test-ordering pat-tern at this investigator’s hospital (44).

In principle, it might be desirable to exclude arelatively common test from the multichannelanalyzer if it were usually ordered alone. In thestudy cited above, for example, the optimal setof 12 tests in the SMA 12/60 excluded calcium,inorganic phosphorus, and cholesterol, eventhough these tests are ordered more often thansome tests that were included in the efficient set.

Optimal Number of Channels.—Our reviewof the literature revealed no studies evaluatingthe most efficient number of channels to buildinto a multichannel analyzer, given a particulartest-ordering pattern. Specifically, the merits of


6, 12, or 20 channels have not been evaluated.Assuming any specified test-ordering pattern,the question of whether to add a channel to thedesign of a continuous-flow multichannel ana-lyzer ought to depend on the following factors:

1.2.

3.

4.

5.

the fixed cost of the new channel;the variable cost of using existing channelswhen the test on the new channel is or-dered alone;the variable cost of using the new channelwhen only tests on existing channels arerequested;the differences, for the new channel, be-tween the variable cost per test on multi-channel equipment and the cost per test ondedicated equipment or by manual meth-ods; andthe cost savings for aliquot preparationfor orders involving the new channel andat least one existing channel.

Only if there exists a test for which the cost in-creases described by items 1 through 3 are out-weighed by the cost reductions described by theitems 4 and 5 is it efficient to add a channel tothe design.

Further complicating the problem of evalua-tion is the fact that test-ordering patterns differfrom institution to institution. Hence, equip-ment-design decisions must be based on aggre-gates across the range of institutions that con-stitute the potential market for the instrument.

Of course, multichannel analyzers may in-duce changes in test-ordering behavior whichmay alter the initial findings from analyses thatare conducted before the instrument is in place.Under such circumstances, the question is notmerely one of economicclinical cost effectivenessordered.

Cost Effectiveness

efficiency, but of theof the additional tests

Overview

For the second, more complex question, con-cerning the cost effectiveness of alternativetechnologies, we present a framework that canbe used to evaluate the cost effectiveness—in thebroader sense—of automated multichannel ana-

lyzers. This framework is based on more de-tailed presentations available elsewhere (11,50,51,52) and on research in progress.

As stated earlier, cost analyses have shownthat the more advanced analyzers are cost effec-tive at high test volumes, but not at lower ones(35). This result can be seen in figure 1. As thenumber of samples to be analyzed increases, theincremental cost of additional tests falls. Thissituation, known to economists as “diminishingmarginal cost, ” is juxtaposed with “diminishingmarginal benefit. ” That is, additional testsordered are decreasingly beneficial from theclinical perspective (assuming that physiciansare already making the best use of the level oftesting they utilize). This juxtaposition neces-sitates explicit channel-by-channel and use-by-use analysis of individual chemistry tests toevaluate ‘the overall cost effectiveness of theautomated multichannel analyzer.

The complexity of this situation is illustratedin figure 2. Let us suppose that costs and healthbenefits have been measured on the same scale.This could be done by valuing each year of lifesaved by the opportunity cost of saving a yearof life by alternative means (i. e., by somespecified cutoff value of a cost-effectivenessratio) (52). This assumption will be relaxed lateron. In the present formulation, however, thedifference between the height of the curvemeasuring the total cost and the height of thecurve measuring total benefit is the variable ofinterest—it represents the net benefit of testing.The segments of the benefit curve representdecreasingly beneficial uses of the test as thenumber of available tests increases. The slopesof these segments represent the expected valueper test. The segments of the cost curve (as infigure 1) represent successively more advancedtechnologies exhibiting diminishing incremen-tal, but increasing fixed, costs.

By inspection of figure 2, we see that at thepoint of switchover to the most advanced tech-nology (where the number of tests is N2), theslope of the benefit curve becomes steeper thanthat of the cost curve; incremental benefits ex-ceed incremental costs. This does not necessar-ily mean that it is cost effective to use the ad-


Figure 2.—Hypothetical Analysis of the Cost= Effective Level of Testing

Totalcostperyear

f

vanced technology to test beyond this level. Totake advantage of this low variable cost, wemust pay the high fixed cost, F. If the incremen-tal benefit continues up to the level N+ , andthen flattens out as the beneficial uses of the testare exhausted (solid line in the figure), thensociety is actually better off testing at level N*with the intermediate technology than at N+

with the advanced technology. At N*, the netbenefit is maximized. If, however, the incremen-tal benefit of testing continues all the way up toN + + (dotted line in the figure), then it wouldbe best to use the advanced equipment for thislarger number of tests.

The above remarks relate only to evaluationfrom a societal perspective, in which the costs ofhealth care resources are considered. From theperspective of patients, considered here not aspayers for their own health care, it would bebest to test at the highest possible level at whichbenefits accrue to the patient. This would occurat N+ if the solid benefit curve were to obtain,a result that would be at odds with the societalinterest (1 1).

Totalbenefitperyear

Thus, the cost effectiveness of multichanneltechnology depends on the slope of the benefitcurve as the number of tests increases. This, inpractice, depends on the value of such tests inscreening patients who are asymptomatic forthe conditions being tested. It is not clearwhether the most advanced multichannel ana-lyzers can pay for themselves, in efficiencyterms, without increasing the use of the chem-istry laboratory for screening purposes. If theycannot, then the evaluation must turn to theissue of clinical efficacy of such uses.

Levels of Evaluation of Clinical Efficacyand Cost Effectiveness

The cost effectiveness of multichannel ana-lyzers can be considered at each of four levels:1) technical quality of measurement; 2) diagnos-tic value of test results; 3) clinical efficacy, interms of expected impact on-treatment decisionsand patient outcomes; and 4) overall cost effec-tiveness.

Technical Quality of Measurement.—Thetechnical quality of measurement in automated

-——

18 . Background Paper #2: Case Studies of Medical Technologies

analyzers may be described in terms of precision(absence of test/retest variability) and accuracy(absence of bias or systematic departure fromthe true value). The precision and accuracy ofthe major automated analyzers are generally re-garded as quite good, largely because such ana-lyzers eliminate human error in measurement(5,28,54). For some tests, however, althoughwithin-laboratory reliability is excellent, be-tween-laboratory variations may be important.Between-laboratory variations have been notedespecially in relation to cholesterol and SGOTdeterminations (12).

Furthermore, there is a tradeoff between ac-curacy and speed because of the problem of con-tamination of results from one specimen to thenext. This tradeoff also arises when the rate-limiting reaction among, say, 17 or 20 tests isaccelerated beyond optimal rates to keep stepwith the others. In weighing this tradeoff from acost-effectiveness point of view, the costs ofrepeat measurements induced by lack of accu-racy must be considered.

Finally, although minimizing human error,machines suffer from fatigue and are prone tosystematic failures that must be monitoredclosely by liberal interspersing of controls ineach batch of tests (5,28). The costs of these con-trols must also be considered and compared tothose for manual testing, because they compro-mise both the unit cost reduction attributable toautomation and the effective rate of testing (28).

Diagnostic Value.—There are many possiblemeasures of the diagnostic value of a test (31).These include its predictive value positive andpredictive value negative (49). The former is theproportion of positive test results that truly cor-respond to disease or abnormality; the latter isthe proportion of negative test results that trulycorrespond to the absence of disease.

The diagnostic value of a test can only bemeasured in the context of how the test is used.It is well known that the predictive valuepositive of a test depends not only on the pro-perties of the test (i.e., its sensitivity andspecificity y), 7 but also on the prevalence of the

“rt,~t sen~ltlvlty I\ the pr(lp{~rtlt~n (d p(w]tlve tests amtmg dls-

eawd patwnts: test speclticlty IS the pr(}p(~rtlt~n (~t n e g a t i v e tests

a m ~Jng ntlnd]wased pat Ien t$.

disease in the tested population. The lower theprevalence, the lower the predictive value of apositive result. As the number of tests run on amultichannel analyzer increases, most determi-nations will be performed despite the absence ofany particular suspicion of abnormality; hence,the prevalence in the tested population of theconditions being tested for will be very low. Itfollows that the probability that any singlepositive result will be a true positive is very low.

One example cited in the literature relates tothe use of the serum calcium value to screen forparathyroid cancer (12). Suppose that the prev-alence of this tumor in the screened populationwere 1 per 1,000. Suppose also that the prob-ability that a patient with no parathyroid ab-normality had a 5-percent chance of having anelevated serum calcium level, according to theusual definition of normal limits. Then, theprobability that a patient who tests positive willactually haveas follows:

+ + h i g h ]

a parathyroid tumor is calculated

=

( 1 )(0.001).

= 0.02

Forty-nine of 50 patients who screen positivewould have no tumor. Many of these wouldeventually be ruled out on the basis of repeatedblood and urine tests. Some might be subjectedto biopsies that prove to be negative.

As the number of tests increases, the numberof false positives also increases. If 20 independ-ent tests, each with a false-positive rate of only 5percent, are run on a normal patient, odds arenearly two-to-one (64 percent) in favor of ob-serving at least one positive result. That positiveresult must then be verified and possibly fol-lowed up at some cost. The induced costs thatresult from positive tests generally—and falsepositives in particular—must be consideredcarefully in evaluating the overall economic im-pact of multichannel analyzers.

It is not clear what the physician’s attitudeshould be with respect to unexpected positive re-sults (25). If the result is so unexpected that the


physician considers an option to be simply to ig-nore it, then perhaps the test should not havebeen ordered. If, on the other hand, the test isavailable at virtually zero incremental cost on amultichannel profile, then should the physiciandiscard the information? This issue, the appro-priate inference to be drawn from test results inthe context of a multiple-test panel, is unre-solved both in theoretical and practical terms.

Clinical Efficacy.—The clinical efficacy of themultichannel analyzer also depends on the pat-tern of its utilization. One might differentiatebetween its clinical efficacy in optimal usage(defined, perhaps, by cost-effectiveness criteria)and that for actual usage (i. e., its effectiveness).Evidence for the former requires analysis thatbecause of both data limitations and methodo-logical obstacles has not yet been undertaken(50). There is, however, some fragmentary evi-dence concerning the latter.

One measure of effectiveness, albeit imper-fect, is the degree to which test results contributeto altering or confirming an existing treatmentplan. Dixon and Laszlo (8) reported that only 5percent of automated chemistry tests ordered byhouse officers in their study were used in thisway. These investigators found that, on aver-age, only 3 of the 12 chemical values providedby the analyzer were actually wanted by thehouse officers. After an intervention in whichthe physicians were restricted in the number oftests they could order, the percentage of testresults that contributed to altering or confirm-ing the course of treatment increased from 5 to23 percent.

A study by Durbridge, et al. (9) corroboratedthese findings. They found that the effects of in-troducing hospital admission screening with amultichannel analyzer were to increase the num-ber of tests during the patients’ hospitalizationby 78 percent (not including the screening pro-file itself), to increase consultations by 25 per-cent, and to increase laboratory costs by 64 per-cent. These effects had no measurable healthbenefit to any patient in the study.

Other studies have found multichannel chem-istry screening to have appreciable value, how-ever. Carmalt, et al. (7), in evaluating a 14-test

screen on hospital admission, reported that newor additional diagnoses were found in 16.9 per-cent of patients. An additional 21.6 percent had“unexplained abnormal” results, such as hyper-cholesterolemia or hyperuricemia. Of all pa-tients tested, 1.4 percent were found to havedeteriorating or newly discovered renal disease,a finding that led to treatment in four patients.

Belliveau, et al. (3) found clinical value in an18-channel chemical admission screen in a com-munity hospital. In 7.4 percent of patientsscreened, test results indicated or “possibly” in-dicated a new diagnosis. These included: 21cases of diabetes mellitus, 9 of gout, 6 of cir-rhosis, 5 of hypercholesterolemia, and 1 ofmyeloma. Tests most often found to be abnor-mal were uric acid (15.1 percent), SGOT (13.5percent), and LDH (12.3 percent). In all, nearlyhalf (43.2 percent) of patients had at least oneabnormal result, leading in most cases to retest-ing and further diagnostic workups.

Thus, the evidence as to the effect of chem-istry screening on treatment decisions is mixed.There are undoubtedly some treatable condi-tions discovered as a result of such screening,but the value of early intervention in those con-ditions must be examined. Clearly, better meas-ures of clinical effectiveness of tests and theiruses are needed, as are studies of the benefits ofindividual tests in the full range of their clinicaluses. The methods of decision analysis may beuseful in carrying out such evaluations, usingexpected improvement in patient outcome as theultimate measure of effectiveness (19,51).

Cost Effectiveness. —Weinstein and Fineberg(51) describe an analytical approach for evaluat-ing the cost effectiveness of an individual chem-istry test in one of its uses. The methodologicalproblems of cost-effectiveness analysis increasedramatically, however, when one moves froman individual test in a single use (e. g., VMAin the diagnosis of pheochromocytoma in anasymptomatic hypertensive patient) to an in-strument that performs multiple tests, each ofwhich may have several clinical uses.

The problem is complicated further because itwould be inappropriate, in analyzing the costeffectiveness of state-of-the-art multichannel in-

20 . Background Paper #.?: Case Studies of Medical Technologies

struments, to compare them with fully manualtesting. It is, however, appropriate to ask:1) whether it is cost effective for a particular testto occupy a channel in any such instrument, and2) under what circumstances the reduced varia-ble cost of an instrument justifies its increasedfixed cost relative to alternative equipment.

The cost of any diagnostic technology largelydepends on the costs induced by test results.These include the costs of repeat tests requiredto compensate for measurement error or to con-firm presumptive diagnoses. Running more testsgenerates more positive results, leading to great-er induced costs to verify these findings (9).These induced costs must be balanced againstthe decreasing cost per test in laboratories thatuse multichannel analyzers at high volumes.The issue of induced costs must also be consid-ered in evaluating the relative cost effectivenessof continuous-flow (unselective) and discrete-sample (selective) analyzers. Technicon’s SMA11 and SMAC systems address this concern byautomating a function that many hospital lab-oratories have provided all along: selectivereporting of only those tests requested by theclinician.

One of the authors of this study (Weinstein),under a grant from the National Center forHealth Services Research, is now developing ananalytical model for cost-effectiveness analysis(CEA) of the multichannel analyzer. The modeladdresses two questions: 1) What is the cost ef-fectiveness of a test in each of its clinical uses?and 2) Given the answer to that question, whatis the cost effectiveness of automating a par-ticular test? The model is used here to analyzethe marginal value of a single test. Methods toevaluate the joint cost effectiveness of manytests remain to be developed.

The analysis proceeds in several steps. Thefirst is to estimate the cost envelope for the test,i.e., the cost of testing as a function of thenumber of tests ordered per unit time, assumingthe most efficient testing technology at eachlevel (this was shown in figure 1). The next stepis to specify the clinical indications for the test,including the possibility of using it for screen-ing. Estimates of the frequencies with whichthese indications arise must also be obtained.

Next, for each clinical indication or use, theexpected value of clinical information (EVCI) isestimated using a decision-analytic framework.The EVCI is a measure of the average amount ofhealth benefit (e.g., quality-adjusted life yearssaved (QALYs) (52)) per test. Also, for eachclinical use, the expected induced costs per testare estimated. Methods for performing both ofthese analyses have been demonstrated (51).

Assuming that the first indication occurs inN, patients per year in a particular setting, thatthe second indication occurs in Nz patients peryear, and so forth, the aggregate expected bene-fit and aggregate expected induced cost can beplotted as a function of the number of tests foreach use (see figure 3). Each curve is actually aline segment whose slope is the benefit or in-duced cost per test, and which extends from theorigin to the maximum possible number of tests(N,, Nz, etc.). For screening, the correspondingsegment may be replaced by an open-ended ray,assuming that there is no limit to the number ofpotential candidates for screening. Alternative-ly, the limit for a hospital may be the total num-

Figure 3.—Analysis of Cost Effectiveness fora Test With Multiple Uses

Benefits as a function of number of tests, by use

B, B2

Use 1 Use 2 Use 3

Induced costs as a function of number of tests, by use

c1

ExpectedInducedcost


ber of hospitalizecl patients, less the numberwith some specified indication.

The next step in the analysis is to construct anet benefit curve for each possible cost-effective-ness cutoff level. This cutoff level, k, mightrange from $1,000 per QALY to arbitrarily highvalues of cost per unit of health benefit. For eachvalue of A, and for each clinical use, the netbenefit per test is calculated as:

Bi is the expected benefit per test (in life years orQALYs), Ci is the expected induced cost pertest, and A is the cost per unit of benefit. A netbenefit curve, analogous to that in figure 2, isthen constructed by concatenating line segmentswith slopes B~et(X) and with horizontal spans Ni,in decreasing order of their slopes. This isshown in figure 4. For each cost-effectivenesscutoff value, A, the optimal level of testing,N*(X), can be found as the point of maximumvertical distance between the net benefit curveand the testing cost curve.

Figure 4.—Composite Analysis of theCost-Effective Level of Testing

N“ (A) N, Optimal Number(global (level at level if of tests, Noptimum) which auto- automated

mation becomesefficient)

aTh e variable A IS the cost-effectiveness cutoff level ($/OALY)

For the hypothetical illustration in figure 4,automation is not cost effective at this particularcutoff level. For sufficiently high values of A,reflecting the ability to pay more per unit ofbenefit despite scarce resources, the cost-effec-tive level of testing will shift to the right andinto the domain in which automation is effi-cient.

Concluding Observations Based onthe Conceptual Framework

The foregoing conceptual framework forevaluating the efficiency and cost effectivenessof the multichannel automated analyzer sug-gests that the cost effectiveness of automationcannot be assessed independently of the level ofresources in the health care system, nor of theprice society is able and/or willing to pay forthe diminishing incremental benefits to be de-rived from marginal indications for testing. Inorder to reach volumes at which multichannelinstrumentation is efficient, for example, wemust be willing and able to pay for relativelylow-yield uses of tests. If resources were trulyscarce, however, we would be able to affordonly the most essential uses of tests—and thesewould be most efficiently performed by manualmethods. In summary, the question may bewhether or not we, as a society, really wish topay enough for diminishing incremental healthbenefits to justify sufficient levels of testing to

justify, in turn , the reduction in unit testingcosts that the multichannel analyzers offer.

The empirical question, yet to be analyzed, isat what point on the spectrum of clinical usesfor each test the multichannel analyzer becomesefficient. Is it necessary to adopt hospital ad-mission screening or well-patient ambulatoryscreening to render the multichannel analyzerefficient? Or does the level of testing necessaryto render the multichannel analyzers efficient( i . e . , N2 in figure 1 ) occur well before we ex-haust the generally accepted and cost-effectiveindications for testing? In the answer to thisquestion may lie the answer to the overall ques-tion as to the cost effectiveness of multichanneldn.] 1 vz~’f% In 01] r h~’.j I t h cdr~) Svstem.

—


ECONOMICS OF THE MULTICHANNEL ANALYZER

Overview

In this part of the case study, we review theavailable data concerning the costs of perform-ing chemistry tests on multichannel equipment.The data reviewed are intended to answer thequestion of economic efficiency: What is thecost of producing test results according to aspecified pattern?

In the first section, we review data pertainingto each component of chemistry laboratory costthat may be influenced by the choice of equip-ment. These cost components, reviewed sepa-rately below, are as follows: 1) direct nonlaborcosts (equipment, service and maintenance, re-agents and consumables); 2) direct labor costs;and 3) indirect costs.

Data on direct nonlabor costs were obtainedfrom manufacturers as the prices charged toinstitutional purchasers (10,24,46). For pur-poses of this review, we concentrate on the fol-lowing products: Technicon’s SMAC and SMA12/60, Hycel’s Super-Seventeen, and Du Pont’sACA II.

Data on direct labor costs are the most dif-ficult to obtain, because estimates of these costsrequire estimates of worker productivity in lab-oratories with various configurations of equip-ment. Later in this case study, we will refer totwo published studies in an effort to derive ap-proximate estimates of the magnitude of laborcosts and their sensitivity to the availability ofautomated equipment. Unpublished data onlabor costs obtained by the College of AmericanPathologists were not available to us in perform-ing these analyses, but could be of great value incost-effectiveness analyses (CEAs) to be per-formed in the future.

No published studies suggest that indirectcosts are affected by the types of equipmentused. In the review below, therefore, we discussthis component of cost only briefly.

In the section following the review of costdata, we suggest analyses to address the follow-ing two efficiency questions, applied to current-ly available multichannel equipment. First,

what is the relation between the unit cost perdetermination and test volume, assuming afixed number of determinations per sample?(These cost functions may be compared to unitcosts with dedicated equipment or manualmethods to determine breakeven test volumes. )Second, what is the relation between the unitcost per determination and the number of deter-minations per sample, for various types of ana-lyzers? The published data on labor costs aretoo fragmentary to allow us to resolve thesequestions, but we do suggest an analyticframework for their resolution.

In the third section below, we discuss the costimplications of R&D policies in this area and ofthe rate of product turnover. Then we restatethe problem of induced costs and repeat testsand give some indication of the magnitude oftheir economic impact. Finally, in the last sec-tion, we make some observations on theeconomic incentives faced by hospital labora-tories, given the current regulatory and reim-bursement environment.

Review of Cost Data

Direct Nonlabor Costs

Equipment.—Equipment prices are shown intable 1. The prices for the instruments to be re-viewed here are as follows:

Technicon SMAC . . . . . . . . . . . . . . . . $271,000”Technicon SMA 12/60 . . . . . . . . . . . . . 99,500Hycel Super-Seventeen . . . . . . . . . . . . . 75,000Du Pont ACA II . . . . . . . . . . . . . . . . . . 69,0009

These are list prices, obtained directly from themanufacturers (10,24,46).

The annual cost of owning a machine dependson its useful life and the interest rate. Assuminga 7-year amortization period (22) and a real in-terest rate of 5 percent (after inflation), theannual equipment costs, in constant 1979 dol-lars, are:

Technicon SMAC ... ... ... ... ... .. $45,000

‘1’lus an addit]t~nal $4, OOO it a test c(~nti~urati(~n other than one(}t the 14 standard c~~ntl~ura t l[~ns IS ch(wen.

“Thlrt y-channel m(del.

Technicon SMA 12/60 . . . . . . . . . . . . . 16,500Hycel Super-Seventeen . , . . . . . , . . . . . 12,400Du Pont ACA II . . . . . . . . . . . . . . . . . . 11,400

Service and Maintenance.—Costs for equip-ment service and maintenance (including partsand labor) can be estimated from experiences oflaboratories that perform their own mainte-nance, or from prices charged by manufacturersfor full-service plans. One study reported annu-al maintenance costs in a hospital laboratory tobe $6,100 in 1971, or slightly more than $10,000in 1979 dollars (35). Others estimate that laborcosts for maintenance amount to approximately20 percent of direct labor costs (32).

Although the prices charged for full-serviceplans may overestimate true costs, many lab-or-a tories buy these plans. Technicon offers aplan that includes preventive maintenance,three emergency service calls per year, parts,and unlimited consultation. The annual pricesare $25,000 for the SMAC and $7,900 for theSMA 12/60, or slightly less than 10 percent oftheir purchase prices.

Hycel’s plan for the Super-Seventeen includesservice and maintenance, plus all reagents, sup-plies, control samples, and other items. Theprice is based on test usage. For individual tests,it averages between 12 and 22 cents per deter-mination, depending (in block-rate fashion) onthe volume of usage and actual tests performed.For complete profiles, it averages between 4 and17 cents per test, also depending on the volumeof usage. Charges are determined by readingmeters built into the reach inc.

Du Pont has two service plans for the ACA.One is a rental plan; the user does not buy themachine, but leases it from Du Pont. The fixedprice of $1.13 per determination includes use ofthe machine, maintenance, unlimited emergen-cy visits, parts, consultation, reagents, non-reagent consumables such as control and cali-bration products, and a training course for threetechnologists. Alternatively, customers whoown their equipment may purchase a servicecontract for $6,200 per year. This covers parts,labor, unlimited emergency visits, 24-hour tele-phone consultation, replacement of mechanicaland electrical parts as they are updated, and

replacement of reagents lost as the result of in-strument malfunction.

Reagents and Consumables.—For Techni-con’s equipment, the reagent prices per sampleare shown in table 3.10 Note that for these con-tinuous-flow analyzers, all reagents (20 forSMAC and 12 for SMA 12/60) are consumed oneach sample. Total reagent and consumable costper sample for the SMAC is approximately$1.67; for the SMA 12/60, it is $1.26. Note the

variation among reagent prices; the prices rangefrom 36.8 cents per sample for triglycerides on

“’Under Technlct~n’s block-rate pric]n~, these prices c(~rresp~~ndt~~ a user vvlth a vc~]ume of 100 samples per day ( 26,000 per year-).Unit reagent prices WIJUICI be somewhat Iesi t t~r higher v(~lumes,

Table 3.—Reagent and Consumable Prices forTwo Technicon Analyzers

Cost per sampleItem SMAC SMA 12/60ReagentsAlbumin. . . . . . . . . . . . . . . . . . . . . . . . . $0.011Alkaline phosphatase . . . . . . . . . . . . . 0.036Calcium. . . . . . . . . . . . . . . . . . . . . . . . . 0.005Carbon dioxide. . . . . . . . . . . . . . . . . . . 0.021Chloride. . . . . . . . . . . . . . . . . . . . . . . . . 0.004Cholesterol. . . . . . . . . . . . . . . . . . . . . . 0.022CPK . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.266Creatinine. . . . . . . . . . . . . . . . . . . . . . . 0.003Direct bilirubin . . . . . . . . . . . . . . . . . . . 0.026Glucose (oxidase). . . . . . . . . . . . . . . . . 0.070Inorganic phosphorus. . . . . . . . . . . . . 0.004Iron . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.009LDH . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.053Potassium. . . . . . . . . . . . . . . . . . . . . . . 0.015SGOT. . . . . . . . . . . . . . . . . . . . . . . . . . . 0.031SGPT . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.040Sodium . . . . . . . . . . . . . . . . . . . . . . . . . 0.017Total bilirubin . . . . . . . . . . . . . . . . . . . . 0.020Total protein . . . . . . . . . . . . . . . . . . . . . 0.005Triglycerides. . . . . . . . . . . . . . . . . . . . . 0.254BUN . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.004Uric acid . . . . . . . . . . . . . . . . . . . . . . . . 0.024

ConsumablesCalibrators and controls . . . . . . . . . . . 0.332Other consumables . . . . . . . . . . . . . . . 0.430Total reagent cost per sample . . . . . . $0.91a

Total reagent and consumable costper samplea. . . . . . . . . . . . . . . . . . . . $1.67

Total reagent and consumable costper determination. . . . . . . . . . . . . ~. $0.08

$0.0220.0860.0090.0110.0110.0670.3220.0040.0230.1060.0130.0160.1470.0090.0410.2350.0090.0220.0090.3680.0050.051

0.1840.178

$0.90 b

$1.26

$0.10aASSurneS SMAC contains all tests except iron and direct billrubin.bAssumes SMA IZ60 Contains the following: albumin, alkaline Phosphatase,

calclum, CPK, creatlnlne, glucose, inorganic phosphorus, LDH, SGOT, totalbllirubln, total protein, BUN.

SOURCE: Personal communications and marketing materials from Technicon(46).


the SMA 12/60 to 0.3 cents per sample for cre-atinine on the SMAC. Consumables account forroughly one-half of variable costs per sample inthe SMAC and for about one-third of variablecosts per sample in the SMA 12/60. Reagentprices are lower for the SMAC than for theSMA 12/60, in part because of the smaller re-agent volumes required for the SMAC.

Hycel’s reagents and consumables are in-cluded in the price per sample under their main-tenance plan for the Super-Seventeen, describedabove. Alternatively, they can be purchasedseparately at prices ranging from 10 cents pertest for inorganic phosphorus to 23 cents per testfor alkaline phosphatase, except for triglycer-ides, which are priced at 83 cents per test. Notethat the Hycel analyzer, unlike Technicon’scontinuous-flow analyzers, consumes only thereagents for the tests requested.

Reagents and consumables for the Du PontACA are available on an individual basis forthose who own the analyzer, but the “full-serv-ice” price of $1.13 per test includes reagents andconsumables as well as equipment use and serv-ice.

Summary of Direct Nonlabor Costs.—On thebasis of the data presented so far, the fixed andvariable components of direct nonlabor cost canbe calculated for a selection of instruments,service and use plans, and test configurations.Examples are shown in table 4.

Direct Labor Costs

Economic analyses of automated chemistrytests have consistently found that direct laborcosts account for 40 to 60 percent of the totaldirect costs (30,32,35). McLaughlin, in a 1971study at a hospital with a 12-channel analyzer,found the direct cost per determination to be 28cents, or 51 percent of total direct costs (30).This would be 46 cents at 1979 price levels.Pegels, in a study based on 1971 data, foundaverage direct labor costs of 21 cents per test, or62 percent of total direct costs for operating a20-channel analyzer (35).

The effect of increased analyzer speed onoverall labor costs has not been studied in recentyears. However, McLaughlin’s study suggests

Table 4.—Direct Nonlabor Costs forSelected Analyzers

Annual Variable costAnalyzer fixed cost Per sampleTechnicon SMACa with

service plan. ... ... ... ... .. .$70,000Technicon SMA 12/60a with

service plan. . ...............24,400Hycel Super-17 with service plan,

full profiles only< 12,000 samples/year. . ......12,40012,001-24,000 samples/year. .. .27,40024,001-36,000 samples/year. .. .44,200>36,000 samples/year. . ......55,000

Hycel Super-1 7 with service plan,individual testsb< 12,000 tests/year. . .........12,40012,001-120,000 tests/year. . ....12,760120,001-228,000 tests/year. . ...17,560>228,000 tests/year. . . . . . . . . .24,400

Du Pont ACA II with rental plan. . . 0

$1.67

1.26

3.001.751.050.75

0.22 x NC

0.19 x N0.15 x N0.12 x N1.13 x N

aA5sumes test configurations given in legend to table 3.bApproximation of Hycel’s block-rate Pricin9.CN = average rlurnber of determinations per SamPle.

SOURCE: Personal communications and marketing materials from the man-ufacturers (10,24,46).

that labor accounted for approximately 50 per-cent of direct costs both in hospitals with andwithout multichannel analyzers (30). The im-portant effect of automation, however, mayshift to certain labor costs from variable to fixed(4,35). Some labor costs (e.g., for specimen col-lection and coding) remain variable, irrespectiveof the method of analysis, and are relativelyunaffected in magnitude (4,28). Others (e. g.,operation) become fixed in the short run. Thecosts of some fixed components of labor (e. g.,supervision and training) may be expected to in-crease (28,30).

There are insufficient recent data from institu-tions with automated analyzers on which tobase quantitative estimates of the effect ofmultichannel technology on labor costs. Qual-itatively, for analyses such as those presentedimmediately below, it may be reasonable toassume that: 1) the variable component of laboris proportional to the number of samples tested(not determinations), and 2) the fixed andvariable components are roughly in the sameproportion to each other as are the fixed andvariable components of nonlabor direct costs,The second assumption reflects the observationthat supervision and training requirements tendto increase as a function of the complexity of in-


strumentation (and, hence, fixed capital andmaintenance costs). Studies of laboratory pro-ductivity are needed to validate these assump-tions and to provide more reliable quantitativeestimates.

Indirect Costs

No recently published data are available onindirect costs of clinical chemistry laboratories.Moran suggests a rule of thumb that indirectcosts (space, administration, utilities, etc. ) mayadd 25 percent to direct costs (32). In any event,it is unlikely that multichannel analyzers havehad much impact on indirect costs, compared toother automated analyzers or even manualmethods.

Hypothetical Analyses of Efficiency

The available data on labor costs are inade-quate for performing realistic analyses, but it ispossible to suggest the kinds of analyses that canbe used to evaluate the need for automatedmultichannel equipment as a function of thenumber of tests required per serum sample,

For purposes of these hypothetical analyses,we assume that the fixed and variable com-ponents of labor are each equal to the fixed andvariable components of nonlabor direct costsfor each equipment configuration. This assump-tion is consistent with the finding that laborcosts consistently account for about half of totalcosts. The variable component of labor isassumed to be proportional to the number ofsample aliquots that the technologist mustprepare.

The following analysis is based on the costdata for the SMA 12/60. The question is: Howlarge an annual demand for tests (determina-tions) is required to justify purchase of an SMA12/60? Unlike some previous analysts who haveaddressed this question in the past, we adopt thesocietal interest in minimizing true cost, not thelaboratory director’s interest in maximizing netrevenues.

Suppose that the average number of deter-minations required per sample is two. (This isconsistent with the data of Taylor, cited above(44). ) The direct fixed costs for the analyzer are

assumed to be $24,400 for nonlabor items (seetable 4), plus $24,400 for the fixed component oflabor. The direct variable costs are $1.26 p e rsample, or $0.63 per determination, plus $0.63per determination for labor. Hence, total cost(TC) is given by the expression:

where N is the number of determinations. Aver-age cost per determination is equal to:

As shown in figure 5, the average cost declinesas the number of determinations increases.

Suppose that the next best alternative methodhas a unit cost of $1.50 per determination. (Thisfigure is consistent with data from Pegels, ad-justed for inflation (35). ) Then, as shown infigure 5, the analyzer is efficient, provided thatat least 200,000 tests on 100,000 samples areperformed per year.

If, instead of two determinations per sample,four per sample were requested, then the break-even volume would be only 56,100 determina-tions on only 14,000 samples. Thus, we con-

Figure 5.—Hypothetical “Breakeven” Analysis for aMultichannel Analyzer

Averagecost perdetermination$3.00

$2.OCI

$1.50

$1.26$1.00

$0.63

56,100 100,000 200,000 300,000

Number of determinations per year


elude the efficiency of multichannel analyzers atlower test volumes is very sensitive to thenumber of determinations performed per sam-ple. Their cost effectiveness, therefore, dependscritically on the benefits to be derived fromthose additional determinations.

Fixed Costs of Production andthe Rate of Innovation

From societal perspective, the price of equip-ment to the purchaser overstates the actual in-cremental cost of production. It includes asignificant component that reflects the fixed,hidden costs of product research, development,and marketing (22). For example, developmentcosts for Technicon’s SMAC alone were $7million (34). Allocated among 1,000 machinessold, on average, 5 years later, and assuming anannual interest rate (not inflation-adjusted) of10 percent, these costs alone would account for$11,400 in the $271,000 purchase price ofSMAC. Furthermore, as Holy points out, prod-uct development is but one of many costs in theprocess from product conception to marketing(22). Patent protection costs perhaps $100,000or more for a major product and several thou-sand dollars for each patented component; engi-neering documentation for production may cost$250,000. Perhaps the largest hidden costs arethose that a company must absorb for researchand development of ideas that never reach pro-duction (22).

From society’s viewpoint, it maybe appropri-ate to factor these elements into the cost of atechnology. But the question must be asked: Arethe resources devoted to innovation in multi-channel analyzer technology appropriatelyspent? Do the reductions in variable costs justifythe fixed costs that finance the rapid turnover incost-saving innovations? These issues deservestudy, and they apply to the cost-effectivenessevaluation of technologies ranging far beyondthe realm of clinical chemistry analyzers.

Induced Costs

Given the ability of multichannel technol-ogies to produce additional determinations atvery low incremental cost, perhaps their in-duced costs make their greatest economic im-

pact. These induced costs include repeat tests,additional laboratory tests required to rule outor confirm a diagnosis, radiographic or otherdiagnostic tests, and treatments.

Obviously, the magnitude of induced costsdepends on the particular tests ordered, theclinical procedures followed, and the popula-tion tested. In one study of hospital admissionscreening, use of a multichannel analyzer led toa 78-percent increase in the number of othertests performed, a 64-percent increase in otherlaboratory costs, and a 25-percent increase inthe number of consultations (9). Explicit deci-sion analyses have indicated that induced costscan easily account for half of all costs at-tributable to laboratory tests, even in popula-tions in which the proportion of patients with acondition requiring intervention is small (51).

The high rate of false positives in multichan-nel test panels can amplify this effect (12). Thus,even if the cost per determination can be re-duced, one must consider the effect of multi-channel technology on total costs, and askwhether these increases in “rule-out” costs arejustified by the benefits. Clearly, further studiesare needed in which the magnitude of inducedcosts attributable to particular chemistry tests isestimated.

Financial Incentives

It would be inappropriate to conclude a dis-cussion of the economics of the multichannelchemistry analyzer without some discussion ofthe financial incentives faced by laboratorydirectors and other users of the technology.Typical third-party reimbursement rates forchemistry tests clearly exceed not only marginalcosts but also average costs for most labora-tories. If the sole concern of the laboratorydirector were to maximize net revenues, the ob-vious incentive would be to increase the numberof tests, at low incremental cost but high in-cremental revenue (32).

Whether this incentive motivates laboratorydirectors in practice has not been shown, butthat it does motivate them certainly seemsplausible and is generally accepted as fact. Onan aggregate level, society seems willing to per-


mit hospitals to use laboratory revenues to sub-sidize other meritorious, but low-income, serv-ices. The result, of course, has been to makemultichannel analyzers even more attractive,because their costs per test decrease with in-creasing test volumes.

Other subtler financial and organizational in-centives may also contribute to the adoption ofmultichannel analyzers. Mather points out thatmany hospital laboratories with only moderatevolumes may purchase advanced multichannelequipment to shift costs from labor to capital,thus relieving pressure from the overburdenedlabor supply (28). This inefficiency can be ra-tionalized to the hospital administration if thelabor supply is relatively fixed and if addedrevenues can more than pay for the loss of effi-ciency.

The countervailing effects of governmentregulation of capital expenditures throughcertificate-of-need programs have been negligi-ble. No instance of denial for multichannelanalyzer purchase was found in the Statessurveyed by the authors. (At current pricelevels, only Technicon’s SMAC and the newHycel-M would be covered by such regulationsin any case. ) Other regulatory policies, in-cluding denial of reimbursement for certain usesof the laboratory (e.g. such as hospital admis-sion profiles) may ultimately alter the financialincentives to acquire multichannel chemistryanalyzers, but such measures are not likely tohave much effect for several years.

COST EFFECTIVENESS OF CARDIAC ENZYMES IN DIAGNOSIS OFMYOCARDIAL INFARCTION: A STRUCTURED REVIEW

Overview

The analytical framework for cost-effective-ness analysis (CEA) of multichannel chemistryanalyzers presented earlier strongly suggests theneed for cost-effectiveness evaluations of the in-dividual chemistry tests that constitute testpanels. Such evaluations should consider notonly the quality of measurement and diagnosticefficacy of such tests, but also the expectedclinical value of the information obtained, thecosts induced or averted as a result of the tests,and the costs of producing various quantities oftest results on different chemistry analyzers. Fora full evaluation of the cost effectiveness ofautomating a particular test, according to themethodology presented earlier, each clinical in-dication for its use must be considered sepa-rately before proceeding to an overall eval-uation of the test.

In this part of the case study, we present astructured review of the information required toconduct such CEAs for three chemistry tests thatare commonly used to diagnose myocardial in-farction (heart attack). These tests are the car-diac enzymes: 1) creatine phosphokinase (CPK),

2) lactic dehydrogenase (LDH), 3) and aspartateaminotransferase, also known as serum glu-tamic oxaloacetic transferaese (SGOT). Theyare all available on the multichannel analyzersreviewed in this study. We also review informa-tion pertaining to the use of certain isoenzymesof these compounds. Some of these isoenzymeshave been widely acclaimed for their diagnosticvalue in myocardial infarction, and some arebeing considered for use in improved multichan-nel analyzers. One (CPK-MB) is now availableon a limited basis in Du Pont’s ACA.

It should be emphasized that our review of thecardiac enzymes and isoenzymes pertains onlyto a single clinical use of these tests, namely,diagnosis (i.e., “rule-in/rule-out”) of myocar-dial infarction in patients with symptoms sug-gesting it. A complete evaluation of these testswould consider other uses as well, including, forexample, the use of LDH and SGOT to diagnoseliver disease. (Tests for CPK, however, areprobably performed almost exclusively in thediagnosis of myocardial infarction or screeningfor cardiac damage. )

The clinical properties of these enzymes and


isoenzymes are briefly reviewed in the first sec-tion below. In the next section, we structure aclinical decision tree for the typical “rule-in/rule-out” decision problem. We then proceedin the subsequent sections to evaluate theavailable evidence at each of the four levels ofanalysis described in the part of this case studyon the framework for cost-effectiveness evalua-tion: 1) technical quality of measurement, 2)diagnostic value, 3) clinical efficacy, and 4) costeffectiveness. Our review concludes with a briefdiscussion of efficiency.

Clinical Properties of theCardiac Enzymes11

Elevation of the serum level of either LDH,SGOT, or CPK suggests tissue damage, but notnecessarily damage to the muscles in the heartwall (myocardium). Cardiac damage usually,but not always, causes all three enzymes to rise.LDH, however, may also be elevated in thepresence of liver disease (such as hepatitis),pulmonary infarction, renal disease, musculardamage, shock, leukemias and lymphomas, andother cancers. SGOT may be elevated due toliver disease, skeletal muscle damage, braindamage, kidney disease, or shock. CPK may beelevated due to damage to skeletal muscles orthe brain; it rises easily in the presence of evenmild skeletal trauma, such as that followingphysical exercise. All three may be elevated dueto congestive heart failure or trauma of thoracicsurgery.

All three enzyme tests are typically ordered inthe workup for myocardial infarction, either inthe coronary care unit to determine whether toreturn the patient to a regular hospital bed (orpossibly to discharge the patient), or in theemergency room to decide whether to admit thepatient to the hospital. The three tests are usual-ly ordered together because of their individuallack of specificity. If all three are positive, ithelps to confirm the diagnosis, The enzyme testsare usually accompanied by an electrocardio-gram (EKG), which is considered a very specifictest if the classical signs of myocardial infarction

I I IJn]es\ (lt herwlse nt~ted, most (d the bdck~r(wnd md terld I t(~r

this secti(;n IS based (In a review by Rapapt}rt (36,37).

are found. An equivocal or negative EKG, how-ever, is usually not considered sufficient evi-dence to rule out a myocardial infarction, andthe enzymes and patient history become theprincipal diagnostic tools.

If the tests are given too soon or too latefollowing a myocardial infarction, they are like-ly to produce more false negatives than if givenat the proper time intervals. LDH elevationstypically appear within 24 hours of the myocar-dial infarction, peak between 48 and 72 hours,and return to normal 7 to 10 days later. SGOTelevations run a shorter course, beginningwithin 8 to 12 hours, peaking at 36 to 48 hours,and returning to normal 4 to 5 days after themyocardial infarction. CPK follows the shortestcourse of all. The elevations first appear inabout 6 hours, peak around 24 hours, and re-turn to normal levels in 3 to 4 days. Thus, a pa-tient whose enzymes are measured within a fewhours of the myocardial infarction is likely toshow a normal LDH. A patient whose enzymesare measured 48 hours later may show a normalCPK. For this reason, these enzymes are usuallymeasured for at least 3 to 4 days before a diag-nosis is made. This practice, of course, can becostly, since it means prolonged hospitalizationfor many patients who will turn out not to havesuffered a myocardial infarction.

Isoenzymes are slightly different molecularforms of an enzyme, all of which catalyze thesame reaction. The difference between isoen-zymes is important because the composition oftotal enzyme levels differs among body tissues.Some isoenzymes or isoenzyme profiles arequite specific for particular organs (e.g., heartor liver), and therein lies their diagnostic value.Methods for separating isoenzymes usually takeadvantage of their physical properties, andseparation is often accomplished by electroph-oresis. Measurement methods based on radio-immunoassay and on calorimetry have alsobeen developed, and the latter are the basis fordevelopments toward automation in multichan-nel analyzers.

The isoenzymes most valuable in the diagno-sis of myocardial infarction are the isoenzymeCPK-MB and the enzymes LDH1 and LDH 2.CPK has three isoenzymes: CPK 1 (or CPK-BB),


C P K2 (or CPK-MB), and CPK3 (or C P K - M M ) .CPK-MB is found only in heart tissue. Itspresence in serum is thought to be a very spe-cific indicator of cardiac damage. However, car-diac damage is not synonymous with myocar-dial infarction. Congestive heart failure, severeangina, myocardial ischemia, and cardiac sur-gery or chest trauma can also cause CPK-MBlevels to rise in the blood, as can certain formsof muscular dystrophy, polymyositis, or signifi-cant myoglobinuria. While CPK-MB is quitespecific for myocardial infarction, its clinicalsensitivity is reduced by the time course of theelevation, typically less than that for total CPK.

LDH has five isoenzymes. Those that are use-ful in the diagnosis of myocardial infarction areLDH 1 and LDH2. In particular, if the serum levelof LDH1 exceeds the level of LDH2, then thenumber of diagnostic possibilities is reducedconsiderably. This “flipped LDH pattern” (socalled because in normal serum the LDH2 levelexceeds that of LDH1) may be observed in pa-tients with a myocardial infarction, but also inthose with acute renal infarction or hemolysis.

Figure 6.—Structure of Decision Analysis for Evaluation

However, since elevated CPK-MB and flippedLDH have only one cause in common—myocar-dial infarction—the combination is extremelyspecific. Some enthusiasts claim that 80 percentof “rule-in/rule-out” tests for myocardial infarc-tion cases can be resolved with virtual certaintyon the basis of CPK and LDH isoenzyme results(15). These authors suggest that SGOT wouldhave virtually no clinical value in the diagnosisof myocardial infarction if CPK and LDH isoen-zymes were routinely available (15).

Structure for Decision Analysis

The structure of the decision whether to orderenzymes and/or isoenzymes for the patient withsuspected myocardial infarction is diagramed inthe decision tree shown in figure 6. The in-formation required to complete the analysisconsists of data on: 1) probabilities at chancenodes, and 2) costs and health outcomes at theend of each path.

The tree begins at the tar left with the order-ing of an EKG. Next, at ❑ , is the key decision

of Enzymes in Diagnosis of Myocardial Infarction (Ml)

EKG I

—.——— — —— —. ——.— ——..——


under analysis: what combination of enzymetests to order. At the first probability node, @,the EKG results are obtained. At the next prob-ability node, ~, the enzyme (and isoenzyme)results are obtained. The probability of any par-ticular combination of test results depends,among other factors, on: 1) the conditionalprobability of this set of values, given the trueenzyme levels that would be measured by a per-fectly accurate and precise test; 2) the condi-tional probabilities of combinations of enzymelevels, given the presence or absence of amyocardial infarction; and 3) the prevalence ofmyocardial infarction in patients with similarhistories and EKG results. This tree structureassumes that the EKG results are not availableat the time of the decision whether to order en-zymes; in some clinical situations, the order ofnodes ❑ and @ may be reversed.

Next comes a decision, at node ❑ : whetherto treat for or to rule out a myocardial infarc-tion. The value of the tests, presumably, lies intheir ability to inform this decision-and, inturn, in the value of a correct diagnosis ofmyocardial infarction, if present, or correct rul-ing out of myocardial infarction, if absent. Theconsequences of false negatives (inappropriaterule-out ) may be added risk of complications oreven death if the patient does not get proper bedrest or if the complication occurs outside thecoronary care unit. The consequences of falsepositives (inappropriate rule-in) may be pri-marily economic costs in the form of prolongedhospitalization or prolonged in the coronarycare unit.

Next, for purposes of analysis, estimates areneeded of the probabilities of the true enzymelevels, given the measured values (a function oftest precision and accuracy). These are incorpo-rated at probability node @. Given the enzymelevels, the next probability node, @, indicateswhether myocardial infarction was present ornot. Finally, the clinical outcome unfolds at ~,given the treatment decision at ❑ and the truediagnosis as revealed at@. The final outcomesare grouped, for purposes of cost-effectivenessevaluation, into two categories: health outcome(mortality and morbidity) and resource cost (fordiagnosis and treatment ).

The decision tree in figure 6 provides a struc-ture for the review that follows in the remainingpages of this case study. The discussion of en-zyme test measurement relates to the probabil-ities required at node @, the conditional prob-abilities of true e n z y m e levels, given themeasured values. The discussion of diagnosticvalue relates to the test result probabilities at

onode B and their relation to the conditionaldisease-state probabilities at node@, the latterbeing the predictive values of the test results(e.g., the probability of myocardial infarction,given a particular set of test values). The discus-sion of clinical efficacy concerns the probabil-

$7ities at node G in relation to both the treatmentdecision at and the true condition as re-vealed at @: Does intervention make a dif-ference? The final section considers the cost datato be associated with each path of the decisiontree and discusses the cost-effectiveness implica-tions of altering the sensitivity and/or thespecificity of the diagnostic workup by the addi-tion or deletion of enzyme and isoenzyme deter-minations.

Quality of MeasurementMeasurement errors for calorimetric determi-

nations of CPK, SGOT, and LDH have beenfound to lie well within the bounds recom-mended by the College of American Patholo-gists, at least if the tests are carried out underideal laboratory conditions. The coefficients ofvariation reported, however, depend on thelaboratory.12 An in-house evaluation by Techni-con found coefficients of variation of approx-imately 5 percent for all three enzymes, with ex-cellent linearity of response over the relevantrange of values (41). Other studies havedetected reliability problems in automatedmeasurement of SGOT, however, with coeffi-cients of variation closer to 10 percent (2,12).Even more serious than intralaboratory varia-tion is interlaboratory variability, which hasbeen reported to be as high as 25 percent forSGOT (12). Despite these caveats, however, theability of existing technology to provide reliable

‘: The c{>etllctent c~t variation in a sample of measurements is therat[[~ (~t the sample standard deviation to the sample mean.


measurements of these enzymes is generallyacknowledged.

Data on test precision and accuracy areavailable from the manufacturers for each oftheir instruments and can provide estimates ofthe probabilistic relations between measuredand true values. These data may be viewed,however, only as lower bounds on variability inactual practice. Studies of laboratory perform-ance are needed to provide more realistic prob-ability estimates. Such data can be obtainedfrom quality-control studies, which are per-formed by all laboratories under Federal law.

As for the quality of isoenzyme measurement,coefficients of variation for immunologic meth-ods have been reported to be quite acceptable—approximately 5 to 10 percent for CPK-MB.Other, notably calorimetric, methods have beenfound to be less reliable (6). The new methodused by Du Pont on its ACA, however, is saidto be as precise as the best available manualmethods, with coefficients of variation wellunder 10 percent (10). One problem in meas-uring CPK-MB is storage of samples, sinceCPK-MB may undergo chemical change at high-er temperatures, leading to false negatives (re-duced sensitivity) (33).

Overall, data relating to quality of measure-ment are readily obtainable. Published data todate suggest that measurement error is not a ma-jor consideration in the overall evaluation ofcardiac enzymes.

Diagnostic Value

Several studies report estimates of the sen-sitivity and specificity of the cardiac enzymesand isoenzymes in the diagnosis of myocardial

infarction. These results, based largely onhospital patients, are summarized in table 5.With a few exceptions, the results are quite con-sistent across studies. These data support theview that CPK, LDH, and SGOT are rather sen-sitive, but only moderately specific. In com-bination, considering a test positive only if allthree are positive, the specificity improves con-siderably, but at some loss of sensitivity.

CPK-MB and the flipped LDH do appear tobe rather specific, but probably not perfectly so.No data on the joint sensitivity and specificityof the isoenzymes were available, except for thestudy by Galen, et al. (15), the results of whichare given in table 6. They find that CPK-MB ele-vation and flipped LDH never occur together inthe absence of myocardial infarction and thatCPK-MB is always elevated in its presence.However, CPK-MB elevation and flipped LDHboth occur in only 80 percent of myocardial in-farction cases, and CPK-MB is elevated in 15percent of nonmyocardial infarction cases.

The data reported in the literature are defi-cient in several respects. First, there are fewestimates of probabilities of combinations of en-zyme test results, Such data would be needed toevaluate the incremental diagnostic value of onetest (e.g., SGOT), given that other tests arealready available. It would be inappropriate toassume that the tests are conditionally inde-pendent, since many of the causes of serum en-zyme elevation are not unique to a single en-zyme. Assuming conditional independencewould tend to overestimate the incrementaldiagnostic value of any single test.

A second, and more serious, deficiency in thedata is that there is no independent definition of

.— . .


Table 6.—ResuIts of Galen, et al., on JointOccurrence of Elevated CPK-MB Isoenzyme and

Flipped LDH Isoenzymes

Probability, Probability y,Finding given Ml given no Ml

CPK-MB +, flipped LDH . . . 0.8 0CPK-MB +, normal LDH . . . 0.2 0.15CPK-MB – . . . . . . . . . . . . . . . 0 0.85

SOURCE: R. S. Galen, et al., ‘(Diagnosis of Acute Myocardial Infarction: RelativeEfficiency of Serum Enzyme and Isoenzyme Measurements,”J.A M.A. 232:145, 1975.

myocardial infarction. In most studies, the testsbeing evaluated are also used to define the pres-ence or absence of disease. This can lead to twoscenarios, both of which lead to bias in the esti-mation process. One procedure is to restrict thepopulation under study to those with over-whelming evidence of myocardial infarction orno myocardial infarction, either as confirmedby a positive EKG finding or as ruled out by theabsence of enzyme elevations other than the oneunder study. Such a procedure overestimatestest sensitivities and specificities by restrictingthe study population to those who may be mostor least likely to show signs of myocardialinfarction-the equivocal cases are thus ex-cluded. The other procedure reported by someinvestigators is to use the enzyme values them-selves to classify patients as having myocardialinfarction or not. This obvious circularity alsocreates bias in the estimation process and willtend to overestimate sensitivities and specifici-ties if the enzyme levels are conditionally de-pendent with positive correlations.

The resolution of this problem of evaluatingthe diagnostic value of tests which themselvesprovide the only basis for defining the diseasestate remains a topic of methodologic research.One promising approach may be to bypass theevaluation of diagnostic value, which is reallyonly a way-station in the overall evaluationprocess, and to proceed directly to measures ofclinical efficacy in terms of prognostic value andpatient outcome. For example, one might esti-mate the survival curves and morbidity rates forpatients with a particular pattern of test results.The rates would be conditional on the treatmentalternative (e. g., hospitalized or not), but ir-respective of any attempt to classify the patientas having had a myocardial infarction or not.

The incremental value of any enzyme test wouldbe measured in terms of its expected informa-tion value (calculated by averaging out anappropriate decision tree). This value would de-pend on the consequences to the patient of mak-ing the right or wrong treatment decision, interms of patient outcomes, rather than in termsof the test’s ability to classify patients intodiagnostic categories.

A third limitation of the available data is that,for the most part, the data ignore the question ofdefining a positive or abnormal enzyme level. Inactuality, enzyme levels are measured on a con-tinuum, and the choice of a positivity criterionis itself a decision. By selecting a stringentcriterion, one can gain specificity but at a loss ofsensitivity; with a lax criterion, one gains sen-sitivity but loses specificity. In order to includethese considerations in the analysis, one wouldneed data on the probability distributions ofvalues for each enzyme individually and in com-bination for patients with and without myocar-dial infarction, Data on the joint distribution ofresults are not available in the literature,although some data for individual tests havebeen reported (14).

The final missing datum needed for evalua-tion of diagnostic value is the prevalence ofmyocardial infarction in the population tested.As argued earlier, the prevalence, or prior prob-ability of disease, has an important effect on thepredictive value of a test result. Several studiesagree that for patients hospitalized with symp-toms suggesting myocardial infarction, theprevalence of myocardial infarction is approx-imately 50 percent (13,18). In patients withequivocal symptoms, who would be tested ifcardiac enzymes were included in routinescreening of hospital patients, the prevalence ofmyocardial infarction in the population testedwould be much lower. Therefore, the predictivevalue positive would be lower than it would beamong those with symptoms of myocardial in-farction. Moreover, patients with myocardialinfarction detected by screening would be morelikely to have uncomplicated cases which wouldbe less likely to benefit from prolonged hospital-ization than would symptomatic cases (29).


Clinical Efficacy

The final piece of information on which theultimate value of cardiac enzyme tests reliesconcerns the value of a correct diagnosis.Specifically, what are the consequences of “rul-ing out” a patient with myocardial infarction?What are the consequences of “ruling in” a pa-tient without it? Since the consequences of afalse “rule-in” (false positive) are largely eco-nomic, we defer most of the discussion of thelatter question to the next section of this casestudy. It should be noted, however, that an er-roneous “rule-in” may have the adverse conse-quence of diverting the physician’s attentionfrom the investigation of other, treatable causesof the patient’s symptoms. The probability thatsuch a treatable condition might be presentwould depend on the nature of the symptomsand the rest of the patient’s history.

The consequences of a false “rule-out” (falsenegative) depend on the value of interventions,which may range from care in a coronary careunit, to hospitalization with bed rest, to bed restat home. The value of bed rest, whether at homeor in a hospital, during the first few days follow-ing a myocardial infarction, when the risk ofventricular fibrillation is greatest, is widelyacknowledged. It is less clear whether bed restcan reduce the subsequent risks of reinfarctionor arrhythmia and sudden death.

If bed rest were the only intervention ofvalue, then the consequences of a false “rule-out” might be that the patient would not benefitfrom the reduction in risk afforded by that in-tervention. However, for patients with severechest pain, or other symptoms bringing them tothe hospital, bed rest is often the treatment ofchoice for other conditions they may have, suchas acute ischemia. The intervention that hashistorically been specific for myocardial infarc-tion is hospitalization, and in recent years,hospitalization in a coronary care unit. But theconsequences of not hospitalizing a patient withmyocardial infarction are a matter of some con-troversy, and therein lies the problem in eval-uating the clinical efficacy of the enzyme testsused in the diagnostic workup.

A complete review of the evidence on the ef-fects of prolonged hospitalization and intensivecare units on mortality and morbidity of pa-tients with myocardial infarction would con-stitute a case study in itself. A review of theliterature reveals numerous studies, many ofthem British, indicating that home care is atleast as effective as hospital care followingmyocardial infarction, and that shorter hospitalstays are at least as effective as longer stays(21,29). Despite this evidence, the consensus inthe United States is now at about 10 to 12 dayshospitalization, possibly preceded by 3 to 7days in the coronary care unit. Objective datasupporting this practice are not available,however.

It is on this perceived benefit, measured as thedifference in risk for myocardial infarction pa-tients who are hospitalized and those who arenot, that the clinical value of the diagnosticworkup for myocardial infarction, including theenzymes, rests. If this difference were zero, theinformation value of the cardiac enzymes wouldbe zero (excluding their purely prognostic value,independent of effects on health outcomes). Theuncertainty surrounding this issue is the majorobstacle to evaluation of the clinical value ofcardiac enzyme and isoenzyme determinationsin the diagnosis of myocardial infarction.

Cost Effectiveness

The diagnostic value of the cardiac enzymeslies in their ability to improve the predictivevalue in diagnosing myocardial infarctions.This involves reducing either the false-negativeor false-positive rate,

Reducing the false-negative rate is tanta-mount to improving test sensitivity. If the car-diac enzymes are as sensitive as the literaturereviewed above suggests, then their cost effec-tiveness would depend on the clinical value ofidentifying and treating a patient with myocar-dial infarction, compared with the added cost ofdiagnosis and treatment. As noted earlier, how-ever, the health benefits from treatment in ahospital remain in doubt. Hence, the costs of thetests themselves, and the induced costs of hospi-

—— . . .


talization of patients “ruled-in,” must beweighed against the hypothetical benefit of in-tervention.

Reducing the false-positive rate is tanta-mount to improving test specificity; it is in thisdomain that isoenzymes are said to have thegreatest value. The value of improving the cor-rect “rule-out” rate is largely economic; manypatients may be spared unnecessary hospitaliza-tion, with ensuing savings both to society and toindividual patients.

A study by Gerber, et al. (18) estimates thenet economic savings that may be expected if thediagnostic test battery for myocardial infarctionis expanded to include routine CPK-MB andLDH isoenzymes. Even imputing a cost as greatas the institution’s charge for these tests ($23 foreach isoenzyme, $7 for each total enzyme, and$21 for an EKG) and assuming a complete work-up for each of 4 days in the hospital, the costsavings resulting from earlier discharge of theadditional “rule-outs” easily outweigh the diag-nostic testing costs (18). This appears to be acase where the induced savings in treatmentcosts exceed the direct costs of testing.

Gerber’s analysis, however, assumes that thecurrent practice of hospitalizing patients unlessthere is strong evidence against myocardial in-farction will prevail. Given that state of affairs,whether clinically justified or not, the evidenceseems to suggest that if physicians trust thespecificities of EKGs and enzyme and isoenzymetests, their net economic effect in symptomaticpatients who are candidates for hospitalizationmay actually be to reduce costs. A limitation ofthe Gerber study for purposes of evaluating theenzymes and isoenzymes individually, how-ever, is that the incremental contributions of theenzymes and isoenzymes to the overall cost sav-ings were not reported. It could be that the EKGalone accounted for most of the savings. Thisstudy does suggest, however, that this would bea promising area for further research.

The net economic effect of enzyme determina-tions, if applied to patients asymptomatic for

myocardial infarction, as distinguished fromacute cases, is undoubtedly to increase costs,while the corresponding health benefit is ques-tionable. Therefore, the overall cost effec-tiveness of this use of LDH, SGOT, CPK, andthe isoenzymes rests on evidence not yet avail-able on the health benefits of intervention in thenatural course of myocardial infarction.

The overall cost effectiveness of automatingany test must be based on an assessment of all ofits principal clinical uses. It is possible that theuses of CPK, SGOT, and LDH in the diagnosisof myocardial infarction alone can justify auto-mating these tests. Ifto examine as wellcosts associated withthe use of LDH andliver disease.

Efficiency

not, then one would havethe benefits and inducedother clinical uses, such asSGOT in the diagnosis of

Whether it is efficient, given a particular test-ordering policy, to reserve three analyzer chan-nels for cardiac enzymes may depend on wheth-er the machine to be used is a continuous-flowor discrete-sample analyzer. Reagent costs forCPK are among the highest of all tests (see table3). Therefore, if CPK is requested for a suffi-ciently small proportion of samples, the costs ofreagents consumed by continuous-flow analyz-ers could be considerable. For discrete-sampleanalyzers, this would not be a concern becausereagents would be consumed only for those testsactually ordered. However, the fixed costs asso-ciated with a CPK channel and its maintenancewould apply to both kinds of analyzers. Wheth-er a continuous-flow multichannel analyzer, adiscrete-sample multichannel analyzer, or sin-gle-channel equipment is the most efficientmethod of producing CPK, LDH, and SGOT re-sults depends on the pattern of test ordering,and this pattern, in turn, would optimally de-pend on an assessment of the effectiveness andinduced costs for each of their diagnostic uses.

cost effectiveness of automated multichannel chemistry

Documents